def __init__(self, **args):
UC = unifiedConfiguration()
self.default = {
'no_JL': {
'default': False,
'action': 'store_true',
'help': 'Do not get the job logs'
},
'no_CL': {
'default': False,
'action': 'store_true',
'help': 'Do not get the condor logs'
},
'all_errors': {
'default': False,
'action': 'store_true',
'help': 'Bypass and expose all error codes'
},
'cache': {
'default':
0,
'type':
float,
'help':
'The age in second of the error report before reloading them'
},
'expose': {
'default': UC.get('n_error_exposed'),
'help': 'Number of logs to retrieve',
'type': int
},
'not_from_eos': {
'default': False,
'action': 'store_true',
'help': 'Do NOT retrieve from job logs from eos'
},
'threads': {
'default': 5,
'type': int,
'help': 'The number of parallel workers to get reports'
},
'log_threads': {
'default': 3,
'type': int,
'help': 'The number of parallel workers to get logs per report'
},
}
for opt, defv in self.default.items():
setattr(self, opt, args.get(opt, defv['default']))
def parse(options):
RDI = remainingDatasetInfo()
UC = unifiedConfiguration()
spec_site = filter(None, options.site.split(','))
## fetching global information
locks = [
l.item.split('#')[0]
for l in session.query(Lock).filter(Lock.lock == True).all()
]
waiting = {}
stuck = {}
missing = {}
si = siteInfo()
sis = si.disk.keys()
random.shuffle(sis)
n_site = options.nsites
i_site = 0
threads = []
for site in sis:
if spec_site and not site in spec_site:
continue
space = si.disk[site]
if space and not spec_site:
continue
if n_site and i_site > n_site:
break
i_site += 1
print site, "has", space, "[TB] left out of", si.quota[site]
threads.append(
SiteBuster(site=site,
UC=UC,
RDI=RDI,
SI=si,
locks=copy.deepcopy(locks),
waiting=copy.deepcopy(waiting),
stuck=copy.deepcopy(stuck),
missing=copy.deepcopy(missing),
options=copy.deepcopy(options)))
run_threads = ThreadHandler(threads=threads,
label='Site Threads',
n_threads=5,
start_wait=0,
timeout=None,
verbose=True)
run_threads.run()
def parse( options ):
RDI = remainingDatasetInfo()
UC = unifiedConfiguration()
spec_site = filter(None,options.site.split(','))
## fetching global information
locks = [l.item.split('#')[0] for l in session.query(Lock).filter(Lock.lock == True).all()]
waiting = {}
stuck = {}
missing = {}
si = siteInfo()
sis = si.disk.keys()
random.shuffle( sis )
n_site = options.nsites
i_site = 0
threads = []
for site in sis:
if spec_site and not site in spec_site:
continue
space = si.disk[site]
if space and not spec_site:
continue
if n_site and i_site>n_site:
break
i_site += 1
print site,"has",space,"[TB] left out of",si.quota[site]
threads.append( SiteBuster( site = site,
UC = UC,
RDI = RDI,
SI = si,
locks = copy.deepcopy(locks),
waiting = copy.deepcopy(waiting),
stuck = copy.deepcopy(stuck),
missing = copy.deepcopy(missing),
options = copy.deepcopy(options)
))
run_threads = ThreadHandler( threads = threads,
label = 'Site Threads',
n_threads = 5 ,
start_wait = 0,
timeout = None,
verbose=True)
run_threads.run()
def completor(url, specific):
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
if use_mcm:
mcm = McMClient(dev=False)
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
wfs = []
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance')).all() )
## just take it in random order so that not always the same is seen
random.shuffle( wfs )
## by workflow a list of fraction / timestamps
completions = json.loads( open('%s/completions.json'%monitor_dir).read())
good_fractions = {}
timeout = {}
for c in CI.campaigns:
if 'force-complete' in CI.campaigns[c]:
good_fractions[c] = CI.campaigns[c]['force-complete']
if 'force-timeout' in CI.campaigns[c]:
timeout[c] = CI.campaigns[c]['force-timeout']
long_lasting = {}
overrides = getForceCompletes()
if use_mcm:
## add all workflow that mcm wants to get force completed
mcm_force = mcm.get('/restapi/requests/forcecomplete')
## assuming this will be a list of actual prepids
overrides['mcm'] = mcm_force
print "can force complete on"
print json.dumps( good_fractions ,indent=2)
print json.dumps( overrides, indent=2)
max_force = UC.get("max_force_complete")
#wfs_no_location_in_GQ = set()
#block_locations = defaultdict(lambda : defaultdict(list))
#wfs_no_location_in_GQ = defaultdict(list)
set_force_complete = set()
for wfo in wfs:
if specific and not specific in wfo.name: continue
print "looking at",wfo.name
## get all of the same
wfi = workflowInfo(url, wfo.name)
pids = wfi.getPrepIDs()
skip=False
if not any([c in wfo.name for c in good_fractions]): skip=True
for user,spec in overrides.items():
if wfi.request['RequestStatus']!='force-complete':
if any(s in wfo.name for s in spec) or (wfo.name in spec) or any(pid in spec for pid in pids) or any(s in pids for s in spec):
sendEmail('force-complete requested','%s is asking for %s to be force complete'%(user,wfo.name))
wfi = workflowInfo(url, wfo.name)
forceComplete(url , wfi )
skip=True
wfi.notifyRequestor("The workflow %s was force completed by request of %s"%(wfo.name,user), do_batch=False)
wfi.sendLog('completor','%s is asking for %s to be force complete'%(user,wfo.name))
if user == 'mcm' and use_mcm:
for pid in wfi.getPrepIDs():
mcm.delete('/restapi/requests/forcecomplete/%s'%pid)
#sendEmail('completor test','not force completing automatically, you have to go back to it')
#skip=False
break
if wfo.status.startswith('assistance'): skip = True
if skip:
continue
priority = wfi.request['RequestPriority']
if not 'Campaign' in wfi.request: continue
if not wfi.request['RequestStatus'] in ['acquired','running-open','running-closed']: continue
c = wfi.request['Campaign']
if not c in good_fractions: continue
good_fraction = good_fractions[c]
ignore_fraction = 2.
lumi_expected = None
event_expected = None
if not 'TotalInputEvents' in wfi.request:
if 'RequestNumEvents' in wfi.request:
event_expected = wfi.request['RequestNumEvents']
else:
print "truncated, cannot do anything"
continue
else:
lumi_expected = wfi.request['TotalInputLumis']
event_expected = wfi.request['TotalInputEvents']
now = time.mktime(time.gmtime()) / (60*60*24.)
running_log = filter(lambda change : change["Status"] in ["running-open","running-closed"],wfi.request['RequestTransition'])
if not running_log:
print "\tHas no running log"
# cannot figure out when the thing started running
continue
then = running_log[-1]['UpdateTime'] / (60.*60.*24.)
delay = now - then ## in days
(w,d) = divmod(delay, 7 )
print "\t"*int(w)+"Running since",delay,"[days] priority=",priority
monitor_delay = 7
allowed_delay = 14
if c in timeout:
allowed_delay = timeout[c]
monitor_delay = min(monitor_delay, allowed_delay)
### just skip if too early
if delay <= monitor_delay: continue
long_lasting[wfo.name] = { "delay" : delay }
percent_completions = {}
for output in wfi.request['OutputDatasets']:
if "/DQM" in output: continue ## that does not count
if not output in completions: completions[output] = { 'injected' : None, 'checkpoints' : [], 'workflow' : wfo.name}
## get completion fraction
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
lumi_completion=0.
event_completion=0.
if lumi_expected:
lumi_completion = lumi_count / float( lumi_expected )
if event_expected:
event_completion = event_count / float( event_expected )
#take the less optimistic
percent_completions[output] = min( lumi_completion, event_completion )
completions[output]['checkpoints'].append( (now, event_completion ) )
if all([percent_completions[out] >= good_fraction for out in percent_completions]):
wfi.sendLog('completor', "all is above %s \n%s"%( good_fraction,
json.dumps( percent_completions, indent=2 )
))
else:
long_lasting[wfo.name].update({
'completion': sum(percent_completions.values()) / len(percent_completions),
'completions' : percent_completions
})
## do something about the agents this workflow is in
long_lasting[wfo.name]['agents'] = wfi.getAgents()
wfi.sendLog('completor', "%s not over bound %s\n%s"%(percent_completions.values(), good_fraction,
json.dumps( long_lasting[wfo.name]['agents'], indent=2) ))
continue
if all([percent_completions[out] >= ignore_fraction for out in percent_completions]):
print "all is done, just wait a bit"
continue
for output in percent_completions:
completions[output]['injected'] = then
#further check on delays
cpuh = wfi.getComputingTime(unit='d')
ran_at = wfi.request['SiteWhitelist']
wfi.sendLog('completor',"Required %s, time spend %s"%( cpuh, delay))
##### WILL FORCE COMPLETE BELOW
# only really force complete after n days
if delay <= allowed_delay: continue
## find ACDCs that might be running
if max_force>0:
forceComplete(url, wfi )
set_force_complete.add( wfo.name )
print "going for force-complete of",wfo.name
wfi.sendLog('completor','going for force completing')
wfi.notifyRequestor("The workflow %s was force completed for running too long"% wfo.name)
max_force -=1
else:
wfi.sendLog('completor',"too many completion this round, cannot force complete")
## do it once only for testing
#break
if set_force_complete:
sendLog('completor','The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
#sendEmail('set force-complete', 'The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
open('%s/completions.json'%monitor_dir,'w').write( json.dumps( completions , indent=2))
text="These have been running for long"
open('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 ))
for wf,info in sorted(long_lasting.items(), key=lambda tp:tp[1]['delay'], reverse=True):
delay = info['delay']
text += "\n %s : %s days"% (wf, delay)
if 'completion' in info:
text += " %d%%"%( info['completion']*100 )
#if wfs_no_location_in_GQ:
# sendEmail('workflow with no location in GQ',"there won't be able to run anytime soon\n%s"%( '\n'.join(wfs_no_location_in_GQ)))
#sendEmail("long lasting workflow",text)
## you can check the log
print text
def equalizor(url , specific = None, options=None):
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
if not specific:
workflows = getWorkflows(url, status='running-closed', details=True)
workflows.extend(getWorkflows(url, status='running-open', details=True))
## start from scratch
modifications = defaultdict(dict)
## define regionality site => fallback allowed. feed on an ssb metric ??
mapping = defaultdict(list)
reversed_mapping = defaultdict(list)
regions = defaultdict(list)
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
for site in SI.sites_ready:
region = site.split('_')[1]
if not region in ['US'
,'DE','IT','FR',
'ES',
'UK' ### latest addition
]: continue
regions[region] = [region]
def site_in_depletion(s):
return True
if s in SI.sites_pressure:
(m, r, pressure) = SI.sites_pressure[s]
if float(m) < float(r):
print s,m,r,"lacking pressure"
return True
else:
print s,m,r,"pressure"
pass
return False
for site in SI.sites_ready:
region = site.split('_')[1]
## fallback to the region, to site with on-going low pressure
mapping[site] = [fb for fb in SI.sites_ready if any([('_%s_'%(reg) in fb and fb!=site and site_in_depletion(fb))for reg in regions[region]]) ]
for site in SI.sites_ready:
if site.split('_')[1] == 'US': ## to all site in the US
## add NERSC
mapping[site].append('T3_US_NERSC')
## add OSG
mapping[site].append('T3_US_OSG')
pass
#mapping['T2_IT_Rome'].append('T3_US_OSG')
#mapping['T1_US_FNAL'].append('T3_US_NERSC')
use_T0 = ('T0_CH_CERN' in UC.get("site_for_overflow"))
if options.t0: use_T0 = True
#if options.augment : use_T0 = True
use_HLT = ('T2_CH_CERN_HLT' in UC.get("site_for_overflow"))
if options.hlt: use_HLT = True
#if options.augment : use_HLT=True
if use_HLT:
mapping['T2_CH_CERN'].append('T2_CH_CERN_HLT')
if use_T0:
mapping['T2_CH_CERN'].append('T0_CH_CERN')
mapping['T1_IT_CNAF'].append('T0_CH_CERN')
mapping['T1_FR_CCIN2P3'].append('T0_CH_CERN')
mapping['T1_DE_KIT'].append('T0_CH_CERN')
## temptatively
#mapping['T0_CH_CERN'].append( 'T2_CH_CERN' )
## all europ can read from CERN
for reg in ['IT','DE','UK','FR','BE','ES']:
mapping['T2_CH_CERN'].extend([fb for fb in SI.sites_ready if '_%s_'%reg in fb])
pass
## all europ T1 among each others
europ_t1 = [site for site in SI.sites_ready if site.startswith('T1') and any([reg in site for reg in ['IT','DE','UK','FR','ES']])]
print europ_t1
for one in europ_t1:
for two in europ_t1:
if one==two: continue
mapping[one].append(two)
pass
## fnal can read from cnaf ?
#mapping['T1_IT_CNAF'].append( 'T1_US_FNAL' )
mapping['T1_IT_CNAF'].extend( [site for site in SI.sites_ready if '_US_' in site] ) ## all US can read from CNAF
mapping['T1_IT_CNAF'].append( 'T2_CH_CERN' )
mapping['T1_DE_KIT'].append( 'T2_CH_CERN' )
mapping['T2_CH_CERN'].append( 'T1_IT_CNAF' )
mapping['T2_CH_CERN'].append( 'T1_US_FNAL' )
#mapping['T2_UK_London_IC'].append( 'T2_CH_CERN' )
#mapping['T1_UK_RAL'].append( 'T2_BE_IIHE' )
mapping['T2_UK_London_IC'].append( 'T2_BE_IIHE' )
mapping['T2_UK_London_IC'].append( 'T2_FR_CCIN2P3' )
for site in SI.sites_ready:
if '_US_' in site:
mapping[site].append('T2_CH_CERN')
## make them appear as OK to use
force_sites = []
## overflow CERN to underutilized T1s
upcoming = json.loads( open('%s/GQ.json'%monitor_dir).read())
for possible in SI.sites_T1s:
if not possible in upcoming:
mapping['T2_CH_CERN'].append(possible)
pass
## remove add-hoc sites from overflow mapping
prevent_sites = []#'T2_US_Purdue']
for prevent in prevent_sites:
if prevent in mapping: mapping.pop( prevent )
for src in mapping:
for prevent in prevent_sites:
if prevent in mapping[src]:
mapping[src].remove( prevent )
## create the reverse mapping for the condor module
for site,fallbacks in mapping.items():
for fb in fallbacks:
if not site in reversed_mapping[fb]:
reversed_mapping[fb].append(site)
## this is the fallback mapping
print "Direct mapping : site => overflow"
print json.dumps( mapping, indent=2)
print "Reverse mapping : dest <= from origin"
print json.dumps( reversed_mapping, indent=2)
altered_tasks = set()
def running_idle( wfi , task_name):
gmon = wfi.getGlideMon()
#print gmon
if not gmon: return (0,0)
if not task_name in gmon: return (0,0)
return (gmon[task_name]['Running'], gmon[task_name]['Idle'])
def needs_action( wfi, task, min_idled = 100):
task_name = task.pathName.split('/')[-1]
running, idled = running_idle( wfi, task_name)
go = True
if not idled and not running :
go = False
if idled < 100:
go = False
if (not running and idled) or (running and (idled / float(running) > needs_action.pressure)):
go = True
else:
go = False
return go, task_name, running, idled
needs_action.pressure = UC.get('overflow_pressure')
def getPerf( task , stats_to_go = 200):
task = task.split('/')[1]+'/'+task.split('/')[-1]
try:
u = 'http://cms-gwmsmon.cern.ch/prodview/json/history/memoryusage720/%s'%task
print u
perf_data = json.loads(os.popen('curl -s --retry 5 %s'%u).read())
except Exception as e:
print str(e)
return (None,None)
buckets = filter(lambda i:i['key']!=0,perf_data['aggregations']["2"]['buckets'])
buckets.sort( key = lambda i:i['key'])
s=0
for bucket in buckets:
s+= bucket['doc_count']
bucket['cum'] = s
s_m = sum( bucket['key']*bucket['doc_count'] for bucket in buckets)
w_m = sum( bucket['doc_count'] for bucket in buckets)
m_m = max( bucket['key'] for bucket in buckets) if buckets else None
#90% percentile calculation
percentile_m = int(0.90 * w_m)
p_m = 0
s=0
for bucket in buckets:
p_m = bucket['key']
if bucket['cum'] > percentile_m:
break
p_m *= 1.1
print "percentile mem",p_m
max_count_m = None
max_count = 0
for bucket in buckets:
if bucket['doc_count'] > max_count:
max_count_m = bucket['key']
max_count = bucket['doc_count']
if max_count_m:
max_count_m *= 1.1
print "max count mem",max_count_m
b_m = None
if w_m > stats_to_go:
if p_m:
b_m = int(p_m)
else:
b_m = int(m_m) ## this is very bad if there are just a couple of outliers
b_m = int((s_m / float(w_m)) * 1.2)
else:
print "not enough stats for memory",w_m
try:
perf_data = json.loads(os.popen('curl -s --retry 5 http://cms-gwmsmon.cern.ch/prodview/json/history/runtime720/%s'%task).read())
except Exception as e:
print str(e)
return (b_m,None)
buckets = filter(lambda i:i['key']!=0,perf_data['aggregations']["2"]['buckets'])
buckets.sort( key = lambda i:i['key'])
s=0
for bucket in buckets:
s+= bucket['doc_count']
bucket['cum'] = s
s_t = sum( bucket['key']*bucket['doc_count'] for bucket in buckets)
w_t = sum( bucket['doc_count'] for bucket in buckets)
m_t = max( bucket['key'] for bucket in buckets) if buckets else None
percentile_t = int(0.90 * w_t)
p_t = 0
for bucket in buckets:
p_t = bucket['key']
if bucket['cum'] > percentile_t:
break
p_t *= 1.1
print "percentile time",p_t
max_count_t = None
max_count = 0
for bucket in buckets:
if bucket['doc_count'] > max_count:
max_count_t = bucket['key']
max_count = bucket['doc_count']
if max_count_t:
max_count_t *= 1.1
print "max count time",max_count_t
b_t = None
if w_t > stats_to_go:
b_t = m_t
else:
print "not enough stats for time",w_t
return (b_m,b_t)
def getcampaign( task ):
try:
taskname = task.pathName.split('/')[-1]
if hasattr( task, 'prepID'):
return task.prepID.split('-')[1]
elif taskname.count('-')>=1:
return taskname.split('-')[1]
else:
return None
except Exception as e :
print "Inconsistent prepid very likely"
print str(e)
return None
def close( interface ):
open('%s/equalizor.json.new'%monitor_pub_dir,'w').write( json.dumps( interface, indent=2))
os.system('mv %s/equalizor.json.new %s/equalizor.json'%(monitor_pub_dir,monitor_pub_dir))
os.system('cp %s/equalizor.json %s/logs/equalizor/equalizor.%s.json'%(monitor_pub_dir,monitor_dir,time.mktime(time.gmtime())))
interface = {
'mapping' : mapping,
'reversed_mapping' : reversed_mapping,
'modifications' : {},
'time' : {},
'memory' : {}
}
if options.augment or options.remove:
interface['modifications'] = json.loads( open('%s/equalizor.json'%monitor_pub_dir).read())['modifications']
interface['memory'] = json.loads( open('%s/equalizor.json'%monitor_pub_dir).read())['memory']
interface['time'] = json.loads( open('%s/equalizor.json'%monitor_pub_dir).read())['time']
if options.remove:
if specific in interface['modifications']:
print "poping",specific
interface['modifications'].pop(specific)
close( interface )
return
PU_locations = {}
PU_overflow = {}
PRIM_overflow = {}
PREMIX_overflow = {}
LHE_overflow = {}
tune_performance = []
pending_HLT = 0
max_HLT = 60000
pending_T0 = 0
max_T0 = 60000
try:
gmon = json.loads(os.popen('curl -s http://cms-gwmsmon.cern.ch/prodview/json/T2_CH_CERN_HLT').read())
pending_HLT += gmon["Running"]
pending_HLT += gmon["MatchingIdle"]
except:
pass
restricting_to_ready = ['pdmvserv_HIG-RunIISummer15wmLHEGS-00420_00157_v0__160909_001612_2018',
'pdmvserv_HIG-RunIISummer15wmLHEGS-00418_00157_v0__160909_001621_321',
'pdmvserv_HIG-RunIISummer15wmLHEGS-00419_00157_v0__160909_001621_2641'
]
remove_from = {
#'cerminar_Run2016B-v1-BTagCSV-23Sep2016_8020_160923_163224_2174' : ['T2_CH_CERN_HLT']
}
add_to = {
#'pdmvserv_EXO-RunIISpring16MiniAODv2-05060_00552_v0__161001_151813_7925' : ['T3_US_OSG'],
#'cerminar_Run2016C-v2-SingleElectron-23Sep2016_8020_160923_182146_3498' : ['T3_US_NERSC'],
#'cerminar_Run2016C-v2-Tau-23Sep2016_8020_160923_182336_5649' : ['T3_US_NERSC'],
}
stay_within_site_whitelist = False
specific_task=None
if specific and ":" in specific:
specific,specific_task = specific.split(':')
if specific:
wfs = session.query(Workflow).filter(Workflow.name.contains(specific)).all()
else:
wfs = session.query(Workflow).filter(Workflow.status == 'away').all()
performance = {}
resizing = {}
no_routing = [
]
random.shuffle( wfs )
for wfo in wfs:
if not wfo.status in ['away']: continue
if wfo.name in no_routing and not options.augment:
continue
if specific and not specific in wfo.name:
continue
if specific:
wfi = workflowInfo(url, wfo.name)
else:
cached = filter(lambda d : d['RequestName']==wfo.name, workflows)
if not cached : continue
wfi = workflowInfo(url, wfo.name, request = cached[0])
## only running-* should get re-routed, unless done by hand
if not wfi.request['RequestStatus'] in ['running-open','running-closed'] and not specific: continue
is_chain = (wfi.request['RequestType'] in ['TaskChain','StepChain'])
tasks_and_campaigns = []
for task in wfi.getWorkTasks():
tasks_and_campaigns.append( (task, getcampaign(task) ) )
lhe,prim,_,sec,sites_allowed = wfi.getSiteWhiteList()#getIO()
ncores = wfi.getMulticore()
memory_allowed = SI.sitesByMemory( float(wfi.request['Memory']) , maxCore=ncores)
if not lhe and not prim and not sec:
## no input at all: go for OSG!!!
add_to[wfo.name] = ['T3_US_OSG']
## check needs override
needs_overide = False
if not needs_overide and options.augment: needs_overide=True
def overide_from_agent( wfi, needs_overide):
bad_agents = []#'http://cmssrv219.fnal.gov:5984']
if not bad_agents: return needs_overide
if needs_overide: return True
agents = wfi.getAgents()
wqss = ['Running','Acquired']
if any([agent in agents.get(wqs,{}).keys() for wqs,agent in itertools.product( wqss, bad_agents)]):
print "overriding the need for bad agent"
needs_overide = True
return needs_overide
## now parse this for action
for i_task,(task,campaign) in enumerate(tasks_and_campaigns):
if options.augment:
print task.pathName
print campaign
resize = CI.get(campaign,'resize',{})
if resize and not is_chain:
resizing[task.pathName] = resize
tune = CI.get(campaign,'tune',options.tune)
if tune and not campaign in tune_performance:
tune_performance.append( campaign )
overflow = CI.get(campaign,'overflow',{})
if overflow:
if "PRIM" in overflow and not campaign in PRIM_overflow:
PRIM_overflow[campaign] = copy.deepcopy(overflow['PRIM'])
print "adding",campaign,"to PRIM overflow"
if "PREMIX" in overflow and not campaign in PREMIX_overflow:
PREMIX_overflow[campaign] = copy.deepcopy(overflow['PREMIX'])
print "adding",campaign,"to PREMIX overflow"
if "PU" in overflow and not campaign in PU_overflow:
PU_overflow[campaign] = copy.deepcopy(overflow['PU'])
print "adding",campaign,"to PU overflow rules"
if "LHE" in overflow and not campaign in LHE_overflow:
site_list = overflow['LHE'].get('site_list',"")
if site_list:
if type(site_list)==list:
LHE_overflow[campaign] = site_list
else:
print site_list
if hasattr(SI,site_list):
LHE_overflow[campaign] = copy.deepcopy( getattr(SI,site_list) )
else:
LHE_overflow[campaign] = site_list.split(',')
print "adding",campaign,"to light input overflow rules",LHE_overflow[campaign]
### setup the resizing
### get the task performance, for further massaging.
if campaign in tune_performance or options.tune:
print "performance",task.taskType,task.pathName
if task.taskType in ['Processing','Production']:
set_memory,set_time = getPerf( task.pathName )
#print "Performance %s GB %s min"%( set_memory,set_time)
wfi.sendLog('equalizor','Performance tuning to %s GB %s min for %s'%( set_memory,set_time,task.pathName.split('/')[-1] ))
## get values from gmwsmon
# massage the values : 95% percentile
performance[task.pathName] = {}
if set_memory:
performance[task.pathName]['memory']=min(set_memory,15000) ## max to 15GB
if set_time:
performance[task.pathName]['time'] = min(set_time, 1440) ## max to 24H
## rule to remove from the site whitelist site that do not look ready for unified (local banning)
if wfo.name in restricting_to_ready:
if task.taskType in ['Production']:
new_list = list(set(SI.sites_ready)&set(wfi.request['SiteWhitelist']))
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : new_list }
if campaign in PREMIX_overflow:
## figure out secondary location and neighbors
## figure out primary presence and neighbors
## do the intersection and add if in need.
needs, task_name, running, idled = needs_action(wfi, task)
#needs = True
## trick to be removed once all wf are passed through the agent patch
assigned_log = filter(lambda change : change["Status"] in ["assigned","acquired"],wfi.request['RequestTransition'])
if assigned_log:
then = assigned_log[0]['UpdateTime']
if then < 1479481842:
print "assigned too early"
needs = False
else:
print "assigned later enough"
else:
needs = False
if is_chain and task.pathName.endswith('_1'):
print i_task,"in chain prevents overflowing"
needs = False
if task.taskType in ['Processing','Production'] and needs:
secondary_locations = set(SI.sites_ready + force_sites)
for s in sec:
if not s in PU_locations:
presence = getDatasetPresence( url, s)
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
PU_locations[s] = one_secondary_locations
print "secondary is at",sorted(PU_locations[s])
secondary_locations = set([SI.SE_to_CE(site) for site in PU_locations[s]]) & secondary_locations
aaa_sec_grid = set(secondary_locations)
for site in sorted(aaa_sec_grid):
aaa_sec_grid.update( mapping.get(site, []) )
if len(prim):
dataset = list(prim)[0]
all_blocks,blocks = wfi.getActiveBlocks()
count_all = sum([len(v) for k,v in all_blocks.items()])
presence = getDatasetPresence(url, dataset, only_blocks=blocks )
aaa_prim_grid = set([SI.SE_to_CE(site) for site in presence.keys()])
for site in sorted(aaa_prim_grid):
aaa_prim_grid.update( mapping.get(site, []) )
print sorted(aaa_prim_grid),"around primary location",sorted(presence.keys())
print sorted(aaa_sec_grid),"aroudn secondary location",sorted(secondary_locations)
## intersect
aaa_grid = aaa_sec_grid & aaa_prim_grid
else:
print "premix overflow from a taskchain"
### hack hack hack
#modifications[wfo.name][task.pathName]= {"ReplaceSiteWhitelist" : ['T2_CH_CERN','T1_US_FNAL']}
aaa_grid = set(wfi.request['SiteWhitelist'])
banned_until_you_find_a_way_to_do_this = ['T3_US_OSG']
aaa_grid = filter(lambda s : not s in banned_until_you_find_a_way_to_do_this, aaa_grid)
if aaa_grid:
wfi.sendLog('equalizor','Extending site whitelist to %s'%sorted(aaa_grid))
modifications[wfo.name][task.pathName]= {"AddWhitelist" : sorted(aaa_grid)}
## rule to overflow jobs on the primary input
if campaign in PRIM_overflow:
if task.taskType in ['Processing','Production']:
if not wfi.request['TrustSitelists']:
###xrootd is OFF
dataset = list(prim)[0]
all_blocks,blocks = wfi.getActiveBlocks()
count_all = sum([len(v) for k,v in all_blocks.items()])
presence = getDatasetPresence(url, dataset, only_blocks=blocks )
in_full = [SI.SE_to_CE(site) for site,(there,_) in presence.items() if there]
aaa_grid= set()
aaa_grid_in_full = set(in_full)
for site in sorted(aaa_grid_in_full):
aaa_grid_in_full.update( mapping.get(site, []) )
## just add the neighbors to the existing whitelist. we could do more with block classAd
for site in wfi.request['SiteWhitelist']:
aaa_grid.update( mapping.get(site, []) )
aaa_grid = aaa_grid & set(sites_allowed + ['T3_US_NERSC']) ## and restrict to site that would be allowed at all (mcore, mem)
aaa_grid_in_full = aaa_grid_in_full & set(sites_allowed + ['T3_US_NERSC']) ## and restrict to site that would be allowed at all (mcore, mem)
gmon = wfi.getGlideMon()
needs, task_name, running, idled = needs_action(wfi, task)
print needs,running,idled
site_in_use = set(gmon[task_name]['Sites']) if gmon and task_name in gmon and 'Sites' in gmon[task_name] else set()
print dataset,"at",sorted(in_full),len(blocks),"/",count_all
print "running at",sorted(site_in_use)
print "set for",sorted(wfi.request['SiteWhitelist'])
print "around current whitelist" ,sorted(aaa_grid)
print "around where the data is now in full", sorted(aaa_grid_in_full)
if needs and not (site_in_use & set(in_full)) and aaa_grid_in_full:
print "we could be going for replace at that point"
wfi.sendLog('equalizor','Replaceing site whitelie to %s dynamically'% sorted(aaa_grid_in_full))
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : sorted( aaa_grid_in_full) }
else:
if aaa_grid:
print wfo.name
wfi.sendLog('equalizor','Adding in site white list %s dynamically'% sorted(aaa_grid) )
if wfo.name in modifications and task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
modifications[wfo.name][task.pathName]["AddWhitelist"].extend(sorted(aaa_grid))
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : sorted(aaa_grid) }
else:
## the request is already is in xrootd mode (either too generous, or just about right with neighbors of full data)
dataset = list(prim)[0]
all_blocks,blocks = wfi.getActiveBlocks()
count_all = sum([len(v) for k,v in all_blocks.items()])
fraction_left = float(len(blocks))/ count_all
#if fraction_left< 0.5: print '\n'.join( blocks )
presence = getDatasetPresence(url, dataset, only_blocks=blocks )
## in full is really the only place we can go to safely, since we have no job-data matching
in_full = [SI.SE_to_CE(site) for site,(there,_) in presence.items() if there]
gmon = wfi.getGlideMon()
needs, task_name, running, idled = needs_action(wfi, task)
site_in_use = set(gmon[task_name]['Sites']) if gmon and task_name in gmon and 'Sites' in gmon[task_name] else set()
print needs,running,idled
aaa_grid = set(in_full)
for site in list(aaa_grid):
aaa_grid.update( mapping.get(site, []) )
new_ones = set(in_full) - set(wfi.request['SiteWhitelist']) ## symptomatic of data have been repositionned
common = set(in_full) & set(wfi.request['SiteWhitelist'])
extra_shit = set(wfi.request['SiteWhitelist']) - aaa_grid ## symptomatic of too generous site-whitelist
aaa_grid = aaa_grid & set(sites_allowed+ ['T3_US_NERSC']) ## restrict to site that would be allowed at all (mcore, mem)
new_grid = aaa_grid - set(wfi.request['SiteWhitelist'])
print dataset,"is in full ",len(blocks),"/",count_all," at",in_full
print '\n'.join( sorted(blocks) )
print "running at",site_in_use
print "in common of the site whitelist",sorted(common)
print "site now also hosting the data",sorted(new_ones)
print "site in whitelist with no data",sorted(extra_shit)## with no data and not within aaa reach
if new_ones:
## we will be add sites
if needs and aaa_grid:
print wfo.name,"would replace for",sorted(aaa_grid)
print "but no thanks"
wfi.sendLog('equalizor','Changing the site whitelist to %s dynamically'%(sorted(aaa_grid)))
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : sorted(aaa_grid) }
elif new_grid:
print wfo.name,"would complement up to",sorted(aaa_grid)
wfi.sendLog('equalizor','Adding site white list to %s dynamically'% sorted(new_grid) )
modifications[wfo.name][task.pathName] = { "AddWhitelist" : sorted(new_grid) }
elif len(extra_shit)>5:
if aaa_grid:
print wfo.name,"would be restricting down to",sorted(aaa_grid),"because of",sorted(extra_shit)
wfi.sendLog('equalizor','Restricting the white list to %s dynamically'% sorted(aaa_grid) )
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : sorted(aaa_grid) }
else:
print wfo.name,"don't do anything"
if wfo.name in remove_from and task.taskType in ['Processing','Production']:
remove = remove_from[wfo.name]
restrict_to = set(wfi.request['SiteWhitelist'])
intersection= set(remove)&set(restrict_to)
if intersection:
print intersection,"is indeed in the original whitelist"
restrict_to = restrict_to - set(remove)
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : sorted(restrict_to) }
if wfo.name in add_to:
if task.taskType in ['Production','Processing']:
augment_to = add_to[wfo.name]
print "adding",sorted(augment_to),"to",wfo.name
if wfo.name in modifications and task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
modifications[wfo.name][task.pathName]['AddWhitelist'].extend( augment_to )
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : augment_to }
### rule to avoid the issue of taskchain secondary jobs being stuck at sites processing the initial step
if campaign in LHE_overflow:
#if not is_chain and task.taskType in ['Processing']:
if task.taskType in ['Processing']:
needs, task_name, running, idled = needs_action(wfi, task)
needs_overide = overide_from_agent( wfi, needs_overide)
extend_to = list(set(copy.deepcopy( LHE_overflow[campaign] )))
if stay_within_site_whitelist:
extend_to = list(set(extend_to) & set(wfi.request['SiteWhitelist'])) ## restrict to stupid-site-whitelist
extend_to = list(set(extend_to) & set(SI.sites_ready + force_sites))
if is_chain:
print "further restricting to initially allowed sites"
## restrict to initial allowed sites
extend_to = list(set(extend_to) & set(sites_allowed))
if not extend_to:
print "Nowhere to extend to"
continue
if extend_to and needs or needs_overide:
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : extend_to ,"Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : ReplaceSiteWhitelist \n %s'%( task_name,
wfo.name,
running,
idled ,
json.dumps( sorted(modifications[wfo.name][task.pathName]['ReplaceSiteWhitelist']))))
altered_tasks.add( task.pathName )
else:
wfi.sendLog('equalizor','%s of %s is running %d and pending %d'%( task_name, wfo.name, running, idled))
### overflow the 76 digi-reco to the site holding the pileup
if campaign in PU_overflow:
force = PU_overflow[campaign]['force'] if 'force' in PU_overflow[campaign] else False
secondary_locations = set(SI.sites_ready + force_sites)
for s in sec:
if not s in PU_locations:
presence = getDatasetPresence( url, s)
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
PU_locations[s] = one_secondary_locations
print "secondary is at",sorted(PU_locations[s])
secondary_locations = set([SI.SE_to_CE(site) for site in PU_locations[s]]) & secondary_locations
## we should add all sites that hold the secondary input if any
### given that we have the secondary location available, it is not necessary to use the add-hoc list
##secondary_locations = list(set(PU_overflow[campaign]['sites']) & set( SI.sites_ready ))
## intersect with the sites that are allowed from the request requirement
secondary_locations = secondary_locations & set(memory_allowed)
if any([task.pathName.endswith(finish) for finish in ['_0','StepOneProc','Production']]) :
needs, task_name, running, idled = needs_action(wfi, task)
## removing the ones in the site whitelist already since they encode the primary input location
if stay_within_site_whitelist:
original_site_in_use = set(wfi.request['SiteWhitelist'] & set(secondary_locations))
else:
original_site_in_use = set(secondary_locations)
mode = 'AddWhitelist'
if not prim and i_task==0:
print "because there isn't any input, one should be able to just replace the sitewhitelist instead of adding, with the restriction of not reaching every possible sites"
mode='ReplaceSiteWhitelist'
## remove the sites that have already running jobs
gmon = wfi.getGlideMon()
if gmon and task_name in gmon and 'Sites' in gmon[task_name] and mode=='AddWhitelist':
site_in_use = set(gmon[task_name]['Sites'])
site_in_use = set([]) ## at this time I cannot find a reason to apply such limitation
print "removing",sorted(site_in_use)
## that determines where you want to run in addition
augment_by = list((set(secondary_locations)- site_in_use) & original_site_in_use)
else:
print "no existing running site"
augment_by = list(original_site_in_use)
if not augment_by: print "Nowhere to extend to"
needs_overide = overide_from_agent( wfi, needs_overide)
if augment_by and (needs or needs_overide or force) and PU_overflow[campaign]['pending'] < PU_overflow[campaign]['max']:
PU_overflow[campaign]['pending'] += idled
print "raising overflow to",PU_overflow[campaign]['pending'],"for",PU_overflow[campaign]['max']
## the step with an input ought to be the digi part : make this one go anywhere
modifications[wfo.name][task.pathName] = { mode : augment_by , "Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
altered_tasks.add( task.pathName )
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : %s \n %s'%( task_name, wfo.name,
running, idled,
mode,
json.dumps( sorted(augment_by), indent=2 )))
else:
print task_name,"of",wfo.name,"running",running,"and pending",idled
### overflow the skims back to multi-core
if campaign in ['Run2015D','Run2015C_25ns'] and task.taskType =='Skim':
original_swl = wfi.request['SiteWhitelist']
needs, task_name, running, idled = needs_action(wfi, task)
if (needs or needs_overide):
modifications[wfo.name][task.pathName] = { 'AddWhitelist' : original_swl,
"Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
altered_tasks.add( task.pathName )
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : AddWhitelist \n %s'%( task_name, wfo.name,
running, idled,
json.dumps( sorted(original_swl), indent=2 )))
if options.augment:
#print "uhm ....",sorted(wfi.request['SiteWhitelist']),i_task,use_HLT
pass
### this is a hack when we need to kick gensim out of everything
if campaign in [
#'RunIIWinter15GS',
#'RunIISummer15GS',
#'RunIISummer15wmLHEGS',
#'Summer12',
] and task.taskType in ['Production'] and is_chain:
#what are the site you want to take out. What are the jobs in whitelist, make the diff and replace
t1s = set([site for site in SI.all_sites if site.startswith('T1')])
ust2s = set([site for site in SI.all_sites if site.startswith('T2_US')])
#ust2s = set([site for site in SI.sites_mcore_ready if site.startswith('T2_US')])
allmcores = set(SI.sites_mcore_ready)
#set_for = set(wfi.request['SiteWhitelist']) - t1s
#set_for = set(wfi.request['SiteWhitelist']) - t1s - ust2s
#set_for = set(wfi.request['SiteWhitelist']) - allmcores
set_for = set(wfi.request['SiteWhitelist']) & t1s
print wfo.name,"going for",set_for
print task.pathName
if set_for:
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : sorted(set_for) }
### add the HLT at partner of CERN
if 'T2_CH_CERN' in wfi.request['SiteWhitelist'] and i_task in [0,1] and use_HLT and not wfi.request['TrustSitelists']:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs=True
needs = True
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide) and pending_HLT < max_HLT:
pending_HLT += idled
if task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
modifications[wfo.name][task.pathName]["AddWhitelist"].append( "T2_CH_CERN_HLT" )
wfi.sendLog('equalizor','also adding the HLT in whitelist of %s to %d for %d'%( task.pathName, pending_HLT, max_HLT))
## this Replace does not work at all for HLT
elif task.pathName in modifications[wfo.name] and 'ReplaceSiteWhitelist' in modifications[wfo.name][task.pathName]:
#modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"].append( "T2_CH_CERN_HLT" )
#print "\t",wfo.name,"adding replace HLT up to",pending_HLT,"for",max_HLT
print "already having a site replacement, not adding the HLT for now"
pass
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : ["T2_CH_CERN_HLT"],
"Priority" : wfi.request['RequestPriority'],
"Running" : running,
"Pending" : idled}
wfi.sendLog('equalizor','adding the HLT in whitelist of %s to %d for %d'%( task.pathName, pending_HLT, max_HLT))
if i_task==0 and not sec and use_T0:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs=True
#needs = True
good_type = wfi.request['RequestType'] in ['MonteCarlo','MonteCarloFromGEN']
read_lhe = ((not 'LheInputFiles' in wfi.request) or bool(wfi.request['LheInputFiles']))
good_type &= not read_lhe
if not good_type and not options.augment: needs = False
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide):
pending_T0 += idled
if task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][task.pathName]["AddWhitelist"]:
modifications[wfo.name][task.pathName]["AddWhitelist"].append( "T0_CH_CERN" )
wfi,sendLog('equalizor','adding the T0 for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
elif task.pathName in modifications[wfo.name] and 'ReplaceSiteWhitelist' in modifications[wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"]:
modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"].append( "T0_CH_CERN" )
wfi,sendLog('equalizor','adding the T0 to replacement for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : ["T0_CH_CERN"],
"Priority" : wfi.request['RequestPriority'],
"Running" : running,
"Pending" : idled}
wfi,sendLog('equalizor','adding the T0 for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
interface['modifications'].update( modifications )
### manage the number of core and job resizing
#interface['cores']={'T2_CH_CERN_HLT': {'min':4,'max':16}, 'default': {'min':1, 'max':4}}
#interface['resizes'] = ['RunIISpring16DR80']
interface['resizing'] = resizing
### manage the modification of the memory and target time
max_N_mem = 2
max_N_time = 4
## discretize the memory to 10 at most values
mems = set([o['memory'] for t,o in performance.items() if 'memory' in o])
times = set([o['time'] for t,o in performance.items() if 'time' in o])
if len(mems)>max_N_mem:
mem_step = int((max(mems) - min(mems))/ float(max_N_mem))
print "rebinning memory"
for t in performance:
if not 'memory' in performance[t]: continue
(m,r) = divmod(performance[t]['memory'], mem_step)
performance[t]['memory'] = (m+1)*mem_step
if len(times)>max_N_time:
print "rebinning memory"
time_step = int((max(times) - min(times))/float(max_N_time))
for t in performance:
if not 'time' in performance[t]: continue
(m,r) = divmod(performance[t]['time'], time_step)
performance[t]['time'] = (m+1)*time_step
new_times = defaultdict(list)
new_memories = defaultdict(list)
for t,o in performance.items():
if 'time' in o:
new_times[str(o['time'])].append( t )
if 'memory' in o:
new_memories[str(o['memory'])].append( t )
interface['time'].update( new_times )
interface['memory'].update( new_memories )
## close and save
close( interface )
def injector(url, options, specific):
mlock = moduleLock()
if mlock(): return
use_mcm = True
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): return
use_mcm = up.status['mcm']
UC = unifiedConfiguration()
transform_keywords = UC.get('convert_to_stepchain')
workflows = getWorkflows(url, status=options.wmstatus, user=options.user)
for user in UC.get("user_rereco"):
workflows.extend(
getWorkflows(url,
status=options.wmstatus,
user=user,
rtype="ReReco"))
for user in (options.user_relval.split(',')
if options.user_relval else UC.get("user_relval")):
workflows.extend(
getWorkflows(url,
status=options.wmstatus,
user=user,
rtype="TaskChain"))
for user in (options.user_storeresults.split(',') if
options.user_storeresults else UC.get("user_storeresults")):
workflows.extend(
getWorkflows(url,
status=options.wmstatus,
user=user,
rtype="StoreResults"))
print len(workflows), "in line"
cannot_inject = set()
to_convert = set()
status_cache = defaultdict(str)
## browse for assignment-approved requests, browsed for ours, insert the diff
for wf in workflows:
if specific and not specific in wf: continue
exists = session.query(Workflow).filter(Workflow.name == wf).first()
if not exists:
wfi = workflowInfo(url, wf)
## check first that there isn't related here with something valid
can_add = True
## first try at finding a match
familly = session.query(Workflow).filter(
Workflow.name.contains(wfi.request['PrepID'])).all()
if not familly:
pids = wfi.getPrepIDs()
req_familly = []
for pid in pids:
req_familly.extend(getWorkflowById(url, pid, details=True))
familly = []
print len(req_familly), "members"
for req_member in req_familly:
#print "member",req_member['RequestName']
owfi = workflowInfo(url,
req_member['RequestName'],
request=req_member)
other_pids = owfi.getPrepIDs()
if set(pids) == set(other_pids):
## this is a real match
familly.extend(
session.query(Workflow).filter(
Workflow.name ==
req_member['RequestName']).all())
for lwfo in familly:
if lwfo:
## we have it already
if not lwfo.status in [
'forget', 'trouble', 'forget-unlock',
'forget-out-unlock'
]:
wfi.sendLog(
'injector', "Should not put %s because of %s %s" %
(wf, lwfo.name, lwfo.status))
sendLog('injector',
"Should not put %s because of %s %s" %
(wf, lwfo.name, lwfo.status),
level='critical')
print "Should not put", wf, "because of", lwfo.name, lwfo.status
cannot_inject.add(wf)
can_add = False
## add a check on validity of input datasets
_, prim, par, sec = wfi.getIO()
for d in list(prim) + list(par) + list(sec):
if not d in status_cache:
status_cache[d] = getDatasetStatus(d)
if status_cache[d] != 'VALID':
wfi.sendLog(
'injector', "One of the input is not VALID. %s : %s" %
(d, status_cache[d]))
sendLog('injector',
"One of the input of %s is not VALID. %s : %s" %
(wf, d, status_cache[d]),
level='critical')
can_add = False
## check for any file in phedex, to verify existence
_, ph_files, _, _ = getDatasetFiles(url, d)
if not ph_files and not ('StoreResults'
== wfi.request.setdefault(
'RequestType', None)):
wfi.sendLog(
'injector',
"One of the input has no file in phedex: %s" % d)
sendLog('injector',
"One of the input has no file in phedex: %s" % d,
level='critical')
can_add = False
### ban some workflow that you don't like anymore
#outputs = wfi.request['OutputDatasets']
if not can_add: continue
## temporary hack to transform specific taskchain into stepchains
#good_for_stepchain = wfi.isGoodToConvertToStepChain( keywords = transform_keywords)
good_for_stepchain = wfi.isGoodToConvertToStepChain(keywords=None)
## match keywords and technical constraints
#if (not options.no_convert) and good_for_stepchain and not wfi.isRelval():
# to_convert.add( wf )
# wfi.sendLog('injector','Transforming %s TaskChain into StepChain'%wf)
# #sendEmail('convertion to stepchain','Transforming %s TaskChain into StepChain'%wf)
wfi.sendLog('injector', "considering %s" % wf)
new_wf = Workflow(name=wf,
status=options.setstatus,
wm_status=options.wmstatus)
session.add(new_wf)
session.commit()
time.sleep(0.5)
else:
#print "already have",wf
pass
if cannot_inject:
#sendEmail('workflow duplicates','These workflow cannot be added in because of duplicates \n\n %s'%( '\n'.join(cannot_inject)))
sendLog(
'injector',
'These workflow cannot be added in because of duplicates \n\n %s' %
('\n'.join(cannot_inject)),
level='warning')
for wf in to_convert:
os.system(
'./Unified/rejector.py --clone --to_step --comments \"Transform to StepChain\" %s'
% wf)
## passing a round of invalidation of what needs to be invalidated
if use_mcm and (options.invalidate or True):
invalidator(url)
no_replacement = set()
#print "getting all transfers"
#all_transfers=session.query(Transfer).all()
#print "go!"
## pick up replacements
for wf in session.query(Workflow).filter(
Workflow.status == 'trouble').all():
print wf.name
if specific and not specific in wf.name: continue
print wf.name
wfi = workflowInfo(url, wf.name)
wl = wfi.request #getWorkLoad(url, wf.name)
familly = getWorkflowById(url, wl['PrepID'])
true_familly = []
for member in familly:
if member == wf.name: continue
fwl = getWorkLoad(url, member)
if options.replace:
if member != options.replace: continue
else:
if fwl['RequestDate'] < wl['RequestDate']: continue
if fwl['RequestType'] == 'Resubmission': continue
if fwl['RequestStatus'] in ['None', None, 'new']: continue
if fwl['RequestStatus'] in [
'rejected', 'rejected-archived', 'aborted',
'aborted-archived'
]:
continue
true_familly.append(fwl)
if len(true_familly) == 0:
#sendLog('injector','%s had no replacement'%wf.name, level='critical')
if wfi.isRelval():
#wfi.sendLog('injector','the workflow was found in trouble with no replacement. As a relval, there is no clean way to handle this.')
wfi.sendLog(
'injector',
'the workflow was found in trouble with no replacement. As a relval, there is no clean way to handle this. Setting forget'
)
wf.status = 'forget'
session.commit()
else:
wfi.sendLog(
'injector',
'the workflow was found in trouble with no replacement')
no_replacement.add(wf.name)
continue
else:
wfi.sendLog(
'injector',
'the workflow was found in trouble and has a replacement')
print wf.name, "has", len(familly), "familly members"
print wf.name, "has", len(true_familly), "true familly members"
##we cannot have more than one of them !!! pick the last one
if len(true_familly) > 1:
#sendEmail('multiple wf','please take a look at injector for %s'%wf.name)
sendLog('injector',
'Multiple wf in line, will take the last one for %s \n%s' %
(wf.name, ', '.join(fwl['RequestName']
for fwl in true_familly)),
level='critical')
for fwl in true_familly[-1:]:
member = fwl['RequestName']
new_wf = session.query(Workflow).filter(
Workflow.name == member).first()
if not new_wf:
sendLog('injector',
"putting %s as replacement of %s" % (member, wf.name))
status = 'away'
if fwl['RequestStatus'] in ['assignment-approved']:
status = 'considered'
new_wf = Workflow(name=member,
status=status,
wm_status=fwl['RequestStatus'])
wf.status = 'forget'
session.add(new_wf)
else:
if new_wf.status == 'forget': continue
sendLog(
'injector',
"getting %s as replacement of %s" % (new_wf.name, wf.name))
wf.status = 'forget'
for tr in session.query(TransferImp).filter(
TransferImp.workflow_id == wf.id).all():
## get all transfer working for the old workflow
existing = session.query(TransferImp).filter(
TransferImp.phedexid == tr.phedexid).filter(
TransferImp.workflow_id == new_wf.id).all()
tr.active = False ## disable the old one
if not existing:
## create the transfer object for the new dependency
tri = TransferImp(phedexid=tr.phedexid, workflow=new_wf)
session.add(tri)
session.commit()
## don't do that automatically
#wf.status = 'forget'
session.commit()
if no_replacement:
#sendEmail('workflow with no replacement','%s \n are dangling there'%( '\n'.join(no_replacement)))
sendLog('injector',
'workflow with no replacement\n%s \n are dangling there' %
('\n'.join(no_replacement)),
level='critical')
def assignor(url ,specific = None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
#SI = siteInfo()
SI = global_SI()
#NLI = newLockInfo()
#if not NLI.free() and not options.go: return
LI = lockInfo()
if not LI.free() and not options.go: return
n_assigned = 0
n_stalled = 0
wfos=[]
fetch_from = []
if specific or options.early:
fetch_from.extend(['considered','staging'])
if specific:
fetch_from.extend(['considered-tried'])
fetch_from.extend(['staged'])
if options.from_status:
fetch_from = options.from_status.split(',')
print "Overriding to read from",fetch_from
for status in fetch_from:
wfos.extend(session.query(Workflow).filter(Workflow.status==status).all())
## in case of partial, go for fetching a list from json ?
#if options.partial and not specific:
# pass
dataset_endpoints = json.loads(open('%s/dataset_endpoints.json'%monitor_dir).read())
aaa_mapping = json.loads(open('%s/equalizor.json'%monitor_pub_dir).read())['mapping']
all_stuck = set()
all_stuck.update( json.loads( open('%s/stuck_transfers.json'%monitor_dir).read() ))
all_stuck.update( getAllStuckDataset())
max_per_round = UC.get('max_per_round').get('assignor',None)
max_cpuh_block = UC.get('max_cpuh_block')
random.shuffle( wfos )
for wfo in wfos:
if options.limit and (n_stalled+n_assigned)>options.limit:
break
if max_per_round and (n_stalled+n_assigned)>max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name,specific.split(','))): continue
#if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo( url, wfo.name)
if options.priority and int(wfh.request['RequestPriority']) < options.priority:
continue
options_text=""
if options.early: options_text+=", early option is ON"
if options.partial:
options_text+=", partial option is ON"
options_text+=", good fraction is %.2f"%options.good_enough
wfh.sendLog('assignor',"%s to be assigned%s"%(wfo.name, options_text))
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput,primary,parent,secondary, sites_allowed) = wfh.getSiteWhiteList()
output_tiers = list(set([o.split('/')[-1] for o in wfh.request['OutputDatasets']]))
is_stuck = (all_stuck & primary)
if is_stuck:
wfh.sendLog('assignor',"%s are stuck input"%(','.join( is_stuck)))
## check if by configuration we gave it a GO
no_go = False
if not wfh.go(log=True) and not options.go:
no_go = True
allowed_secondary = {}
assign_parameters = {}
check_secondary = False
for campaign in wfh.getCampaigns():
if campaign in CI.campaigns:
assign_parameters.update( CI.campaigns[campaign] )
if campaign in CI.campaigns and 'secondaries' in CI.campaigns[campaign]:
if CI.campaigns[campaign]['secondaries']:
allowed_secondary.update( CI.campaigns[campaign]['secondaries'] )
check_secondary = True
if campaign in CI.campaigns and 'banned_tier' in CI.campaigns[campaign]:
banned_tier = list(set(CI.campaigns[campaign]['banned_tier']) & set(output_tiers))
if banned_tier:
no_go=True
wfh.sendLog('assignor','These data tiers %s are not allowed'%(','.join( banned_tier)))
sendLog('assignor','These data tiers %s are not allowed'%(','.join( banned_tier)), level='critical')
if secondary and check_secondary:
if (set(secondary)&set(allowed_secondary.keys())!=set(secondary)):
wfh.sendLog('assignor','%s is not an allowed secondary'%(', '.join(set(secondary)-set(allowed_secondary.keys()))))
sendLog('assignor','%s is not an allowed secondary'%(', '.join(set(secondary)-set(allowed_secondary.keys()))), level='critical')
if not options.go:
no_go = True
## then get whether there is something more to be done by secondary
for sec in secondary:
if sec in allowed_secondary:# and 'parameters' in allowed_secondary[sec]:
assign_parameters.update( allowed_secondary[sec] )
if no_go:
n_stalled+=1
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] !='assignment-approved':
if not options.test:
wfh.sendLog('assignor',"setting %s away and skipping"%wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name,wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version=wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor',"cannot decide on version number")
n_stalled+=1
wfo.status = 'trouble'
session.commit()
continue
original_sites_allowed = copy.deepcopy( sites_allowed )
wfh.sendLog('assignor',"Site white list %s"%sorted(sites_allowed))
override_sec_location = CI.get(wfh.request['Campaign'], 'SecondaryLocation', [])
blocks = []
if 'BlockWhitelist' in wfh.request:
blocks = wfh.request['BlockWhitelist']
if 'RunWhitelist' in wfh.request and wfh.request['RunWhitelist']:
## augment with run white list
for dataset in primary:
blocks = list(set( blocks + getDatasetBlocks( dataset, runs=wfh.request['RunWhitelist'] ) ))
wfh.sendLog('assignor',"Allowed %s"%sorted(sites_allowed))
secondary_locations=None
primary_aaa = options.primary_aaa
secondary_aaa = options.secondary_aaa
do_partial = False #options.good_enough if options.partial else 0
if 'Campaign' in wfh.request and wfh.request['Campaign'] in CI.campaigns:
assign_parameters.update( CI.campaigns[wfh.request['Campaign']] )
if 'primary_AAA' in assign_parameters:
primary_aaa = primary_aaa or assign_parameters['primary_AAA']
if 'secondary_AAA' in assign_parameters:
secondary_aaa = secondary_aaa or assign_parameters['secondary_AAA']
if 'partial_copy' in assign_parameters:
## can this only work if there is a stuck input ? maybe not
## this is a number. 0 means no
print "Could do partial disk copy assignment"
if is_stuck or options.partial:
do_partial = assign_parameters['partial_copy']
wfh.sendLog('assignor',"Overiding partial copy assignment to %.2f fraction"% do_partial)
#sendEmail('stuck input to assignment','%s is stuck for assigning %s and going fractional'%(','.join( is_stuck), wfo.name))
do_partial = options.good_enough if options.partial else do_partial
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
#sendEmail("tempting to pass sec location check","but we cannot yet IMO")
#pass
if secondary_aaa:
#just continue without checking
continue
presence = getDatasetPresence( url, sec )
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations==None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [site for site in sites_allowed if SI.CE_to_SE(site) in one_secondary_locations]
wfh.sendLog('assignor',"From secondary requirement, now Allowed%s"%sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy( sites_allowed ) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy( sites_allowed )
sites_with_data = copy.deepcopy( sites_allowed )
sites_with_any_data = copy.deepcopy( sites_allowed )
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
endpoints = set()
for prim in list(primary):
if prim in dataset_endpoints:
print "endpoints from stagor",dataset_endpoints[prim]
endpoints.update( dataset_endpoints[prim] )
set_lfn = getLFNbase( prim )
presence = getDatasetPresence( url, prim , only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(url, prim, sites = [SI.CE_to_SE(site) for site in sites_allowed] , only_blocks = blocks)
#sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
#sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,(there,frac)) in presence.items() if there]]
sites_with_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,frac) in presence.items() if frac[1]>90.]]
sites_with_any_data = [site for site in sites_with_any_data if SI.CE_to_SE(site) in presence.keys()]
wfh.sendLog('assignor',"Holding the data but not allowed %s"%sorted(list(set([se_site for se_site in presence.keys() if not SI.SE_to_CE(se_site) in sites_allowed]))))
if primary_locations==None:
primary_locations = presence.keys()
else:
primary_locations = list(set(primary_locations) & set(presence.keys() ))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites=[]
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations)) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
elif primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list(set(primary_locations) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
else:
opportunistic_sites = []
wfh.sendLog('assignor',"We could be running in addition at %s"% sorted(opportunistic_sites))
if any([osite in SI.sites_not_ready for osite in opportunistic_sites]):
wfh.sendLog('assignor',"One of the usable site is in downtime %s"%([osite for osite in opportunistic_sites if osite in SI.sites_not_ready]))
down_time = True
## should this be send back to considered ?
## should be 2 but for the time-being let's lower it to get things going
copies_wanted,cpuh = wfh.getNCopies()
wfh.sendLog('assignor',"we need %s CPUh"%cpuh)
if cpuh>max_cpuh_block and not options.go:
#sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog('assignor','%s requires a large numbr of CPUh %s , not assigning, please check with requester'%( wfo.name, cpuh), level='critical')
wfh.sendLog('assignor',"Requiring a large number of CPUh %s, not assigning"%cpuh)
continue
if 'Campaign' in wfh.request and wfh.request['Campaign'] in CI.campaigns and 'maxcopies' in CI.campaigns[wfh.request['Campaign']]:
copies_needed_from_campaign = CI.campaigns[wfh.request['Campaign']]['maxcopies']
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(1,copies_wanted-less_copies_than_requested) # take one out for the efficiency
else:
## find out whether there is a site in the whitelist, that is lacking jobs and reduce to 1 copy needed to get things going
pass
wfh.sendLog('assignor',"needed availability fraction %s"% copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
## default back to white list to original white list with any data
wfh.sendLog('assignor',"Allowed sites :%s"% sorted(sites_allowed))
if primary_aaa:
## remove the sites not reachable localy if not in having the data
if not sites_all_data:
wfh.sendLog('assignor',"Overiding the primary on AAA setting to Off")
primary_aaa=False
else:
aaa_grid = set(sites_all_data)
for site in list(aaa_grid):
aaa_grid.update( aaa_mapping.get(site,[]) )
sites_allowed = list(set(initial_sites_allowed) & aaa_grid)
wfh.sendLog('assignor',"Selected to read primary through xrootd %s"%sorted(sites_allowed))
if not primary_aaa:
sites_allowed = sites_with_any_data
wfh.sendLog('assignor',"Selected for any data %s"%sorted(sites_allowed))
### check on endpoints for on-going transfers
if do_partial:
if endpoints:
end_sites = [SI.SE_to_CE(s) for s in endpoints]
sites_allowed = list(set(sites_allowed + end_sites))
if down_time and not any(osite in SI.sites_not_ready for osite in end_sites):
print "Flip the status of downtime, since our destinations are good"
down_time = False
print "with added endpoints",sorted(end_sites)
else:
print "Cannot do partial assignment without knowin the endpoints"
n_stalled+=1
continue
#if not len(sites_allowed):
# if not options.early:
# wfh.sendLog('assignor',"cannot be assign with no matched sites")
# sendLog('assignor','%s has no whitelist'% wfo.name, level='critical')
# n_stalled+=1
# continue
low_pressure = SI.sites_low_pressure(0.4)
## if any of the site allowed is low pressure : reduce to 1 copy so that it gets started
allowed_and_low = sorted(set(low_pressure) & set(sites_allowed))
if allowed_and_low:
wfh.sendLog('assignor',"The workflow can run at %s under low pressure currently"%( ','.join( allowed_and_low )))
copies_wanted = max(1., copies_wanted-1.)
if available_fractions and not all([available>=copies_wanted for available in available_fractions.values()]):
not_even_once = not all([available>=1. for available in available_fractions.values()])
above_good = all([available >= do_partial for available in available_fractions.values()])
wfh.sendLog('assignor',"The input dataset is not available %s times, only %s"%( copies_wanted, available_fractions.values()))
if down_time and not options.go and not options.early:
wfo.status = 'considered'
session.commit()
wfh.sendLog('assignor',"sending back to considered because of site downtime, instead of waiting")
#sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
sendLog('assignor','%s is not sufficiently available, due to down time of a site in the whitelist. sending back to considered.'%( wfo.name ), level='delay')
n_stalled+=1
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known and not options.limit and not options.go and not options.early and not (do_partial and above_good):
wfh.sendLog('assignor',"cannot be assigned, %s is not sufficiently available.\n %s"%(wfo.name,json.dumps(available_fractions)))
#sendEmail( "cannot be assigned","%s is not sufficiently available.\n %s"%(wfo.name,json.dumps(available_fractions)))
known.append( wfo.name )
open('cannot_assign.json','w').write(json.dumps( known, indent=2))
if options.early:
if wfo.status == 'considered':
wfh.sendLog('assignor',"setting considered-tried")
wfo.status = 'considered-tried'
session.commit()
else:
print "tried but status is",wfo.status
if do_partial and above_good:
print "Will move on with partial locations"
else:
n_stalled+=1
continue
if not len(sites_allowed):
if not options.early:
wfh.sendLog('assignor',"cannot be assign with no matched sites")
sendLog('assignor','%s has no whitelist'% wfo.name, level='critical')
n_stalled+=1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])]
wfh.sendLog('assignor',"Placing the output on %s"%sites_out)
parameters={
'SiteWhitelist' : sites_allowed,
'NonCustodialSites' : sites_out,
'AutoApproveSubscriptionSites' : list(set(sites_out)),
'AcquisitionEra' : wfh.acquisitionEra(),
'ProcessingString' : wfh.processingString(),
'MergedLFNBase' : set_lfn,
'ProcessingVersion' : version,
}
if primary_aaa:
parameters['TrustSitelists'] = True
wfh.sendLog('assignor',"Reading primary through xrootd at %s"%sorted(sites_allowed))
if secondary_aaa:
parameters['TrustPUSitelists'] = True
wfh.sendLog('assignor',"Reading secondary through xrootd at %s"%sorted(sites_allowed))
if 'parameters' in assign_parameters:
parameters.update( assign_parameters['parameters'] )
## plain assignment here
team='production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
if False and 'T2_CH_CERN' in parameters['SiteWhitelist']:
## add some check on
### the amount pending to HLT
### the size of the request
### the priority of the request (maybe not if we decide to overflow during runs)
parameters['SiteWhitelist'] = ['T2_CH_CERN_HLT']
team = 'hlt'
## reduce the splitting by factor of 4, regardless of type of splitting
sendEmail("sending work to HLT","%s was assigned to HLT"%wfo.name)
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v=getattr(options,key)
if v!=None:
if type(v)==str and ',' in v:
parameters[key] = filter(None,v.split(','))
else:
parameters[key] = v
if lheinput:
## throttle reading LHE article
wfh.sendLog('assignor', 'Setting the number of events per job to 500k max')
parameters['EventsPerJob'] = 500000
## pick up campaign specific assignment parameters
#parameters.update( CI.parameters(wfh.request['Campaign']) )
parameters.update( assign_parameters.get('parameters',{}) )
if not options.test:
parameters['execute'] = True
split_check = wfh.checkWorkflowSplitting()
if split_check!=True:
parameters.update( split_check )
if 'NoGo' in split_check.values():
wfh.sendLog('assignor', "Failing splitting check")
sendLog('assignor','the workflow %s is failing the splitting check. Verify in the logs'% wfo.name, level='critical')
n_stalled+=1
continue
if 'EventBased' in split_check.values():
wfh.sendLog('assignor', "Falling back to event splitting.")
#sendEmail("Fallback to EventBased","the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"%wfo.name)
sendLog('assignor','the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting ?'%wfo.name, level='critical')
## we have a problem here, that EventBased should never be used as a backup
if not options.go:
n_stalled+=1
continue
continue ## skip all together
elif 'EventsPerJob' in split_check.values():
wfh.sendLog('assignor', "Modifying the number of events per job")
#sendEmail("Modifying the job per events","the workflow %s is too heavy in number of jobs explosion"%wfo.name)
sendLog('assignor',"the workflow %s is too heavy in number of jobs explosion"%wfo.name, level='critical')
elif 'EventsPerLumi' in split_check.values():
wfh.sendLog('assignor', "Modifying the number of events per lumi to be able to process this")
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi)/float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents/(reqJobs*1.4))
lumisPerJob = int(eventsPerJob/eventsPerLumi)
if lumisPerJob==0:
#sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog('assignor', "%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob), level='critical')
wfh.sendLog('assignor', "%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl['events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl['avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
#sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog('assignor',"%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob), level='critical')
wfh.sendLog('assignor',"%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
#sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog('assignor',"leaving splitting untouched for %s, please check on %s"%( pstring, wfo.name), level='critical')
wfh.sendLog('assignor',"leaving splitting untouched for PU_RD*, please check.")
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned+=1
wfh.sendLog('assignor',"Properly assigned\n%s"%(json.dumps( parameters, indent=2)))
try:
## refetch information and lock output
new_wfi = workflowInfo( url, wfo.name)
(_,prim,_,sec) = new_wfi.getIO()
for secure in list(prim)+list(sec)+new_wfi.request['OutputDatasets']:
## lock all outputs flat
#NLI.lock( secure )
LI.lock( secure, reason = 'assigning')
#for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output",str(e))
else:
wfh.sendLog('assignor',"Failed to assign. Please check the logs")
print "ERROR could not assign",wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor',"Assigned %d Stalled %s"%(n_assigned, n_stalled))
def stagor(url,specific =None, options=None):
if not componentInfo().check(): return
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
done_by_wf_id = {}
done_by_input = {}
completion_by_input = {}
good_enough = 100.0
lost = json.loads(open('lost_blocks_datasets.json').read())
still_lost = []
for dataset in lost:
l = findLostBlocks(url ,dataset)
if not l:
print dataset,"is not really lost"
else:
still_lost.append( dataset )
open('lost_blocks_datasets.json','w').write( json.dumps( still_lost, indent=2) )
if options.fast:
print "doing the fast check of staged with threshold:",options.goodavailability
for wfo in session.query(Workflow).filter(Workflow.status == 'staging').all():
if specific and not specific in wfo.name: continue
wfi = workflowInfo(url, wfo.name)
sites_allowed = getSiteWhiteList( wfi.getIO() )
if 'SiteWhitelist' in CI.parameters(wfi.request['Campaign']):
sites_allowed = CI.parameters(wfi.request['Campaign'])['SiteWhitelist']
if 'SiteBlacklist' in CI.parameters(wfi.request['Campaign']):
sites_allowed = list(set(sites_allowed) - set(CI.parameters(wfi.request['Campaign'])['SiteBlacklist']))
_,primaries,_,secondaries = wfi.getIO()
se_allowed = [SI.CE_to_SE(site) for site in sites_allowed]
all_check = True
for dataset in list(primaries):#+list(secondaries) ?
#print se_allowed
available = getDatasetBlocksFraction( url , dataset , sites=se_allowed )
all_check &= (available >= options.goodavailability)
if not all_check: break
if all_check:
print "\t\t",wfo.name,"can go staged"
wfo.status = 'staged'
session.commit()
else:
print "\t",wfo.name,"can wait a bit more"
return
for wfo in session.query(Workflow).filter(Workflow.status == 'staging').all():
wfi = workflowInfo(url, wfo.name)
_,primaries,_,secondaries = wfi.getIO()
for dataset in list(primaries)+list(secondaries):
done_by_input[dataset] = {}
completion_by_input[dataset] = {}
print wfo.name,"needs",dataset
for transfer in session.query(Transfer).all():
if specific and str(transfer.phedexid)!=str(specific): continue
skip=True
for wfid in transfer.workflows_id:
tr_wf = session.query(Workflow).get(wfid)
if tr_wf:
if tr_wf.status == 'staging':
print "\t",transfer.phedexid,"is staging for",tr_wf.name
skip=False
if skip: continue
if transfer.phedexid<0: continue
## check the status of transfers
checks = checkTransferApproval(url, transfer.phedexid)
approved = all(checks.values())
if not approved:
print transfer.phedexid,"is not yet approved"
approveSubscription(url, transfer.phedexid)
continue
## check on transfer completion
checks = checkTransferStatus(url, transfer.phedexid, nocollapse=True)
if not specific:
for dsname in checks:
if not dsname in done_by_input: done_by_input[dsname]={}
if not dsname in completion_by_input: completion_by_input[dsname] = {}
done_by_input[dsname][transfer.phedexid]=all(map(lambda i:i>=good_enough, checks[dsname].values()))
completion_by_input[dsname][transfer.phedexid]=checks[dsname].values()
if checks:
print "Checks for",transfer.phedexid,[node.values() for node in checks.values()]
done = all(map(lambda i:i>=good_enough,list(itertools.chain.from_iterable([node.values() for node in checks.values()]))))
else:
## it is empty, is that a sign that all is done and away ?
print "ERROR with the scubscriptions API of ",transfer.phedexid
print "Most likely something else is overiding the transfer request. Need to work on finding the replacement automatically, if the replacement exists"
done = False
## the thing above is NOT giving the right number
#done = False
for wfid in transfer.workflows_id:
tr_wf = session.query(Workflow).get(wfid)
if tr_wf:# and tr_wf.status == 'staging':
if not tr_wf.id in done_by_wf_id: done_by_wf_id[tr_wf.id]={}
done_by_wf_id[tr_wf.id][transfer.phedexid]=done
if done:
## transfer.status = 'done'
print transfer.phedexid,"is done"
else:
print transfer.phedexid,"not finished"
pprint.pprint( checks )
#print done_by_input
print "\n----\n"
for dsname in done_by_input:
fractions = None
if dsname in completion_by_input:
fractions = itertools.chain.from_iterable([check.values() for check in completion_by_input.values()])
## the workflows in the waiting room for the dataset
using_its = getWorkflowByInput(url, dsname)
#print using_its
using_wfos = []
for using_it in using_its:
wf = session.query(Workflow).filter(Workflow.name == using_it).first()
if wf:
using_wfos.append( wf )
if not len(done_by_input[dsname]):
print "For dataset",dsname,"there are no transfer report. That's an issue."
for wf in using_wfos:
if wf.status == 'staging':
if UC.get("stagor_sends_back"):
print "sending",wf.name,"back to considered"
wf.status = 'considered'
session.commit()
sendEmail( "send back to considered","%s was send back and might be trouble"% wf.name)
else:
print "would send",wf.name,"back to considered"
sendEmail( "subscription lagging behind","susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name)
continue
#need_sites = int(len(done_by_input[dsname].values())*0.7)+1
need_sites = len(done_by_input[dsname].values())
#if need_sites > 10: need_sites = int(need_sites/2.)
got = done_by_input[dsname].values().count(True)
if all([wf.status != 'staging' for wf in using_wfos]):
## not a single ds-using wf is in staging => moved on already
## just forget about it
print "presence of",dsname,"does not matter anymore"
print "\t",done_by_input[dsname]
print "\t",[wf.status for wf in using_wfos]
print "\tneeds",need_sites
continue #??
## should the need_sites reduces with time ?
# with dataset choping, reducing that number might work as a block black-list.
if len(done_by_input[dsname].values()) and all(done_by_input[dsname].values()):
print dsname,"is everywhere we wanted"
## the input dataset is fully transfered, should consider setting the corresponding wf to staged
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is with us. setting staged and move on"
wf.status = 'staged'
session.commit()
elif fractions and len(list(fractions))>1 and set(fractions)==1:
print dsname,"is everywhere at the same fraction"
print "We do not want this in the end. we want the data we asked for"
continue
## the input dataset is fully transfered, should consider setting the corresponding wf to staged
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is with us everywhere the same. setting staged and move on"
wf.status = 'staged'
session.commit()
elif got >= need_sites:
print dsname,"is almost everywhere we wanted"
#print "We do not want this in the end. we want the data we asked for"
#continue
## the input dataset is fully transfered, should consider setting the corresponding wf to staged
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is almost with us. setting staged and move on"
wf.status = 'staged'
session.commit()
else:
print "incomplete",dsname
lost = findLostBlocks(url, dsname)
try:
known_lost = json.loads(open('lost_blocks_datasets.json').read())
except:
print "enable to get the known_lost from local json file"
known_lost = []
if lost and not dsname in known_lost:
lost_names = [item['name'] for item in lost]
## make a deeper investigation of the block location to see whether it's really no-where no-where
print "We have lost",len(lost),"blocks",lost_names
#print json.dumps( lost , indent=2 )
sendEmail('we have lost a few blocks', str(len(lost))+" in total.\nDetails \n:"+json.dumps( lost , indent=2 ))
known_lost.append(dsname)
rr= open('lost_blocks_datasets.json','w')
rr.write( json.dumps( known_lost, indent=2))
rr.close()
## should the status be change to held-staging and pending on a ticket
print "\t",done_by_input[dsname]
print "\tneeds",need_sites
print "\tgot",got
for wfid in done_by_wf_id:
#print done_by_wf_id[wfid].values()
## ask that all related transfer get into a valid state
if all(done_by_wf_id[wfid].values()):
pass
def injector(url, options, specific):
mlock = moduleLock()
if mlock(): return
use_mcm = True
up = componentInfo(soft=['mcm','wtc','jira'] )
if not up.check(): return
use_mcm = up.status['mcm']
UC = unifiedConfiguration()
transform_keywords = UC.get('convert_to_stepchain')
workflows = getWorkflows(url, status=options.wmstatus, user=options.user)
for user in UC.get("user_rereco"):
workflows.extend( getWorkflows(url, status=options.wmstatus, user=user, rtype="ReReco"))
for user in (options.user_relval.split(',') if options.user_relval else UC.get("user_relval")) :
workflows.extend( getWorkflows(url, status=options.wmstatus, user=user, rtype="TaskChain"))
for user in (options.user_storeresults.split(',') if options.user_storeresults else UC.get("user_storeresults")) :
workflows.extend( getWorkflows(url, status=options.wmstatus, user=user, rtype="StoreResults"))
print len(workflows),"in line"
cannot_inject = set()
to_convert = set()
status_cache = defaultdict(str)
## browse for assignment-approved requests, browsed for ours, insert the diff
for wf in workflows:
if specific and not specific in wf: continue
exists = session.query(Workflow).filter(Workflow.name == wf ).first()
if not exists:
wfi = workflowInfo(url, wf)
## check first that there isn't related here with something valid
can_add = True
## first try at finding a match
familly = session.query(Workflow).filter(Workflow.name.contains(wfi.request['PrepID'])).all()
if not familly:
pids = wfi.getPrepIDs()
req_familly = []
for pid in pids:
req_familly.extend( getWorkflowById( url, pid, details=True) )
familly = []
print len(req_familly),"members"
for req_member in req_familly:
#print "member",req_member['RequestName']
owfi = workflowInfo(url, req_member['RequestName'], request=req_member)
other_pids = owfi.getPrepIDs()
if set(pids) == set(other_pids):
## this is a real match
familly.extend( session.query(Workflow).filter(Workflow.name == req_member['RequestName']).all() )
for lwfo in familly:
if lwfo:
## we have it already
if not lwfo.status in ['forget','trouble','forget-unlock','forget-out-unlock']:
wfi.sendLog('injector',"Should not put %s because of %s %s"%( wf, lwfo.name,lwfo.status ))
sendLog('injector',"Should not put %s because of %s %s"%( wf, lwfo.name,lwfo.status ), level='critical')
print "Should not put",wf,"because of",lwfo.name,lwfo.status
cannot_inject.add( wf )
can_add = False
## add a check on validity of input datasets
_,prim,par,sec = wfi.getIO()
for d in list(prim)+list(par)+list(sec):
if not d in status_cache:
status_cache[d] = getDatasetStatus(d)
if status_cache[d] != 'VALID':
wfi.sendLog('injector',"One of the input is not VALID. %s : %s"%( d, status_cache[d]))
sendLog('injector',"One of the input of %s is not VALID. %s : %s"%( wf, d, status_cache[d]), level='critical')
can_add = False
#else:
# ##make sure that all blocks get closed
# closeAllBlocks(url, d)
## check for any file in phedex, to verify existence
_,ph_files,_,_ = getDatasetFiles(url, d)
if not ph_files and not ( 'StoreResults' == wfi.request.setdefault('RequestType',None) ):
wfi.sendLog('injector',"One of the input has no file in phedex: %s" % d )
sendLog('injector',"One of the input has no file in phedex: %s"% d, level='critical')
can_add = False
### ban some workflow that you don't like anymore
#outputs = wfi.request['OutputDatasets']
if not can_add: continue
## temporary hack to transform specific taskchain into stepchains
good_for_stepchain = wfi.isGoodToConvertToStepChain( keywords = transform_keywords)
#good_for_stepchain = wfi.isGoodToConvertToStepChain( keywords = None)
## match keywords and technical constraints
if (not options.no_convert) and good_for_stepchain and not wfi.isRelval():
to_convert.add( wf )
wfi.sendLog('injector','Transforming %s TaskChain into StepChain'%wf)
sendEmail('convertion to stepchain','Transforming %s TaskChain into StepChain'%wf)
wfi.sendLog('injector',"considering %s"%wf)
new_wf = Workflow( name = wf , status = options.setstatus, wm_status = options.wmstatus)
session.add( new_wf )
session.commit()
time.sleep(0.5)
else:
#print "already have",wf
pass
if cannot_inject:
#sendEmail('workflow duplicates','These workflow cannot be added in because of duplicates \n\n %s'%( '\n'.join(cannot_inject)))
sendLog('injector','These workflow cannot be added in because of duplicates \n\n %s'%( '\n'.join(cannot_inject)), level='critical')
for wf in to_convert:
os.system('./Unified/rejector.py --clone --to_step --comments \"Transform to StepChain\" %s'% wf)
## passing a round of invalidation of what needs to be invalidated
if use_mcm and (options.invalidate or True):
invalidator(url)
no_replacement = set()
#print "getting all transfers"
#all_transfers=session.query(Transfer).all()
#print "go!"
## pick up replacements
for wf in session.query(Workflow).filter(Workflow.status == 'trouble').all():
print wf.name
if specific and not specific in wf.name: continue
print wf.name
wfi = workflowInfo(url, wf.name )
wl = wfi.request #getWorkLoad(url, wf.name)
familly = getWorkflowById( url, wl['PrepID'] )
true_familly = []
for member in familly:
if member == wf.name: continue
fwl = getWorkLoad(url , member)
if options.replace:
if member != options.replace: continue
else:
if fwl['RequestDate'] < wl['RequestDate']: continue
if fwl['RequestType']=='Resubmission': continue
if fwl['RequestStatus'] in ['None',None,'new']: continue
if fwl['RequestStatus'] in ['rejected','rejected-archived','aborted','aborted-archived']: continue
true_familly.append( fwl )
if len(true_familly)==0:
#sendLog('injector','%s had no replacement'%wf.name, level='critical')
if wfi.isRelval():
#wfi.sendLog('injector','the workflow was found in trouble with no replacement. As a relval, there is no clean way to handle this.')
wfi.sendLog('injector','the workflow was found in trouble with no replacement. As a relval, there is no clean way to handle this. Setting forget')
wf.status = 'forget'
session.commit()
else:
wfi.sendLog('injector','the workflow was found in trouble with no replacement')
no_replacement.add( wf.name )
continue
else:
wfi.sendLog('injector','the workflow was found in trouble and has a replacement')
print wf.name,"has",len(familly),"familly members"
print wf.name,"has",len(true_familly),"true familly members"
##we cannot have more than one of them !!! pick the last one
if len(true_familly)>1:
#sendEmail('multiple wf','please take a look at injector for %s'%wf.name)
sendLog('injector','Multiple wf in line, will take the last one for %s \n%s'%( wf.name, ', '.join(fwl['RequestName'] for fwl in true_familly)), level='critical')
for fwl in true_familly[-1:]:
member = fwl['RequestName']
new_wf = session.query(Workflow).filter(Workflow.name == member).first()
if not new_wf:
sendLog('injector',"putting %s as replacement of %s"%( member, wf.name))
status = 'away'
if fwl['RequestStatus'] in ['assignment-approved']:
status = 'considered'
new_wf = Workflow( name = member, status = status, wm_status = fwl['RequestStatus'])
wf.status = 'forget'
session.add( new_wf )
else:
if new_wf.status == 'forget': continue
sendLog('injector',"getting %s as replacement of %s"%( new_wf.name, wf.name ))
wf.status = 'forget'
for tr in session.query(TransferImp).filter( TransferImp.workflow_id == wf.id).all():
## get all transfer working for the old workflow
existing = session.query(TransferImp).filter( TransferImp.phedexid == tr.phedexid).filter( TransferImp.workflow_id == new_wf.id).all()
tr.active = False ## disable the old one
if not existing:
## create the transfer object for the new dependency
tri = TransferImp( phedexid = tr.phedexid,
workflow = new_wf)
session.add( tri )
session.commit()
## don't do that automatically
#wf.status = 'forget'
session.commit()
if no_replacement:
#sendEmail('workflow with no replacement','%s \n are dangling there'%( '\n'.join(no_replacement)))
sendLog('injector','workflow with no replacement\n%s \n are dangling there'% ( '\n'.join(no_replacement)), level='critical')
def assignor(url, specific=None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
#if notRunningBefore( 'stagor' ): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = global_SI
#LI = lockInfo()
NLI = newLockInfo()
n_assigned = 0
n_stalled = 0
wfos = []
if specific or options.early:
wfos.extend(
session.query(Workflow).filter(
Workflow.status == 'considered').all())
wfos.extend(
session.query(Workflow).filter(Workflow.status == 'staging').all())
if specific:
wfos.extend(
session.query(Workflow).filter(
Workflow.status == 'considered-tried').all())
wfos.extend(
session.query(Workflow).filter(Workflow.status == 'staged').all())
#if specific:
# #wfos = session.query(Workflow).filter(Workflow.name==specific).all()
# wfos = session.query(Workflow).filter(Workflow.name.contains(specific)).all()
#if not wfos:
# if specific:
# wfos = session.query(Workflow).filter(Workflow.status=='considered').all()
# wfos.extend( session.query(Workflow).filter(Workflow.status=='staging').all())
# wfos.extend(session.query(Workflow).filter(Workflow.status=='staged').all())
random.shuffle(wfos)
for wfo in wfos:
if options.limit and (n_stalled + n_assigned) > options.limit:
break
if specific:
if not any(map(lambda sp: sp in wfo.name, specific.split(','))):
continue
#if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo(url, wfo.name)
wfh.sendLog('assignor', "%s to be assigned" % wfo.name)
## check if by configuration we gave it a GO
if not CI.go(wfh.request['Campaign']) and not options.go:
wfh.sendLog('assignor', "No go for %s" % wfh.request['Campaign'])
n_stalled += 1
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] != 'assignment-approved':
if not options.test:
wfh.sendLog('assignor',
"setting %s away and skipping" % wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name, wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version = wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor', "cannot decide on version number")
n_stalled += 1
wfo.status = 'trouble'
session.commit()
continue
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput, primary, parent, secondary,
sites_allowed) = wfh.getSiteWhiteList()
original_sites_allowed = copy.deepcopy(sites_allowed)
wfh.sendLog('assignor', "Site white list %s" % sorted(sites_allowed))
override_sec_location = CI.get(wfh.request['Campaign'],
'SecondaryLocation', [])
blocks = []
if 'BlockWhitelist' in wfh.request:
blocks = wfh.request['BlockWhitelist']
if 'RunWhitelist' in wfh.request and wfh.request['RunWhitelist']:
## augment with run white list
for dataset in primary:
blocks = list(
set(blocks + getDatasetBlocks(
dataset, runs=wfh.request['RunWhitelist'])))
wfh.sendLog('assignor', "Allowed %s" % sorted(sites_allowed))
secondary_locations = None
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
sendEmail("tempting to pass sec location check",
"but we cannot yet IMO")
#pass
presence = getDatasetPresence(url, sec)
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [
site for (site, (there, frac)) in presence.items()
if frac > 98.
]
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations == None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(
set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [
site for site in sites_allowed
if SI.CE_to_SE(site) in one_secondary_locations
]
wfh.sendLog(
'assignor', "From secondary requirement, now Allowed%s" %
sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy(
sites_allowed
) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy(sites_allowed)
sites_with_data = copy.deepcopy(sites_allowed)
sites_with_any_data = copy.deepcopy(sites_allowed)
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
for prim in list(primary):
set_lfn = getLFNbase(prim)
presence = getDatasetPresence(url, prim, only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(
url,
prim,
sites=[SI.CE_to_SE(site) for site in sites_allowed],
only_blocks=blocks)
#sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
#sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in [
psite for (psite, (there, frac)) in presence.items()
if there
]
]
sites_with_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in
[psite for (psite, frac) in presence.items() if frac[1] > 90.]
]
sites_with_any_data = [
site for site in sites_with_any_data
if SI.CE_to_SE(site) in presence.keys()
]
wfh.sendLog(
'assignor', "Holding the data but not allowed %s" % sorted(
list(
set([
se_site for se_site in presence.keys()
if not SI.SE_to_CE(se_site) in sites_allowed
]))))
if primary_locations == None:
primary_locations = presence.keys()
else:
primary_locations = list(
set(primary_locations) & set(presence.keys()))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites = []
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in
list((set(secondary_locations) & set(primary_locations)) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
elif primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in list(
set(primary_locations) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
else:
opportunistic_sites = []
wfh.sendLog(
'assignor', "We could be running in addition at %s" %
sorted(opportunistic_sites))
if any(
[osite in SI.sites_not_ready
for osite in opportunistic_sites]):
wfh.sendLog(
'assignor', "One of the usable site is in downtime %s" % ([
osite in SI.sites_not_ready
for osite in opportunistic_sites
]))
down_time = True
## should this be send back to considered ?
"""
if available_fractions and not all([available>=1. for available in available_fractions.values()]):
print "The input dataset is not located in full over sites"
print json.dumps(available_fractions)
if not options.test and not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known:
sendEmail( "cannot be assigned","%s is not full over sites \n %s"%(wfo.name,json.dumps(available_fractions)))
known.append( wfo.name )
open('cannot_assign.json','w').write(json.dumps( known, indent=2))
n_stalled+=1
continue ## skip skip skip
"""
## should be 2 but for the time-being let's lower it to get things going
copies_wanted, cpuh = wfh.getNCopies()
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns and 'maxcopies' in CI.campaigns[
wfh.request['Campaign']]:
copies_needed_from_campaign = CI.campaigns[
wfh.request['Campaign']]['maxcopies']
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(
1, copies_wanted -
less_copies_than_requested) # take one out for the efficiency
wfh.sendLog('assignor',
"needed availability fraction %s" % copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
if available_fractions and not all([
available >= copies_wanted
for available in available_fractions.values()
]):
not_even_once = not all([
available >= 1. for available in available_fractions.values()
])
wfh.sendLog(
'assignor',
"The input dataset is not available %s times, only %s" %
(copies_wanted, available_fractions.values()))
if down_time and not options.go and not options.early:
wfo.status = 'considered'
session.commit()
wfh.sendLog(
'assignor',
"sending back to considered because of site downtime, instead of waiting"
)
sendEmail(
"cannot be assigned due to downtime",
"%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."
% wfo.name)
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known and not options.limit and not options.go and not options.early:
wfh.sendLog(
'assignor',
"cannot be assigned, %s is not sufficiently available.\n %s"
% (wfo.name, json.dumps(available_fractions)))
sendEmail(
"cannot be assigned",
"%s is not sufficiently available.\n %s" %
(wfo.name, json.dumps(available_fractions)))
known.append(wfo.name)
open('cannot_assign.json',
'w').write(json.dumps(known, indent=2))
n_stalled += 1
if options.early:
if wfo.status == 'considered':
wfh.sendLog('assignor', "setting considered-tried")
wfo.status = 'considered-tried'
session.commit()
else:
print "tried but status is", wfo.status
continue
## default back to white list to original white list with any data
print "Allowed", sites_allowed
if options.primary_aaa:
sites_allowed = initial_sites_allowed
#options.useSiteListAsLocation = True
options.TrustSitelists = True
else:
sites_allowed = sites_with_any_data
wfh.sendLog('assignor',
"Selected for any data %s" % sorted(sites_allowed))
if options.restrict:
print "Allowed", sites_allowed
sites_allowed = sites_with_any_data
print "Selected", sites_allowed
else:
if set(sites_with_data) != set(sites_allowed):
## the data is not everywhere we wanted to run at : enable aaa
print "Sites with 90% data not matching site white list (block choping!)"
print "Resorting to AAA reading for", list(
set(sites_allowed) - set(sites_with_data)), "?"
print "Whitelist site with any data", list(
set(sites_allowed) - set(sites_with_any_data))
#options.useSiteListAsLocation = True
#print "Not commissioned yet"
#continue
#print "We could be running at",opportunistic_sites,"in addition"
##sites_allowed = list(set(sites_allowed+ opportunistic_sites))
if not len(sites_allowed):
wfh.sendLog('assignor', "cannot be assign with no matched sites")
sendEmail("cannot be assigned", "%s has no whitelist" % (wfo.name))
n_stalled += 1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [
SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])
]
## one last modification now that we know we can assign, and to make sure all ressource can be used by the request : set all ON sites to whitelist
###sites_allowed = original_sites_allowed ## not needed, afterall as secondary jobs go their own ways
wfh.sendLog('assignor', "Placing the output on %s" % sites_out)
parameters = {
'SiteWhitelist': sites_allowed,
#'CustodialSites' : sites_custodial,
'NonCustodialSites': sites_out,
'AutoApproveSubscriptionSites': list(set(sites_out)),
'AcquisitionEra': wfh.acquisitionEra(),
'ProcessingString': wfh.processingString(),
'MergedLFNBase': set_lfn,
'ProcessingVersion': version,
}
## plain assignment here
team = 'production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
## high priority team agent
#if wfh.request['RequestPriority'] >= 100000 and (wfh.request['TimePerEvent']*int(wfh.getRequestNumEvents()))/(8*3600.) < 10000:
# team = 'highprio'
# sendEmail("sending work with highprio team","%s"% wfo.name, destination=['*****@*****.**'])
## SDSC redirection
#if "T2_US_UCSD" in sites_with_data and random.random() < -0.5 and wfh.request['Campaign']=='RunIISpring15DR74' and int(wfh.getRequestNumEvents()) < 600000 and not any([out.endswith('RAW') for out in wfh.request['OutputDatasets']]):
# ## consider SDSC
# parameters['SiteWhitelist'] = ['T2_US_UCSD','T3_US_SDSC']
# parameters['useSiteListAsLocation'] = True
# team = 'allocation-based'
# sendEmail("sending work to SDSC","%s was assigned to SDSC/UCSD"% wfo.name, destination=['*****@*****.**'])
## SDSC redirection
#if wfh.request['Campaign']==R'unIIWinter15GS' and random.random() < -1.0:
# parameters['SiteWhitelist'] = ['T3_US_SDSC']
# team = 'allocation-based'
# sendEmail("sending work to SDSC","%s was assigned to SDSC"% wfo.name, destination=['*****@*****.**'])
if False and 'T2_CH_CERN' in parameters['SiteWhitelist']:
## add some check on
### the amount pending to HLT
### the size of the request
### the priority of the request (maybe not if we decide to overflow during runs)
parameters['SiteWhitelist'] = ['T2_CH_CERN_HLT']
team = 'hlt'
## reduce the splitting by factor of 4, regardless of type of splitting
sendEmail("sending work to HLT",
"%s was assigned to HLT" % wfo.name)
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v = getattr(options, key)
if v != None:
if type(v) == str and ',' in v:
parameters[key] = filter(None, v.split(','))
else:
parameters[key] = v
## pick up campaign specific assignment parameters
parameters.update(CI.parameters(wfh.request['Campaign']))
if not options.test:
parameters['execute'] = True
split_check = wfh.checkWorkflowSplitting()
if split_check != True:
parameters.update(split_check)
if 'EventBased' in split_check.values():
wfh.sendLog('assignor', "Falling back to event splitting.")
sendEmail(
"Fallback to EventBased",
"the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"
% wfo.name)
elif 'EventsPerJob' in split_check.values():
wfh.sendLog('assignor',
"Modifying the number of job per event")
sendEmail(
"Modifying the job per events",
"the workflow %s is too heavy in number of jobs explosion"
% wfo.name)
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi) / float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents / (reqJobs * 1.4))
lumisPerJob = int(eventsPerJob / eventsPerLumi)
if lumisPerJob == 0:
sendEmail(
"issue with event splitting for run-dependent MC",
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob))
wfh.sendLog(
'assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl[
'events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl[
'avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
sendEmail(
"setting lumi splitting for run-dependent MC",
"%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob))
wfh.sendLog(
'assignor', "%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
sendEmail("leaving splitting untouched for PU_RD*",
"please check on " + wfo.name)
wfh.sendLog(
'assignor',
"leaving splitting untouched for PU_RD*, please check."
)
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned += 1
try:
## refetch information and lock output
new_wfi = workflowInfo(url, wfo.name)
(_, prim, _, sec) = new_wfi.getIO()
for secure in list(prim) + list(
sec) + new_wfi.request['OutputDatasets']:
## lock all outputs flat
NLI.lock(secure)
#for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output", str(e))
else:
print "ERROR could not assign", wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor', "Assigned %d Stalled %s" % (n_assigned, n_stalled))
def parse_one(url, wfn, options=None):
def time_point(label="", sub_lap=False):
now = time.mktime(time.gmtime())
nows = time.asctime(time.gmtime())
print "[showError] Time check (%s) point at : %s" % (label, nows)
print "[showError] Since start: %s [s]" % (now - time_point.start)
if sub_lap:
print "[showError] Sub Lap : %s [s]" % (now - time_point.sub_lap)
time_point.sub_lap = now
else:
print "[showError] Lap : %s [s]" % (now - time_point.lap)
time_point.lap = now
time_point.sub_lap = now
time_point.sub_lap = time_point.lap = time_point.start = time.mktime(
time.gmtime())
task_error_site_count = {}
one_explanation = defaultdict(set)
per_task_explanation = defaultdict(set)
if wfn in [
'vlimant_task_EXO-RunIISummer15wmLHEGS-04800__v1_T_170906_141738_1357'
]:
return task_error_site_count, one_explanation
time_point("Starting with %s" % wfn)
threads = []
SI = global_SI()
UC = unifiedConfiguration()
wfi = workflowInfo(url, wfn)
time_point("wfi", sub_lap=True)
where_to_run, missing_to_run, missing_to_run_at = wfi.getRecoveryInfo()
time_point("acdcinfo", sub_lap=True)
all_blocks, needed_blocks_loc, files_in_blocks, files_and_loc_notin_dbs = wfi.getRecoveryBlocks(
)
time_point("inputs", sub_lap=True)
ancestor = workflowInfo(url, wfn)
lhe, prim, _, sec = ancestor.getIO()
high_order_acdc = 0
while ancestor.request['RequestType'] == 'Resubmission':
ancestor = workflowInfo(url, ancestor.request['OriginalRequestName'])
lhe, prim, _, sec = ancestor.getIO()
high_order_acdc += 1
no_input = (not lhe) and len(prim) == 0 and len(sec) == 0
cache = options.cache
print "cache timeout", cache
err = wfi.getWMErrors(cache=cache)
time_point("wmerrors", sub_lap=True)
stat = wfi.getWMStats(cache=cache)
time_point("wmstats", sub_lap=True)
#adcd = wfi.getRecoveryDoc()
total_by_code_dash = defaultdict(int)
total_by_site_dash = defaultdict(int)
r_dashb = defaultdict(lambda: defaultdict(int))
dash_board_h = 1
if False:
## NB get the since from when the wf has started, not a fixed value
## no dashboard until we get a better api
#dashb = wfi.getFullPicture(since=dash_board_h,cache=cache)
dashb = {}
#print json.dumps( dashb , indent=2)
for site, sinfo in dashb.items():
for s_code, counts in sinfo.items():
d_statuses = ['submitted', 'pending', 'app-unknown', 'done']
total_by_code_dash[str(s_code)] += counts.get('submitted', 0)
total_by_site_dash[site] += counts.get('submitted', 0)
r_dashb[str(s_code)][site] += counts.get('submitted', 0)
print json.dumps(total_by_code_dash, indent=2)
print json.dumps(total_by_site_dash, indent=2)
time_point("Got most input")
status_per_task = defaultdict(lambda: defaultdict(int))
if not 'AgentJobInfo' in stat:
stat['AgentJobInfo'] = {}
#print "bad countent ?"
#print json.dumps( stat, indent=2)
for agent in stat['AgentJobInfo']:
for task in stat['AgentJobInfo'][agent]['tasks']:
if not 'status' in stat['AgentJobInfo'][agent]['tasks'][task]:
continue
for status in stat['AgentJobInfo'][agent]['tasks'][task]['status']:
info = stat['AgentJobInfo'][agent]['tasks'][task]['status'][
status]
#print status,stat['AgentJobInfo'][agent]['tasks'][task]['status'][status]
if type(info) == dict:
status_per_task[task][status] += sum(
stat['AgentJobInfo'][agent]['tasks'][task]['status']
[status].values())
else:
status_per_task[task][status] += stat['AgentJobInfo'][
agent]['tasks'][task]['status'][status]
#print json.dumps( status_per_task, indent=2)
db_total_per_site = defaultdict(int)
db_total_per_code = defaultdict(int)
## cannot do that since there is no task count in dashboard and we have to take away the submitted
#for site in dashb:
# for error in dashb[site]:
# db_total_per_site[site] += dashb[site][error]
# db_total_per_code[code] += dashb[site][error]
print "ACDC Information"
print "\t where to re-run"
print json.dumps(where_to_run, indent=2)
print "\t Missing events"
print json.dumps(missing_to_run, indent=2)
print "\t Missing events per site"
print json.dumps(missing_to_run_at, indent=2)
if not where_to_run and not missing_to_run and not missing_to_run_at:
print "showError is unable to run"
#return task_error_site_count, one_explanation
pass
do_JL = not options.no_JL
do_CL = not options.no_CL
do_all_error_code = options.all_errors
if high_order_acdc >= 1:
print high_order_acdc, "order request, pulling down all logs"
do_all_error_code = True
if wfi.isRelval():
print "getting all codes for relval"
do_all_error_code = True
tasks = sorted(set(err.keys() + missing_to_run.keys()))
if not tasks:
print "no task to look at"
#return task_error_site_count
html = "<html> <center><h1>%s, Updated on %s (GMT)" % (
wfn, time.asctime(time.gmtime()))
html += '</center>'
html += '<a href=https://cmsweb.cern.ch/reqmgr2/fetch?rid=%s>dts</a>, ' % (
wfn)
html += '<a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?prep_id=%s>ac</a>, ' % (
wfi.request['PrepID'])
html += '<a href=https://cms-gwmsmon.cern.ch/prodview/%s>Job Progress</a>, ' % (
wfn)
r_type = wfi.request.get('OriginalRequestType',
wfi.request.get('RequestType', 'NaT'))
if r_type in ['ReReco']:
html += '<a href=../datalumi/lumi.%s.html>Lumisection Summary</a>, ' % wfi.request[
'PrepID']
html += '<a href="https://its.cern.ch/jira/issues/?jql=text~%s AND project = CMSCOMPPR" target="_blank">jira</a>' % (
wfi.request['PrepID'])
html += '<hr>'
html += '<a href=#IO>I/O</a>, <a href=#ERROR>Errors</a>, <a href=#BLOCK>blocks</a>, <a href=#FILE>files</a>, <a href=#CODES>Error codes</a><br>'
html += '<hr>'
time_point("Header writen")
html += '<a name=IO></a>'
if prim:
html += 'Reads in primary<br>'
rwl = wfi.getRunWhiteList()
lwl = wfi.getLumiWhiteList()
for dataset in prim:
html += '<b>%s </b>(events/lumi ~%d)' % (
dataset, getDatasetEventsPerLumi(dataset))
blocks = getDatasetBlocks(dataset, runs=rwl) if rwl else None
blocks = getDatasetBlocks(dataset, lumis=lwl) if lwl else None
available = getDatasetBlocksFraction(url,
dataset,
only_blocks=blocks)
html += '<br><br>Available %.2f (>1 more than one copy, <1 not in full on disk)<br>' % available
html += '<ul>'
presence = getDatasetPresence(url, dataset, only_blocks=blocks)
for site in sorted(presence.keys()):
html += '<li>%s : %.2f %%' % (site, presence[site][1])
html += '</ul><br>'
if sec:
html += 'Reads in secondary<br>'
for dataset in sec:
presence = getDatasetPresence(url, dataset)
html += '<b>%s</b><ul>' % dataset
for site in sorted(presence.keys()):
html += '<li>%s : %.2f %%' % (site, presence[site][1])
html += '</ul>'
outs = sorted(wfi.request['OutputDatasets'])
if outs:
html += 'Produces<br>'
for dataset in outs:
presence = getDatasetPresence(url, dataset)
html += '<b>%s </b>(events/lumi ~ %d)<ul>' % (
dataset, getDatasetEventsPerLumi(dataset))
for site in sorted(presence.keys()):
html += '<li>%s : %.2f %%' % (site, presence[site][1])
html += '</ul>'
time_point("Input checked")
html += """
<hr><br>
<a name=ERROR></a>
<ul>
<li> <b><i>dashboard numbers over %d days</b></i>
<li> ↑ %% with respect to total number of error in the code
<li> → %% with respect to total number of error at the site
</ul>
""" % (dash_board_h)
html += '<br>'
n_expose_base = options.expose # if options else UC.get('n_error_exposed')
print "getting", n_expose_base, "logs by default"
if tasks:
min_rank = min([task.count('/') for task in tasks])
for task in tasks:
n_expose = n_expose_base
expose_archive_code = dict([(str(code), defaultdict(lambda: n_expose))
for code in UC.get('expose_archive_code')])
expose_condor_code = dict([(str(code), defaultdict(lambda: n_expose))
for code in UC.get('expose_condor_code')])
#print task
task_rank = task.count('/')
task_short = task.split('/')[-1]
total_per_site = defaultdict(int)
time_point("Starting with task %s" % task_short, sub_lap=True)
notreported = 'NotReported'
total_count = defaultdict(int)
error_site_count = defaultdict(lambda: defaultdict(int))
all_not_reported = set()
for agent in stat['AgentJobInfo']:
for site in stat['AgentJobInfo'][agent]['tasks'].get(task, {}).get(
'skipped', {}):
info = stat['AgentJobInfo'][agent]['tasks'][task]['skipped'][
site]
#print info
all_not_reported.add(site)
ce = SI.SE_to_CE(site)
error_site_count[notreported][ce] += info.get(
'skippedFiles', 0)
total_count[notreported] += info.get('skippedFiles', 0)
for site in stat['AgentJobInfo'][agent]['tasks'].get(task, {}).get(
'sites', {}):
info = stat['AgentJobInfo'][agent]['tasks'][task]['sites'][
site]
for s in ['success', 'failure', 'cooloff', 'submitted']:
if not s in info: continue
data = info[s]
if type(data) == dict:
total_per_site[site] += sum(data.values())
else:
total_per_site[site] += data
#is the task relevant to recover (discard log, cleanup)
if any([v in task.lower() for v in ['logcol', 'cleanup']]): continue
#total_count= defaultdict(int)
#error_site_count = defaultdict( lambda : defaultdict(int))
if not task in err:
print task, "has not reported error"
err[task] = {}
#print err[task].keys()
for exittype in err[task]:
#print "\t",err[task][exittype].keys()
for errorcode_s in err[task][exittype]:
if errorcode_s == '0': continue
#print "\t\t",err[task][exittype][errorcode_s].keys()
for site in err[task][exittype][errorcode_s]:
ce = SI.SE_to_CE(site)
count = err[task][exittype][errorcode_s][site][
'errorCount']
total_count[errorcode_s] += count
#error_site_count[errorcode_s][site] += count
error_site_count[errorcode_s][ce] += count
## show the total
all_sites = set()
all_codes = set()
for code in error_site_count:
for site in error_site_count[code]:
all_sites.add(site)
if code != '0':
all_codes.add(code)
s_per_code = defaultdict(int)
for site in all_sites:
for code in sorted(all_codes):
s_per_code[code] += error_site_count[code][site]
expose_top_N = UC.get('expose_top_N')
count_top_N = min(
sorted(s_per_code.values(),
reverse=True)[:expose_top_N]) if s_per_code else -1
for exittype in err[task]:
#print "\t",err[task][exittype].keys()
for errorcode_s in err[task][exittype]:
if errorcode_s == '0': continue
#print "\t\t",err[task][exittype][errorcode_s].keys()
force_code = (count_top_N > 0
and s_per_code[errorcode_s] >= count_top_N)
if force_code: print "will expose", errorcode_s, "anyways"
for site in err[task][exittype][errorcode_s]:
ce = SI.SE_to_CE(site)
count = err[task][exittype][errorcode_s][site][
'errorCount']
###total_count[errorcode_s] += count
#error_site_count[errorcode_s][site] += count
###error_site_count[errorcode_s][ce] += count
for sample in err[task][exittype][errorcode_s][site][
'samples']:
#print sample.keys()
for step in sample['errors']:
for report in sample['errors'][step]:
if report['type'] == 'CMSExeption': continue
#if int(report['exitCode']) == int(errorcode_s):
one_explanation[errorcode_s].add(
"%s (Exit code: %s) \n%s" %
(report['type'], report['exitCode'],
report['details']))
per_task_explanation[
"%s:%s" % (task_short, errorcode_s)].add(
"%s (Exit code: %s) \n%s" %
(report['type'], report['exitCode'],
report['details']))
#one_explanation[errorcode_s].add( report['details'] )
#else:
#one_explanation[
agent = sample['agent_name']
wmbs = sample['wmbsid']
workflow = sample['workflow']
if force_code:
if not errorcode_s in expose_condor_code:
expose_condor_code[errorcode_s] = defaultdict(
lambda: n_expose)
if not errorcode_s in expose_archive_code:
expose_archive_code[errorcode_s] = defaultdict(
lambda: n_expose)
if do_CL and ((errorcode_s in expose_condor_code and
expose_condor_code[errorcode_s][agent])
) and 'cern' in agent:
if errorcode_s in expose_condor_code:
expose_condor_code[errorcode_s][agent] -= 1
print errorcode_s, agent, "error count", expose_condor_code.get(
errorcode_s, {}).get(agent, 0)
threads.append(
AgentBuster(agent=agent,
workflow=workflow,
wmbs=wmbs,
errorcode_s=errorcode_s,
base_eos_dir=base_eos_dir,
monitor_eos_dir=monitor_eos_dir,
task_short=task_short))
for out in sample['output']:
#print out
if out['type'] == 'logArchive':
if do_JL and (
(errorcode_s in expose_archive_code and
expose_archive_code[errorcode_s][agent] >
0)):
if errorcode_s in expose_archive_code:
expose_archive_code[errorcode_s][
agent] -= 1
print errorcode_s, agent, "error count", expose_archive_code.get(
errorcode_s, {}).get(agent, 0)
threads.append(
XRDBuster(
out_lfn=out['lfn'],
monitor_eos_dir=monitor_eos_dir,
wfn=wfn,
errorcode_s=errorcode_s,
task_short=task_short,
from_eos=(
not options.not_from_eos
), # if options else True),
))
#print task
#print json.dumps( total_count, indent=2)
#print json.dumps( explanations , indent=2)
all_sites = set()
all_codes = set()
for code in error_site_count:
for site in error_site_count[code]:
all_sites.add(site)
if code != '0':
all_codes.add(code)
## parse the dashboard data
for site in total_by_site_dash:
## no. cannot discriminate by task in dashboard...
#all_sites.add( site )
pass
## parse the acdc data
#notreported='NotReported'
#all_missing_stats = set()
#for site in missing_to_run_at[task] if task in missing_to_run_at else []:
# if not missing_to_run_at[task][site]: continue
# ce = SI.SE_to_CE( site )
# #all_sites.add( ce )
# all_missing_stats.add( ce )
#all_missing_stats = all_missing_stats &set(SI.all_sites)
#all_not_reported = all_missing_stats - all_sites
#print task
#print "site with no report",sorted(all_not_reported)
#print sorted(all_sites)
#print sorted(all_missing_stats)
#all_sites = all_missing_stats | all_sites
#all_sites = all_sites & set(SI.all_sites)
no_error = len(all_not_reported) != 0
if not no_error and notreported in all_codes:
all_codes.remove(notreported)
missing_events = missing_to_run[task] if task in missing_to_run else 0
feff = wfi.getFilterEfficiency(task.split('/')[-1])
html += "<a name=%s>" % task.split('/')[-1]
html += "<b>%s</b>" % task.split('/')[-1]
if missing_events:
if feff != 1.:
html += ' is missing %s events in input and <b>about %s events in output</b>' % (
"{:,}".format(missing_events), "{:,}".format(
int(missing_events * feff)))
else:
html += ' is missing <b>%s events in I/O</b>' % (
"{:,}".format(missing_events))
html += ' <a href="https://cmsweb.cern.ch/couchdb/acdcserver/_design/ACDC/_view/byCollectionName?key=%%22%s%%22&include_docs=true&reduce=false" target=_blank>AC/DC</a>' % (
wfn)
if no_error:
html += "<br><b><font color=red> and has UNreported error</font></b>"
html += "<br><table border=1><thead><tr><th>Sites/Errors</th>"
#for site in all_sites:
# html+='<th>%s</th>'%site
for code in sorted(all_codes):
#html+='<th><a href="#%s">%s</a>'%(code,code)
html += '<th><a href="#%s:%s">%s</a>' % (task_short, code, code)
if (str(code) in expose_archive_code
or do_all_error_code): # and n_expose_base:
html += ' <a href=%s/joblogs/%s/%s/%s>, JobLog</a>' % (
unified_url_eos, wfn, code, task_short)
if (str(code) in expose_condor_code
or do_all_error_code): # and n_expose_base:
html += ' <a href=%s/condorlogs/%s/%s/%s>, CondorLog</a>' % (
unified_url_eos, wfn, code, task_short)
html += '</th>'
html += '<th>Total jobs</th><th>Site Ready</th>'
html += '</tr></thead>\n'
html += '<tr><td>Total</td>'
for code in sorted(all_codes):
html += '<td bgcolor=orange width=100>%d' % (s_per_code[code])
if code in total_by_code_dash:
html += ' (<b><i>%d</i></b>)' % total_by_code_dash[code]
html += '</td>'
ulist = '<ul>'
grand = 0
for status in sorted(status_per_task[task].keys()):
ulist += '<li> %s %d' % (status, status_per_task[task][status])
grand += status_per_task[task][status]
ulist += '<li><b> Total %d </b>' % grand
ulist += '</ul>'
#html += '<td bgcolor=orange> %.2f%% </td>'% (100.*(float(sum(s_per_code.values()))/sum(total_per_site.values())) if sum(total_per_site.values()) else 0.)
html += '<td bgcolor=orange> → %.2f%% ← </td>' % (
100. * (float(sum(s_per_code.values())) / grand) if grand else 0.)
html += '<td bgcolor=orange> %s </td>' % ulist
html += '</tr>'
def palette(frac):
_range = {
0.0: 'green',
0.5: 'green',
0.6: 'darkgreen',
0.7: 'orange',
0.8: 'salmon',
0.9: 'red'
}
which = [k for k in _range.keys() if k <= frac]
if which:
there = max(which)
else:
there = max(_range.keys())
return _range[there]
for site in sorted(all_sites):
site_in = 'Yes'
color = 'bgcolor=lightblue'
if not site in SI.sites_ready:
color = 'bgcolor=indianred'
site_in = '<b>No</b>'
if task in missing_to_run_at and missing_to_run_at[task][
SI.CE_to_SE(site)] == 0 or min_rank == task_rank:
color = 'bgcolor=aquamarine'
site_in = '<b>No</b> but fine'
if not no_error:
site_in += " (%s events)" % ("{:,}".format(
missing_to_run_at[task][SI.CE_to_SE(site)]) if task
in missing_to_run_at else '--')
html += '<tr><td %s>%s</td>' % (color, site)
for code in sorted(all_codes):
if code == notreported:
html += '<td %s width=200>%s events </td>' % (
color, "{:,}".format(
missing_to_run_at[task][SI.CE_to_SE(site)]))
else:
if error_site_count[code][site]:
er_frac = float(
error_site_count[code][site]
) / s_per_code[code] if s_per_code[code] else 0.
si_frac = float(
error_site_count[code][site]) / total_per_site[
site] if total_per_site[site] else 0.
html += '<td %s width=200>%d' % (
color, error_site_count[code][site])
if code in r_dashb and site in r_dashb[code]:
html += ' (<b><i>%d</i></b>)' % (
r_dashb[code][site])
html += ', <font color=%s>↑ %.1f%%</font>, <font color=%s>→ %.1f%%</font></td>' % (
palette(er_frac), 100. * er_frac, palette(si_frac),
100. * si_frac)
else:
html += '<td %s>0</td>' % color
html += '<td bgcolor=orange>%d</td>' % total_per_site[site]
html += '<td %s>%s</td>' % (color, site_in)
html += '</tr>\n'
html += '</table><br>'
task_error_site_count[task] = error_site_count
## run all retrieval
run_threads = ThreadHandler(
threads=threads,
n_threads=options.log_threads, # if options else 5,
sleepy=10,
timeout=UC.get('retrieve_errors_timeout'),
verbose=True)
run_threads.start()
html += '<hr><br>'
html += '<a name=BLOCK></a>'
html += "<b>Blocks (%d/%d) needed for recovery</b><br>" % (
len(needed_blocks_loc), len(all_blocks))
for block in sorted(needed_blocks_loc.keys()):
html += '%s <b>@ %s</b><br>' % (block, ','.join(
sorted(needed_blocks_loc[block])))
html += '<a name=FILE></a>'
html += "<br><b>Files in no block</b><br>"
rthreads = []
check_files = [f for f in files_and_loc_notin_dbs.keys() if '/store' in f]
random.shuffle(check_files)
check_files = check_files[:100]
check_files = [] ## disable it completely
by_f = {}
if check_files:
for f in check_files:
rthreads.append(ReadBuster(file=f))
print "checking on existence of", len(rthreads), "files"
run_rthreads = ThreadHandler(threads=rthreads,
n_threads=20,
timeout=10)
run_rthreads.start()
while run_rthreads.is_alive():
time.sleep(10)
for t in run_rthreads.threads:
by_f[t.file] = t.readable
#print "checked",t.file,t.readable
for f in sorted(files_and_loc_notin_dbs.keys()):
readable = by_f.get(f, -1)
if readable == -1:
fs = '%s' % f
elif readable == 0:
fs = '<font color=light green>%s</font>' % f
#print f,"is readable"
else:
fs = '<font color=red>%s</font>' % f
#print f,"is not readable"
html += '%s <b>@</b> %s<br>' % (fs, ','.join(
sorted(files_and_loc_notin_dbs[f])))
#html +='%s <b>@</b> %s<br>'%(f , ','.join(sorted(files_and_loc_notin_dbs[f])) )
html += '<hr><br>'
html += '<a name=CODES></a>'
html += '<table border=1>'
for code in per_task_explanation:
html += '<tr><td><a name="%s">%s</a><br><a href=https://twiki.cern.ch/twiki/bin/view/CMSPublic/JobExitCodes>code twiki</a></td><td>%s</td></tr>' % (
code, code, '<br><br>'.join(per_task_explanation[code]).replace(
'\n', '<br>'))
#for code in one_explanation:
# html +='<tr><td><a name="%s">%s</a></td><td>%s</td></tr>'% ( code, code, '<br><br>'.join(one_explanation[code]).replace('\n','<br>' ))
html += '</table>'
html += ('<br>' * 30)
html += '</html>'
time_point("Report finished")
wfi.sendLog('error', html, show=False)
fn = '%s' % wfn
time_point("error send to ES")
open('%s/report/%s' % (monitor_dir, fn), 'w').write(html)
open('%s/report/%s' % (monitor_eos_dir, fn), 'w').write(html)
time_point("Finished with showError")
## then wait for the retrivals to complete
ping = 0
while run_threads.is_alive():
ping += 1
if ping % 100:
time_point("waiting for sub-threads to finish")
time.sleep(6)
time_point("Finished with retrieval threads")
return task_error_site_count, one_explanation
def htmlor(caller=""):
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
try:
boost = json.loads(open('%s/equalizor.json' %
monitor_dir).read())['modifications']
except:
boost = {}
cache = getWorkflows(reqmgr_url, 'assignment-approved', details=True)
cache.extend(getWorkflows(reqmgr_url, 'acquired', details=True))
cache.extend(getWorkflows(reqmgr_url, 'running-open', details=True))
cache.extend(getWorkflows(reqmgr_url, 'running-closed', details=True))
def getWL(wfn):
cached = filter(lambda d: d['RequestName'] == wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad(reqmgr_url, wfn)
return wl
def wfl(wf,
view=False,
p=False,
ms=False,
within=False,
ongoing=False,
status=False,
update=False):
wfn = wf.name
wfs = wf.wm_status
wl = None
pid = None
wl_pid = None
pids = filter(lambda seg: seg.count('-') == 2, wf.name.split('_'))
if len(pids):
pids = pids[:1]
pid = pids[0]
if not pids:
wl = getWL(wf.name)
pids = getPrepIDs(wl)
pid = pids[0]
wl_pid = pid
if 'task' in wf.name:
wl_pid = 'task_' + pid
text = ', '.join([
#wfn,
#'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">%s</a> '%(wfn,wfn),
#'<table><tr><td>%s</td></tr></table>'%(wfn),
#'<span>%s</span>'%(wfn),
"%s " % wfn,
'(%s) <br>' % wfs
])
text += ', '.join([
'<a href="https://%s/reqmgr2/fetch?rid=%s" target="_blank">dts</a>'
% (reqmgr_url, wfn),
'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">dts-req1</a>'
% wfn,
#TOFIX '<a href=https://cmsweb.cern.ch/reqmgr/view/showWorkload?requestName=%s target="_blank">wkl</a>'%wfn,
'<a href="https://%s/couchdb/reqmgr_workload_cache/%s" target="_blank">wfc</a>'
% (reqmgr_url, wfn),
'<a href="https://%s/reqmgr2/data/request?name=%s" target="_blank">req</a>'
% (reqmgr_url, wfn),
#'<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/request?requestName=%s" target="_blank">dwkc</a>'%wfn,
#TOFIX '<a href="https://cmsweb.cern.ch/reqmgr/view/splitting/%s" target="_blank">spl</a>'%wfn,
'<a href="https://cms-pdmv.cern.ch/stats/?RN=%s" target="_blank">vw</a>'
% wfn,
'<a href="https://cms-pdmv.cern.ch/stats/restapi/get_one/%s" target="_blank">vwo</a>'
% wfn,
'<a href="https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=full&reverse=0&reverse=1&npp=20&subtext=%s&sall=q" target="_blank">elog</a>'
% pid,
'<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank">pv</a>'
% wfn,
#deprecated '<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/outputDatasetsByRequestName/%s" target="_blank">out</a>'%wfn,
'<a href="closeout.html#%s" target="_blank">clo</a>' % wfn,
'<a href="statuses.html#%s" target="_blank">st</a>' % wfn,
'<a href="https://%s/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s" target="_blank">perf</a>'
% (reqmgr_url, wfn)
])
if within and (not view or wfs == 'completed'):
wl = getWL(wfn)
dataset = None
if 'InputDataset' in wl:
dataset = wl['InputDataset']
if 'Task1' in wl and 'InputDataset' in wl['Task1']:
dataset = wl['Task1']['InputDataset']
if dataset:
text += ', '.join([
'',
'<a href=https://cmsweb.cern.ch/das/request?input=%s target=_blank>input</a>'
% dataset,
'<a href=https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions#state=create_since=0;filter=%s target=_blank>sub</a>'
% dataset,
'<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/subscriptions?dataset=%s&collapse=n target=_blank>ds</a>'
% dataset,
'<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/blockreplicas?dataset=%s target=_blank>rep</a>'
% dataset,
])
if p:
cached = filter(lambda d: d['RequestName'] == wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad('cmsweb.cern.ch', wfn)
text += ', (%s)' % (wl['RequestPriority'])
pass
if pid:
if ms:
mcm_s = json.loads(
os.popen(
'curl https://cms-pdmv.cern.ch/mcm/public/restapi/requests/get_status/%s --insecure'
% pid).read())[pid]
text += ', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm (%s)</a>' % (
pid, mcm_s)
else:
text += ', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm</a>' % (
pid)
text += ', <a href="https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?prep_id=%s" target="_blank">ac</a>' % (
wl_pid)
if status:
if wf.status.startswith('assistance'):
text += ', <a href="assistance.html#%s" target="_blank">assist</a>' % wfn
text += ' : %s ' % (wf.status)
if view and wfs != 'acquired':
text += '<a href="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" target="_blank"><img src="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" style="height:50px"></a>' % (
wfn.replace('_', '/'), wfn.replace('_', '/'))
if ongoing:
text += '<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a>' % (
wfn, wfn)
if ongoing:
date1 = time.strftime(
'%Y-%m-%d+%H:%M',
time.gmtime(time.mktime(time.gmtime()) - (15 * 24 * 60 * 60)))
date2 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime())
text += '<a href="http://dashb-cms-job.cern.ch/dashboard/templates/web-job2/#table=Jobs&date1=%s&date2=%s&sortby=site&task=wmagent_%s">dashb</a>' % (
date1, date2, wfn)
if ongoing and wfn in boost:
for task in boost[wfn]:
overflow = boost[wfn][task].get('ReplaceSiteWhitelist', None)
if not overflow:
overflow = boost[wfn][task].get('AddWhitelist', None)
if overflow:
text += ',boost (<a href=equalizor.json>%d</a>)' % len(
overflow)
#text+="<hr>"
return text
def phl(phid):
text = ', '.join([
str(phid),
'<a href="https://cmsweb.cern.ch/phedex/prod/Request::View?request=%s" target="_blank">vw</a>'
% phid,
'<a href="https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions?reqfilter=%s" target="_blank">sub</a>'
% phid,
])
return text
def ol(out):
return '<a href="https://cmsweb.cern.ch/das/request?input=%s" target="_blank"> %s</a>' % (
out, out)
def lap(comment):
l = time.mktime(time.gmtime())
spend = l - lap.start
lap.start = l
print "Spend %d [s] for %s" % (spend, comment)
lap.start = time.mktime(time.gmtime())
## start to write it
#html_doc = open('/afs/cern.ch/user/v/vlimant/public/ops/index.html','w')
html_doc = open('%s/index.html.new' % monitor_dir, 'w')
print "Updating the status page ..."
UC = unifiedConfiguration()
if not caller:
try:
#caller = sys._getframe(1).f_code.co_name
caller = sys.argv[0].split('/')[-1].replace('.py', '')
print "caller is"
print caller
except Exception as es:
caller = 'none found'
print "not getting frame"
print str(es)
html_doc.write("""
<html>
<head>
<META HTTP-EQUIV="refresh" CONTENT="900">
<script type="text/javascript">
function showhide(id) {
var e = document.getElementById(id);
e.style.display = (e.style.display == 'block') ? 'none' : 'block';
}
</script>
</head>
<body>
Last update on %s(CET), %s(GMT)
<br>
<a href=logs/ target=_blank>logs</a> <a href=logs/last.log target=_blank>last</a> <a href=statuses.html>statuses</a> <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/ target=_blank>prod mon</a> <a href=https://%s/wmstats/index.html target=_blank>wmstats</a> <a href=http://t3serv001.mit.edu/~cmsprod/IntelROCCS/Detox/SitesInfo.txt target=_blank>detox</a> <a href=locked.html>space</a> <a href=logs/subscribor/last.log target=_blank>blocks</a> <a href=logs/agents/last.log>agents</a>
<br>
<a href=https://twiki.cern.ch/twiki/bin/view/CMSPublic/CompOpsWorkflowL3Responsibilities#Automatic_Assignment_and_Unified>what am I</a> <a href=data.html>json interfaces</a> <a href=logs/addHoc/last.log>add-hoc op</a> created from <b>%s <a href=logs/last_running>last running</a></b> <object height=20 type="text/html" data="logs/last_running"><p>backup content</p></object>
<br><br>
""" % (time.asctime(time.localtime()), time.asctime(
time.gmtime()), reqmgr_url, caller))
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(
Workflow.status.startswith('considered')).all():
wl = getWL(wf.name)
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])] += 1
#print wf.name
text += "<li> %s </li> \n" % wfl(wf, p=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(
count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write("""
Worflow next to handle (%d) <a href=https://cms-pdmv.cern.ch/mcm/batches?status=new&page=-1 target="_blank"> batches</a> <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/transferor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('considered')">[Click to show/hide]</a>
<br>
<div id="considered" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('considered_bywf')">[Click to show/hide]</a><div id="considered_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('considered_bycamp')">[Click to show/hide]</a><div id="considered_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
""" % (count, count, text, len(count_by_campaign), text_by_c))
lap('done with considered')
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(
Workflow.status == 'staging').all():
wl = getWL(wf.name)
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])] += 1
text += "<li> %s </li> \n" % wfl(wf, within=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(
count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write("""
Worflow waiting in staging (%d) <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staging')">[Click to show/hide]</a>
<br>
<div id="staging" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staging_bywf')">[Click to show/hide]</a><div id="staging_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staging_bycamp')">[Click to show/hide]</a><div id="staging_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
""" % (count, count, text, len(count_by_campaign), text_by_c))
lap('done with staging')
text = "<ul>"
count = 0
transfer_per_wf = defaultdict(list)
for ts in session.query(Transfer).filter(Transfer.phedexid > 0).all():
hide = True
t_count = 0
stext = ""
for pid in ts.workflows_id:
w = session.query(Workflow).get(pid)
hide &= (w.status != 'staging')
if w.status in ['considered', 'staging', 'staged']:
stext += "<li> %s </li>\n" % (wfl(w, status=True))
transfer_per_wf[w].append(ts.phedexid)
t_count += 1
stext = '<li> %s serves %d workflows<br><a href="javascript:showhide(\'%s\')">[show/hide]</a> <div id="%s" style="display:none;"><ul>\n' % (
phl(ts.phedexid), t_count, ts.phedexid, ts.phedexid) + stext
stext += "</ul></li>\n"
if hide:
#text+="<li> %s not needed anymore to start running (does not mean it went through completely)</li>"%phl(ts.phedexid)
pass
else:
count += 1
text += stext
text += "</ul>"
text_bywf = "<ul>"
for wf in transfer_per_wf:
text_bywf += "<li> %s </li>" % (wfl(wf, within=True))
text_bywf += '<a href=javascript:showhide("transfer_%s")>[Click to show/hide] %d transfers</a>' % (
wf.name, len(transfer_per_wf[wf]))
text_bywf += '<div id="transfer_%s" style="display:none;">' % wf.name
text_bywf += "<ul>"
for pid in sorted(transfer_per_wf[wf]):
text_bywf += "<li> %s </li>" % (phl(pid))
text_bywf += "</ul></div><hr>"
text_bywf += '</ul>'
html_doc.write("""
Transfer on-going (%d) <a href=http://cmstransferteam.web.cern.ch/cmstransferteam/ target=_blank>dashboard</a> <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('transfer')">[Click to show/hide]</a>
<br>
<div id="transfer" style="display:none;">
<ul>
<li> By Workflow
<a href="javascript:showhide('transfer_bywf')">[Click to show/hide]</a>
<div id="transfer_bywf" style="display:none;">
%s
</div>
</li>
<li> By transfer request
<a href="javascript:showhide('transfer_byreq')">[Click to show/hide]</a>
<div id="transfer_byreq" style="display:none;">
%s
</div>
</li>
</ul>
</div>
""" % (count, text_bywf, text))
lap('done with transfers')
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(
Workflow.status == 'staged').all():
wl = getWL(wf.name)
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])] += 1
text += "<li> %s </li> \n" % wfl(wf, p=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(
count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write(
"""Worflow ready for assigning (%d) <a href=logs/stagor/last.log target=_blank>log</a> <a href=logs/assignor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staged')">[Click to show/hide]</a>
<br>
<div id="staged" style="display:none;">
<br>
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staged_bywf')">[Click to show/hide]</a><div id="staged_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staged_bycamp')">[Click to show/hide]</a><div id="staged_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
""" % (count, count, text, len(count_by_campaign), text_by_c))
lap('done with staged')
lines = []
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status == 'away').all():
wl = getWL(wf.name)
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])] += 1
lines.append("<li> %s <hr></li>" % wfl(wf, view=True, ongoing=True))
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/campaign.php?campaign=%s>mon</a> <a href=https://cms-pdmv.cern.ch/pmp/historical?r=%s target=_blank>pmp</a> " % (
c, sum(count_by_campaign[c].values()), c, c)
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
lines.sort()
html_doc.write("""
Worflow on-going (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests_in_production.php target=_blank>ongoing</a> <a href=https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=summary target=_blank>elog</a> <a href=http://cms-gwmsmon.cern.ch/prodview target=_blank>queues</a> <a href=logs/assignor/last.log target=_blank>log</a> <a href=logs/checkor/last.log target=_blank>postlog</a> <a href=logs/equalizor/last.log target=_blank>equ</a> <a href=logs/completor/last.log target=_blank>comp</a>
<a href="javascript:showhide('away')">[Click to show/hide]</a>
<br>
<div id="away" style="display:none;">
<ul>
<li>By workflow (%d) </li>
<a href="javascript:showhide('away_bywf')">[Click to show/hide]</a><div id="away_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('away_bycamp')">[Click to show/hide]</a><div id="away_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
""" % (len(lines), len(lines), '\n'.join(lines), len(count_by_campaign),
text_by_c))
lap('done with away')
text = ""
count = 0
#for wf in session.query(Workflow).filter(Workflow.status == 'assistance-custodial').all():
for wf in session.query(Workflow).filter(
Workflow.status.startswith('assistance')).filter(
Workflow.status.contains('custodial')).all():
text += "<li> %s </li> \n" % wfl(
wf, view=True, update=True, status=True)
count += 1
text += "</ul></div>\n"
html_doc.write("""Worflow that are closing (%d)
<a href=closeout.html target=_blank>closeout</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('closing')">[Click to show/hide]</a>
<br>
<div id="closing" style="display:none;">
<br>
<ul>
""" % count)
html_doc.write(text)
lap('done with closing')
assistance_by_type = defaultdict(list)
text = ""
count = 0
for wf in session.query(Workflow).filter(
Workflow.status.startswith('assistance-')).all():
assistance_by_type[wf.status].append(wf)
count += 1
for assistance_type in assistance_by_type:
text += "<li> %s (%d) <a href=\"javascript:showhide('%s')\">[Click to show/hide]</a><br><div id=\"%s\" style=\"display:none;\"><ul>" % (
assistance_type,
len(assistance_by_type[assistance_type]),
assistance_type,
assistance_type,
)
for wf in assistance_by_type[assistance_type]:
text += "<li> %s <hr></li> \n" % wfl(
wf, view=True, within=True, status=True, update=True)
text += "</ul></div></li>\n"
html_doc.write("""Worflow which need assistance (%d)
<a href=assistance.html target=_blank>assistance</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/recoveror/last.log target=_blank>postlog</a>
<a href="javascript:showhide('assistance')">[Click to show/hide]</a>
<br>
<div id="assistance" style="display:none;">
<br>
<ul>
%s
</ul>
</div>
""" % (count, text))
lap('done with assistance')
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == 'close').all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write("""Worflow ready to close (%d)
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('close')">[Click to show/hide]</a>
<br>
<div id="close" style="display:none;">
<br>
<ul>
""" % count)
html_doc.write(text)
lap('done with annoucing')
text = ""
count = 0
for wf in session.query(Workflow).filter(
Workflow.status == 'trouble').all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""Worflow with issue (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/injector/last.log target=_blank>postlog</a>
<a href="javascript:showhide('trouble')">[Click to show/hide]</a>
<br>
<div id="trouble" style="display:none;">
<br>
<ul>
""" % count)
html_doc.write(text)
lap('done with trouble')
text = ""
count = 0
for wf in session.query(Workflow).filter(
Workflow.status == 'forget').all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write("""
Worflow to forget (%d) <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/lockor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('forget')">[Click to show/hide]</a>
<br>
<div id="forget" style="display:none;">
<br>
<ul>
""" % count)
html_doc.write(text)
lap('done with forget')
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == 'done').all():
text += "<li> %s </li> \n" % wfl(wf) #,ms=True)
count += 1
text += "</ul></div>\n"
html_doc.write("""
Worflow through (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/lockor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('done')">[Click to show/hide]</a>
<br>
<div id="done" style="display:none;">
<br>
<ul>
""" % count)
html_doc.write(text)
lap('done with done')
wfs = session.query(Workflow).filter(
Workflow.status.endswith('-unlock')).all()
html_doc.write(
" Workflows unlocked : %s <a href=logs/lockor/last.log target=_blank>log</a><br>"
% (len(wfs)))
lap('done with unlocked')
text = ""
lines_thisweek = []
lines_lastweek = []
now = time.mktime(time.gmtime())
this_week = int(time.strftime("%W", time.gmtime()))
start_time_two_weeks_ago = time.mktime(
time.gmtime(now - (20 * 24 * 60 * 60))) # 20
last_week = int(time.strftime("%W", time.gmtime(now - (7 * 24 * 60 * 60))))
all_locks = json.loads(open('%s/globallocks.json' % monitor_dir).read())
waiting_custodial = json.loads(
open('%s/waiting_custodial.json' % monitor_dir).read())
all_pending_approval_custodial = dict([
(k, item) for k, item in waiting_custodial.items() if 'nodes' in item
and not any([node['decided'] for node in item['nodes'].values()])
])
n_pending_approval = len(all_pending_approval_custodial)
#n_pending_approval = len([item for item in waiting_custodial.values() if 'nodes' in item and not any([node['decided'] for node in item['nodes'].values() ])])
missing_approval_custodial = json.loads(
open('%s/missing_approval_custodial.json' % monitor_dir).read())
stuck_custudial = json.loads(
open('%s/stuck_custodial.json' % monitor_dir).read())
lagging_custudial = json.loads(
open('%s/lagging_custodial.json' % monitor_dir).read())
if len(stuck_custudial):
stuck_string = ', <font color=red>%d appear to be <a href=stuck_custodial.json>stuck</a></font>' % len(
stuck_custudial)
else:
stuck_string = ''
if len(missing_approval_custodial):
long_approve_string = ', <font color=red>%d more than %d days</font>' % (
len(missing_approval_custodial), UC.get('transfer_timeout'))
else:
long_approve_string = ''
output_within_two_weeks = session.query(Output).filter(
Output.date >= start_time_two_weeks_ago).all()
waiting_custodial_string = ""
waiting_custodial_strings = []
for ds in waiting_custodial:
out = None
## lots of it will be within two weeks
of = filter(lambda odb: odb.datasetname == ds, output_within_two_weeks)
if of:
out = of[0]
else:
out = session.query(Output).filter(
Output.datasetname == ds).first()
if out:
info = waiting_custodial[out.datasetname]
action = 'going'
if out.datasetname in all_pending_approval_custodial:
action = '<font color=red>pending</font>'
try:
size = str(info['size'])
except:
size = "x"
destination = ",".join(info['nodes'].keys())
if not destination:
destination = '<font color=red>NO SITE</font>'
a_waiting_custodial_string = "<li>on week %s : %s %s</li>" % (
time.strftime("%W (%x %X)", time.gmtime(
out.date)), ol(out.datasetname),
' %s [GB] %s to %s on %s (<a href="https://cmsweb.cern.ch/phedex/datasvc/xml/prod/requestlist?dataset=%s&node=T*MSS">%d missing</a>)'
% (size, action, destination,
time.asctime(time.gmtime(
info['checked'])), out.datasetname, info['nmissing']))
waiting_custodial_strings.append(
(out.date, a_waiting_custodial_string))
waiting_custodial_strings.sort(key=lambda i: i[0])
waiting_custodial_string = "\n".join(
[i[1] for i in waiting_custodial_strings])
#start_time_two_weeks_ago = time.mktime(time.strptime("15-0-%d"%(this_week-2), "%y-%w-%W"))
for out in output_within_two_weeks:
if not out.workflow:
print "This is a problem with", out.datasetname
continue
if out.workflow.status in [
'done-unlock', 'done', 'clean', 'clean-out', 'clean-unlock'
]:
custodial = ''
if out.datasetname in waiting_custodial:
info = waiting_custodial[out.datasetname]
try:
try:
size = str(info['size'])
except:
size = "x"
destination = ",".join(info['nodes'].keys())
if not destination:
destination = '<font color=red>NO SITE</font>'
action = 'going'
if out.datasetname in all_pending_approval_custodial:
action = '<font color=red>pending</font>'
custodial = ' %s [GB] %s to %s on %s (<a href="https://cmsweb.cern.ch/phedex/datasvc/xml/prod/requestlist?dataset=%s&node=T*MSS">%d missing</a>)' % (
size, action, destination,
time.asctime(time.gmtime(info['checked'])),
out.datasetname, info['nmissing'])
except Exception as e:
#print info
#print str(e)
pass
elif out.datasetname in all_locks:
custodial = '<font color=green>LOCKED</font>'
out_week = int(time.strftime("%W", time.gmtime(out.date)))
##only show current week, and the previous.
if last_week == out_week:
lines_lastweek.append(
"<li>on week %s : %s %s</li>" %
(time.strftime("%W (%x %X)", time.gmtime(
out.date)), ol(out.datasetname), custodial))
if this_week == out_week:
lines_thisweek.append(
"<li>on week %s : %s %s</li>" %
(time.strftime("%W (%x %X)", time.gmtime(
out.date)), ol(out.datasetname), custodial))
lines_thisweek.sort()
lines_lastweek.sort()
html_doc.write(
"""Output produced (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?in_disagreement=1 target=_blank>disagreements</a>
<a href="javascript:showhide('output')">[Click to show/hide]</a>
<br>
<div id="output" style="display:none;">
<br>
<ul>
<li> %d waiting to go to tape</li>
<ul>
<li> %d waiting for tape approval%s</li>
<li> %d are not completed after %d days%s</li>
<li> Full list (%d) <a href="javascript:showhide('waiting-custodial')">[Click to show/hide]</a>
<div id="waiting-custodial" style="display:none;">
<ul>
%s
</ul>
</div>
</li>
</ul>
<li> Last week (%d) </li><a href="javascript:showhide('output_lastweek')">[Click to show/hide]</a><div id="output_lastweek" style="display:none;"><ul>
%s
</ul></div>
<li> This week (%d) </li><a href="javascript:showhide('output_thisweek')">[Click to show/hide]</a><div id="output_thisweek" style="display:none;"><ul>
%s
</ul></div></div>
""" % (len(lines_lastweek) + len(lines_thisweek),
len(waiting_custodial), n_pending_approval, long_approve_string,
len(lagging_custudial), UC.get('transfer_timeout'), stuck_string,
len(waiting_custodial), waiting_custodial_string, len(lines_lastweek),
'\n'.join(lines_lastweek), len(lines_thisweek),
'\n'.join(lines_thisweek)))
lap('done with output')
html_doc.write("""Job installed
<a href="javascript:showhide('acron')">[Click to show/hide]</a>
<br>
<div id="acron" style="display:none;">
<br>
<pre>
%s
</pre>
""" % (os.popen('acrontab -l | grep Unified | grep -v \#').read()))
per_module = defaultdict(list)
for t in filter(
None,
os.popen('cat %s/logs/*/*.time' % monitor_dir).read().split('\n')):
module_name, run_time, spend = t.split(':')
## then do what you want with it !
if 'cleanor' in module_name: continue
per_module[module_name].append(int(spend))
def display_time(sec):
m, s = divmod(sec, 60)
h, m = divmod(m, 60)
dis = ""
if h:
dis += "%d [h] " % h
if h or m:
dis += "%d [m] " % m
if h or m or s:
dis += "%d [s]" % s
return dis
html_doc.write("Module running time<ul>\n")
for m, spends in per_module.items():
avg = sum(spends) / float(len(spends))
lasttime = spends[-1]
html_doc.write("<li>%s : last %s, avg %s</li>\n" %
(m, display_time(lasttime), display_time(avg)))
html_doc.write("</ul>")
html_doc.write(
"Last running <pre>%s</pre><br>" %
(os.popen("tac %s/logs/running | head -5" % monitor_dir).read()))
html_doc.write("Order in cycle <pre>%s</pre><br>" % ('\n'.join(
map(
lambda l: l.split('/')[-1].replace('.py', ''),
filter(
lambda l: not l.startswith('#') and 'Unified' in l and 'py' in
l.split('/')[-1],
open('%s/WmAgentScripts/cycle.sh' %
base_dir).read().split('\n'))))))
html_doc.write("</div>\n")
lap('done with jobs')
text = ""
count = 0
for (c, info) in campaignInfo().campaigns.items():
#if 'go' in info and info['go']:
text += "<li>%s <br> <pre>%s</pre> </li>" % (
c, json.dumps(info, indent=2))
count += 1
html_doc.write("""Campaign configuration
<a href="javascript:showhide('campaign')">[Click to show/hide]</a>
<br>
<div id="campaign" style="display:none;">
<br>
<ul>
%s
</ul></div>
""" % (text))
text = ""
count = 0
n_column = 4
SI = siteInfo()
date1 = time.strftime(
'%Y-%m-%d+%H:%M',
time.gmtime(time.mktime(time.gmtime()) -
(15 * 24 * 60 * 60))) ## 15 days
date2 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime())
for t in SI.types():
text += "<li>%s<table border=1>" % t
c = 0
for site in getattr(SI, t):
cpu = SI.cpu_pledges[site] if site in SI.cpu_pledges else 'N/A'
disk = SI.disk[SI.CE_to_SE(site)] if SI.CE_to_SE(
site) in SI.disk else 'N/A'
if c == 0:
text += "<tr>"
if not disk:
ht_disk = '<font color=red>Disk available: %s</font>' % disk
else:
ht_disk = 'Disk available: %s' % disk
text += '<td><a href=http://dashb-ssb.cern.ch/dashboard/templates/sitePendingRunningJobs.html?site=%s>%s</a><br><a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a><br><a href="http://dashb-cms-job.cern.ch/dashboard/templates/web-job2/#user=&refresh=0&table=Jobs&p=1&records=25&activemenu=1&site=%s&submissiontool=wmagent&check=submitted&sortby=activity&scale=linear&bars=20&data1=%s&date2=%s">dashb</a><br>CPU pledge: %s<br>%s</td>' % (
site, site, site, site, site, date1, date2, cpu, ht_disk)
if c == n_column:
c = 0
else:
c += 1
text += "</table></li>"
text += "<li> Sites in auto-approved transfer<ul>"
for site in sorted(SI.sites_auto_approve):
text += "<li>%s" % site
text += "</ul></li>"
text += "<li> Sites with vetoe transfer<ul>"
for site in sorted(SI.sites_veto_transfer):
text += "<li>%s" % site
text += "</ul></li>"
text += "<li> Sites banned from production<ul>"
for site in sorted(SI.sites_banned):
text += "<li>%s" % site
text += "</ul></li>"
text += "<li> Approximate Free Tape<ul>"
for mss in SI.storage:
waiting = 0
try:
waiting = float(
os.popen(
"grep '%s is pending . Created since' %s/logs/lockor/last.log -B 3 | grep size | awk '{ sum+=$6 ; print sum }' | tail -1"
% (mss, monitor_dir)).readline())
except Exception as e:
print str(e)
oldest = ""
os.system(
'grep pending %s/logs/lockor/last.log | sort -u > %s/logs/pending.log'
% (monitor_dir, monitor_dir))
try:
oldest = os.popen(
"grep '%s is pending . Created since ' %s/logs/lockor/last.log | sort | awk '{print $10, $11, $12, $13, $14 }' | head -1"
% (mss, monitor_dir)).readline()
except Exception as e:
print str(e)
waiting /= 1024.
text += "<li>%s : %d [TB]. Waiting for approval %d [TB] since %s </li>" % (
mss, SI.storage[mss], waiting, oldest)
text += "</ul></li>"
lap('done with sites')
open('%s/siteInfo.json' % monitor_dir, 'w').write(
json.dumps(dict([(t, getattr(SI, t)) for t in SI.types()]), indent=2))
lap('done with sites json')
chart_data = defaultdict(list)
for site in SI.quota:
chart_data[site].append("""
var data_%s = google.visualization.arrayToDataTable([
['Overall', 'Space in TB'],
//['Quota' , %s],
['Locked' , %s],
['Free' , %s]
]);
""" % (
site,
SI.quota[site],
SI.locked[site],
SI.disk[site],
))
chart_data[site].append("""
var chart_%s = new google.visualization.PieChart(document.getElementById('donutchart_%s'));
chart_%s.draw(data_%s, {title: '%s %s [TB]', pieHole:0.4, slices:{0:{color:'red'},1:{color:'green'}}});
""" % (site, site, site, site, site, SI.quota[site]))
chart_data[site].append("""
<div id="donutchart_%s" style="height: 200px;width: 300px"></div>
""" % (site))
## make the locked/available donut chart
donut_html = open('%s/locked.html' % monitor_dir, 'w')
tables = "\n".join([info[0] for site, info in chart_data.items()])
draws = "\n".join([info[1] for site, info in chart_data.items()])
divs = "\n".join([info[2] for site, info in chart_data.items()])
divs_table = "<table border=0>"
for c, site in enumerate(sorted(chart_data.keys())):
if c % 5 == 0:
divs_table += "<tr>"
divs_table += "<td>%s</td>" % (chart_data[site][2])
divs_table += "</table>"
donut_html.write("""
<html>
<head>
<script type="text/javascript" src="https://www.google.com/jsapi"></script>
<script type="text/javascript">
google.load("visualization", "1", {packages:["corechart"]});
google.setOnLoadCallback(drawChart);
function drawChart() {
%s
%s
}
</script>
</head>
<body>
%s
</body>
</html>
""" % (tables, draws, divs_table))
donut_html.close()
html_doc.write("""Site configuration
<a href="javascript:showhide('site')">[Click to show/hide]</a>
<br>
<div id="site" style="display:none;">
<br>
<ul>
%s
</ul></div>
""" % (text))
lap('done with space')
text = ""
for param in UC.configs:
text += "<li>%s</li><ul>\n" % param
for sub in sorted(UC.configs[param].keys()):
text += "<li> %s : %s </li>\n" % (sub, UC.configs[param][sub])
text += '</ul>\n'
html_doc.write("""Unified configuration
<a href="javascript:showhide('config')">[Click to show/hide]</a>
<br>
<div id="config" style="display:none;">
<br>
<ul>
%s
</ul></div>
""" % (text))
lap('done with configuration')
print "... done with status page."
html_doc.write("""
</body>
</html>
""")
html_doc.close()
## and put the file in place
os.system('mv %s/index.html.new %s/index.html' %
(monitor_dir, monitor_dir))
statuses = json.loads(open('%s/statusmon.json' % monitor_dir).read())
s_count = defaultdict(int)
now = time.mktime(time.gmtime())
for wf in session.query(Workflow).all():
s_count[wf.status] += 1
statuses[now] = dict(s_count)
## remove old entries
for t in statuses.keys():
if (now - float(t)) > 7 * 24 * 60 * 60:
statuses.pop(t)
open('%s/statusmon.json' % monitor_dir,
'w').write(json.dumps(statuses, indent=2))
html_doc = open('%s/statuses.html' % monitor_dir, 'w')
html_doc.write(
""" <html>
<table border=1>
<thead>
<tr>
<th> workflow </th><th> status </th><th> wm status</th>
</tr>
</thead>
""")
wfs = {}
for wfo in session.query(Workflow).all():
## pass all that is unlocked and considered it gone
wfs[wfo.name] = (wfo.status, wfo.wm_status)
open('%s/statuses.json' % monitor_dir, 'w').write(json.dumps(wfs))
for wfn in sorted(wfs.keys()):
## pass all that is unlocked and considered it gone
if 'unlock' in wfs[wfn][0]: continue
html_doc.write(
'<tr><td><a id="%s">%s</a></td><td>%s</td><td>%s</td></tr>\n' %
(wfn, wfn, wfs[wfn][0], wfs[wfn][1]))
html_doc.write("</table>")
html_doc.write("<br>" * 100)
html_doc.write("end of page</html>")
html_doc.close()
def closor(url, specific=None, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
all_late_files = []
check_fullcopy_to_announce = UC.get('check_fullcopy_to_announce')
jump_the_line = options.announce if options else False
if jump_the_line:
wfs = session.query(Workflow).filter(Workflow.status.contains('announce')).filter(sqlalchemy.not_(Workflow.status.contains('announced'))).all()
else:
wfs = session.query(Workflow).filter(Workflow.status=='close').all()
held = set()
print len(wfs),"closing"
max_per_round = UC.get('max_per_round').get('closor',None)
if options.limit: max_per_round = options.limit
random.shuffle( wfs )
if max_per_round: wfs = wfs[:max_per_round]
batch_go = {}
batch_warnings = defaultdict(set)
batch_goodness = UC.get("batch_goodness")
for wfo in wfs:
if specific and not specific in wfo.name: continue
## what is the expected #lumis
wfi = workflowInfo(url, wfo.name )
wfo.wm_status = wfi.request['RequestStatus']
if wfi.isRelval():
has_batch_go = False
batch_name = wfi.getCampaign()
if not batch_name in batch_go:
## do the esimatation whethere this can be announced : only once per batch
in_batches = getWorkflowByCampaign(url , batch_name, details=True)
batch_go[ batch_name ] = all(map(lambda s : not s in ['completed','running-open','running-closed','acquired','assigned','assignment-approved'], [r['RequestStatus'] for r in in_batches]))
## already verified
has_batch_go = batch_go[batch_name]
if not has_batch_go:
wfi.sendLog('closor', 'Cannot close for now because the batch %s is not all close'% batch_name)
continue
if wfi.request['RequestStatus'] in ['announced','normal-archived'] and not options.force:
## manually announced ??
wfo.status = 'done'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog('closor','%s is announced already : %s'%( wfo.name,wfo.wm_status))
session.commit()
if jump_the_line:
wfi.sendLog('closor','Announcing while completing')
expected_lumis = 1
if not 'TotalInputLumis' in wfi.request:
print wfo.name,"has not been assigned yet, or the database is corrupted"
elif wfi.request['TotalInputLumis']==0:
print wfo.name,"is corrupted with 0 expected lumis"
else:
expected_lumis = wfi.request['TotalInputLumis']
## what are the outputs
outputs = wfi.request['OutputDatasets']
## check whether the number of lumis is as expected for each
all_OK = defaultdict(lambda : False)
stats = defaultdict(int)
#print outputs
if len(outputs):
print wfo.name,wfi.request['RequestStatus']
for out in outputs:
event_count,lumi_count = getDatasetEventsAndLumis(dataset=out)
odb = session.query(Output).filter(Output.datasetname==out).first()
if not odb:
print "adding an output object",out
odb = Output( datasetname = out )
odb.workflow = wfo
session.add( odb )
odb.nlumis = lumi_count
odb.nevents = event_count
odb.workfow_id = wfo.id
if odb.expectedlumis < expected_lumis:
odb.expectedlumis = expected_lumis
else:
expected_lumis = odb.expectedlumis
odb.date = time.mktime(time.gmtime())
session.commit()
fraction = lumi_count/float(expected_lumis)*100.
completion_line = "%60s %d/%d = %3.2f%%"%(out,lumi_count,expected_lumis,fraction)
wfi.sendLog('closor',"\t%s"% completion_line)
if wfi.isRelval() and fraction < batch_goodness:
batch_warnings[ wfi.getCampaign()].add( completion_line )
stats[out] = lumi_count
all_OK[out] = True
## check for at least one full copy prior to moving on
in_full = {}
for out in outputs:
in_full[out] = []
presence = getDatasetPresence( url, out )
where = [site for site,info in presence.items() if info[0]]
if where:
all_OK[out] = True
print out,"is in full at",",".join(where)
in_full[out] = copy.deepcopy(where)
else:
going_to = wfi.request['NonCustodialSites']+wfi.request['CustodialSites']
wfi.sendLog('closor',"%s is not in full anywhere. send to %s"%(out, ",".join(sorted(going_to))))
at_destination = dict([(k,v) for (k,v) in presence.items() if k in going_to])
else_where = dict([(k,v) for (k,v) in presence.items() if not k in going_to])
print json.dumps( at_destination )
print json.dumps( else_where, indent=2 )
## do the full stuck transfer study, missing files and shit !
for there in going_to:
late_info = findLateFiles(url, out, going_to = there )
for l in late_info:
l.update({"workflow":wfo.name,"dataset":out})
all_late_files.extend( late_info )
if check_fullcopy_to_announce:
## only set this false if the check is relevant
all_OK[out] = False
## verify if we have to do harvesting
if not options.no_harvest and not jump_the_line:
(OK, requests) = spawn_harvesting(url, wfi, in_full)
all_OK.update( OK )
## only that status can let me go into announced
if all(all_OK.values()) and ((wfi.request['RequestStatus'] in ['closed-out']) or options.force or jump_the_line):
print wfo.name,"to be announced"
results=[]
if not results:
for out in outputs:
if out in stats and not stats[out]:
continue
_,dsn,process_string,tier = out.split('/')
if all_OK[out]:
results.append(setDatasetStatus(out, 'VALID'))
if all_OK[out] and wfi.isRelval():
## make the specific relval rules and the replicas
## figure the destination(s) out
destinations = set()
if tier != "RECO" and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if tier == "GEN-SIM":
destinations.add('T1_US_FNAL_Disk')
if tier == "GEN-SIM-DIGI-RAW":
destinations.add('T1_US_FNAL_Disk')
if tier == "GEN-SIM-RECO":
destinations.add('T1_US_FNAL_Disk')
if "RelValTTBar" in dsn and "TkAlMinBias" in process_string and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if "MinimumBias" in dsn and "SiStripCalMinBias" in process_string and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if destinations:
wfi.sendLog('closor', '%s to go to %s'%(out, ', '.join( sorted( destinations ))))
## call to makereplicarequest under relval => done
for site in destinations:
result = makeReplicaRequest(url, site, [out], 'Copy for release validation consumption', priority='normal', approve=True, mail=False, group='RelVal')
try:
request_id = result['phedex']['request_created'][0]['id']
results.append( True )
except:
results.append( 'Failed relval transfer' )
elif all_OK[out]:
campaign = None
try:
campaign = out.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request and wfi.request['Campaign']:
campaign = wfi.request['Campaign']
to_DDM = False
## campaign override
if campaign and campaign in CI.campaigns and 'toDDM' in CI.campaigns[campaign] and tier in CI.campaigns[campaign]['toDDM']:
to_DDM = True
## by typical enabling
if tier in UC.get("tiers_to_DDM"):
to_DDM = True
## check for unitarity
if not tier in UC.get("tiers_no_DDM")+UC.get("tiers_to_DDM"):
print "tier",tier,"neither TO or NO DDM for",out
results.append('Not recognitized tier %s'%tier)
#sendEmail("failed DDM injection","could not recognize %s for injecting in DDM"% out)
sendLog('closor', "could not recognize %s for injecting in DDM"% out, level='critical')
continue
n_copies = 2
destinations=[]
if to_DDM and campaign and campaign in CI.campaigns and 'DDMcopies' in CI.campaigns[campaign]:
ddm_instructions = CI.campaigns[campaign]['DDMcopies']
if type(ddm_instructions) == int:
n_copies = CI.campaigns[campaign]['DDMcopies']
elif type(ddm_instructions) == dict:
## a more fancy configuration
for ddmtier,indication in ddm_instructions.items():
if ddmtier==tier or ddmtier in ['*','all']:
## this is for us
if 'N' in indication:
n_copies = indication['N']
if 'host' in indication:
destinations = indication['host']
destination_spec = ""
if destinations:
destination_spec = "--destination="+",".join( destinations )
group_spec = "" ## not used yet
### should make this a campaign configuration
## inject to DDM when necessary
if to_DDM:
print "Sending",out," to DDM"
p = os.popen('python assignDatasetToSite.py --nCopies=%d --dataset=%s %s %s --debug 0 --exec'%(n_copies, out,destination_spec, group_spec))
ddm_text = p.read()
print ddm_text
status = p.close()
if status!=None:
print "Failed DDM, retrying to send",out,"a second time"
p = os.popen('python assignDatasetToSite.py --nCopies=%d --dataset=%s %s %s --debug 1 --exec'%(n_copies, out,destination_spec, group_spec))
ddm_text = p.read()
print ddm_text
status = p.close()
if status!=None:
#sendEmail("failed DDM injection","could not add "+out+" to DDM pool. check closor logs.")
sendLog('closor',"could not add "+out+" to DDM pool. check closor logs.", level='critical')
if options.force: status = True
results.append( status )
if status == None:
wfi.sendLog('closor',ddm_text)
wfi.sendLog('closor','%s is send to AnalysisOps DDM pool in %s copies %s'%( out, n_copies, destination_spec))
else:
print wfo.name,"no stats for announcing",out
results.append('No Stats')
if all(map(lambda result : result in ['None',None,True],results)):
if not jump_the_line:
## only announce if all previous are fine
res = reqMgrClient.announceWorkflowCascade(url, wfo.name)
if not res in ['None',None]:
## check the status again, it might well have toggled
wl_bis = workflowInfo(url, wfo.name)
wfo.wm_status = wl_bis.request['RequestStatus']
session.commit()
if wl_bis.request['RequestStatus'] in ['announced','normal-archived']:
res = None
else:
## retry ?
res = reqMgrClient.announceWorkflowCascade(url, wfo.name)
results.append( res )
#print results
if all(map(lambda result : result in ['None',None,True],results)):
if jump_the_line:
if not 'announced' in wfo.status:
wfo.status = wfo.status.replace('announce','announced')
else:
wfo.status = 'done'
session.commit()
wfi.sendLog('closor',"workflow outputs are announced")
else:
wfi.sendLog('closor',"Error with %s to be announced \n%s"%( wfo.name, json.dumps( results )))
elif wfi.request['RequestStatus'] in ['failed','aborted','aborted-archived','rejected','rejected-archived','aborted-completed']:
if wfi.isRelval():
wfo.status = 'forget'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog('closor',"%s is %s, but will not be set in trouble to find a replacement."%( wfo.name, wfo.wm_status))
else:
wfo.status = 'trouble'
wfo.wm_status = wfi.request['RequestStatus']
session.commit()
else:
print wfo.name,"not good for announcing:",wfi.request['RequestStatus']
wfi.sendLog('closor',"cannot be announced")
held.add( wfo.name )
days_late = 0.
retries_late = 10
really_late_files = [info for info in all_late_files if info['retries']>=retries_late]
really_late_files = [info for info in really_late_files if info['delay']/(60*60*24.)>=days_late]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s'%(len(really_late_files), days_late, retries_late, json.dumps( really_late_files , indent=2) )
#sendEmail('waiting for files to announce', subject)
sendLog('closor', subject, level='warning')
sendLog('closor',subject)
print subject
open('%s/stuck_files.json'%monitor_dir,'w').write( json.dumps( really_late_files , indent=2))
if held:
sendLog('closor',"the workflows below are held up \n%s"%("\n".join( sorted(held) )), level='critical')
#batches = json.loads(open('batches.json').read())
for bname,go in batch_go.items():
if go:
subject = "Release Validation Samples Batch %s"% bname
issues=""
if batch_warnings[ bname ]:
issues="The following datasets have outstanding completion (<%d%%) issues:\n\n"% batch_goodness
issues+="\n".join( sorted( batch_warnings[ bname ] ))
issues+="\n\n"
text = """
Dear all,
a batch of release validation workflows has finished.
Batch ID:
%s
Detail of the workflows
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
%s
This is an automated message.
"""%( bname,
bname,
issues)
to = ['*****@*****.**']
sendEmail(subject, text, destination=to )
def closor(url, specific=None, options=None):
if userLock(): return
mlock = moduleLock()
if mlock(): return
up = componentInfo(soft=['mcm','wtc'])
if not up.check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
BI = batchInfo()
CloseI = closeoutInfo()
all_late_files = []
jump_the_line = options.announce if options else False
if jump_the_line:
print "announce option is on. Checking on things on-going ready to be announced"
wfs = session.query(Workflow).filter(Workflow.status.contains('announce')).filter(sqlalchemy.not_(Workflow.status.contains('announced'))).all()
else:
print "regular option. Checking on things done and to be announced"
wfs = session.query(Workflow).filter(Workflow.status=='close').all()
if specific:
wfs = [wfo for wfo in wfs if specific in wfo.name]
wfs_n = [w.name for w in wfs]
print "unique names?"
print len(set(wfs_n)) == len(wfs_n)
held = set()
print len(wfs),"closing"
random.shuffle( wfs )
max_per_round = UC.get('max_per_round').get('closor',None)
if options.limit: max_per_round = options.limit
if max_per_round:
## order them by priority
all_closedout = sorted(getWorkflows(url, 'closed-out', details=True), key = lambda r : r['RequestPriority'])
all_closedout = [r['RequestName'] for r in all_closedout]
def rank( wfn ):
return all_closedout.index( wfn ) if wfn in all_closedout else 0
wfs = sorted( wfs, key = lambda wfo : rank( wfo.name ),reverse=True)
wfs = wfs[:max_per_round]
batch_go = {}
batch_warnings = defaultdict(set)
batch_goodness = UC.get("batch_goodness")
closers = []
print len(wfs),"closing"
th_start = time.mktime(time.gmtime())
for iwfo,wfo in enumerate(wfs):
if specific and not specific in wfo.name: continue
closers.append( CloseBuster(
wfo = wfo,
url = url,
CI = CI,
UC = UC,
jump_the_line = jump_the_line,
batch_goodness = batch_goodness,
batch_go = batch_go,
#stats = stats,
batch_warnings = batch_warnings,
all_late_files = all_late_files,
held = held,
))
run_threads = ThreadHandler( threads = closers,
n_threads = options.threads,
sleepy = 10,
timeout = None,
verbose = True,
label = 'closor')
run_threads.start()
## waiting on all to complete
while run_threads.is_alive():
#print "Waiting on closing threads",time.asctime(time.gmtime())
time.sleep(5)
JC = JIRAClient() if up.status.get('jira',False) else None
print len(run_threads.threads),"finished thread to gather information from"
failed_threads = 0
for to in run_threads.threads:
if to.failed:
failed_threads += 1
continue
if to.outs:
for outO in to.outs:
out = outO.datasetname
odb = session.query(Output).filter(Output.datasetname==out).first()
if not odb:
print "adding an output object",out
session.add( outO )
else:
odb.date = outO.date
if to.to_status:
to.wfo.status = to.to_status
if JC and to.to_status == "done" and to.wfi:
jiras = JC.find({"prepid" : to.wfi.request['PrepID']})
for jira in jiras:
JC.close(jira.key)
if to.to_wm_status:
to.wfo.wm_status = to.to_wm_status
if to.closing:
CloseI.pop( to.wfo.name )
session.commit()
th_stop = time.mktime(time.gmtime())
if wfs:
time_spend_per_workflow = (th_stop-th_start) / float(len(wfs))
print "Average time spend per workflow is", time_spend_per_workflow
if float(failed_threads/run_threads.n_threads) > 0:
sendLog('checkor','%d/%d threads have failed, better check this out'% (failed_threads, run_threads.n_threads), level='critical')
sendEmail('checkor','%d/%d threads have failed, better check this out'% (failed_threads,run_threads.n_threads))
days_late = 0.
retries_late = 10
really_late_files = [info for info in all_late_files if info['retries']>=retries_late]
really_late_files = [info for info in really_late_files if info['delay']/(60*60*24.)>=days_late]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s'%(len(really_late_files), days_late, retries_late, json.dumps( really_late_files , indent=2) )
#sendEmail('waiting for files to announce', subject)
sendLog('closor', subject, level='warning')
sendLog('closor',subject)
print subject
open('%s/stuck_files.json'%monitor_dir,'w').write( json.dumps( really_late_files , indent=2))
if held:
sendLog('closor',"the workflows below are held up \n%s"%("\n".join( sorted(held) )), level='critical')
for bname,go in batch_go.items():
if go:
subject = "Release Validation Samples Batch %s"% bname
issues=""
if batch_warnings[ bname ]:
issues="The following datasets have outstanding completion (<%d%%) issues:\n\n"% batch_goodness
issues+="\n".join( sorted( batch_warnings[ bname ] ))
issues+="\n\n"
text = """
Dear all,
a batch of release validation workflows has finished.
Batch ID:
%s
Detail of the workflows
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
%s
This is an automated message.
"""%( bname,
bname,
issues)
to = ['*****@*****.**']
sendEmail(subject, text, destination=to )
## just announced ; take it out now.
BI.pop( bname )
if os.path.isfile('.closor_stop'):
print "The loop on workflows was shortened"
sendEmail('closor','Closor loop was shortened artificially using .closor_stop')
os.system('rm -f .closor_stop')
def completor(url, specific):
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
if use_mcm:
mcm = McMClient(dev=False)
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
wfs = []
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance')).all() )
## just take it in random order so that not always the same is seen
random.shuffle( wfs )
max_per_round = UC.get('max_per_round').get('completor',None)
if max_per_round and not specific: wfs = wfs[:max_per_round]
all_stuck = set()
## take into account what stagor was saying
all_stuck.update( json.loads( open('%s/stuck_transfers.json'%monitor_dir).read() ))
## take into account the block that needed to be repositioned recently
all_stuck.update( [b.split('#')[0] for b in json.loads( open('%s/missing_blocks.json'%monitor_dir).read()) ] )
## take into account all stuck block and dataset from transfer team
all_stuck.update( getAllStuckDataset())
## by workflow a list of fraction / timestamps
completions = json.loads( open('%s/completions.json'%monitor_dir).read())
good_fractions = {}
truncate_fractions = {}
timeout = {}
for c in CI.campaigns:
if 'force-complete' in CI.campaigns[c]:
good_fractions[c] = CI.campaigns[c]['force-complete']
if 'truncate-complete' in CI.campaigns[c]:
truncate_fractions[c] = CI.campaigns[c]['truncate-complete']
if 'force-timeout' in CI.campaigns[c]:
timeout[c] = CI.campaigns[c]['force-timeout']
long_lasting = {}
overrides = getForceCompletes()
if use_mcm:
## add all workflow that mcm wants to get force completed
mcm_force = mcm.get('/restapi/requests/forcecomplete')
## assuming this will be a list of actual prepids
overrides['mcm'] = mcm_force
print "can force complete on"
print json.dumps( good_fractions ,indent=2)
print "can truncate complete on"
print json.dumps( truncate_fractions ,indent=2)
print "can overide on"
print json.dumps( overrides, indent=2)
max_force = UC.get("max_force_complete")
#wfs_no_location_in_GQ = set()
#block_locations = defaultdict(lambda : defaultdict(list))
#wfs_no_location_in_GQ = defaultdict(list)
set_force_complete = set()
for wfo in wfs:
if specific and not specific in wfo.name: continue
print "looking at",wfo.name
## get all of the same
wfi = workflowInfo(url, wfo.name)
pids = wfi.getPrepIDs()
skip=False
if not any([c in wfo.name for c in good_fractions]) and not any([c in wfo.name for c in truncate_fractions]): skip=True
for user,spec in overrides.items():
if wfi.request['RequestStatus']!='force-complete':
if any(s in wfo.name for s in spec) or (wfo.name in spec) or any(pid in spec for pid in pids) or any(s in pids for s in spec):
wfi = workflowInfo(url, wfo.name)
forceComplete(url , wfi )
skip=True
wfi.notifyRequestor("The workflow %s was force completed by request of %s"%(wfo.name,user), do_batch=False)
wfi.sendLog('completor','%s is asking for %s to be force complete'%(user,wfo.name))
break
if wfo.status.startswith('assistance'): skip = True
if skip:
continue
priority = wfi.request['RequestPriority']
if not 'Campaign' in wfi.request: continue
if not wfi.request['RequestStatus'] in ['acquired','running-open','running-closed']: continue
if wfi.request['RequestStatus'] in ['running-open','running-closed'] and (random.random() < 0.01):
try:
parse_one(url, wfo.name )
except Exception as e:
print "failed error parsing"
print str(e)
c = wfi.request['Campaign']
#if not c in good_fractions: continue
good_fraction = good_fractions.get(c,1000.)
truncate_fraction = truncate_fractions.get(c,1000.)
print "force at",good_fraction,"truncate at",truncate_fraction
ignore_fraction = 2.
lumi_expected = None
event_expected = None
if not 'TotalInputEvents' in wfi.request:
if 'RequestNumEvents' in wfi.request:
event_expected = wfi.request['RequestNumEvents']
else:
print "truncated, cannot do anything"
continue
else:
lumi_expected = wfi.request['TotalInputLumis']
event_expected = wfi.request['TotalInputEvents']
now = time.mktime(time.gmtime()) / (60*60*24.)
running_log = filter(lambda change : change["Status"] in ["running-open","running-closed"],wfi.request['RequestTransition'])
if not running_log:
print "\tHas no running log"
# cannot figure out when the thing started running
continue
then = running_log[-1]['UpdateTime'] / (60.*60.*24.)
delay = now - then ## in days
## this is supposed to be the very initial request date, inherited from clones
try:
injected_on =time.mktime(time.strptime("-".join(map(lambda n : "%02d"%n, wfi.request['RequestDate'])), "%Y-%m-%d-%H-%M-%S")) / (60.*60.*24.)
injection_delay = now - injected_on
except:
injection_delay = None
(w,d) = divmod(delay, 7 )
print "\t"*int(w)+"Running since",delay,"[days] priority=",priority
if injection_delay!=None and injection_delay > 50 and False:
tail_cutting_priority = 200000
## bump up to 200k
print "Injected since",injection_delay,"[days] priority=",priority
if wfi.request['RequestPriority'] != tail_cutting_priority:
wfi.sendLog('completor','bumping priority to %d for being injected since'% tail_cutting_priority)
reqMgrClient.changePriorityWorkflow(url, wfo.name, tail_cutting_priority)
_,prim,_,_ = wfi.getIO()
is_stuck = all_stuck & prim
if is_stuck: wfi.sendLog('completor','%s is stuck'%','.join(is_stuck))
monitor_delay = 7
allowed_delay = 14
if c in timeout:
allowed_delay = timeout[c]
monitor_delay = min(monitor_delay, allowed_delay)
### just skip if too early
if delay <= monitor_delay: continue
long_lasting[wfo.name] = { "delay" : delay,
"injection_delay" : injection_delay }
percent_completions = {}
for output in wfi.request['OutputDatasets']:
if "/DQM" in output: continue ## that does not count
if not output in completions: completions[output] = { 'injected' : None, 'checkpoints' : [], 'workflow' : wfo.name}
## get completion fraction
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
lumi_completion=0.
event_completion=0.
if lumi_expected:
lumi_completion = lumi_count / float( lumi_expected )
if event_expected:
event_completion = event_count / float( event_expected )
#take the less optimistic
## taking the max or not : to force-complete on over-pro
percent_completions[output] = min( lumi_completion, event_completion )
completions[output]['checkpoints'].append( (now, event_completion ) )
if all([percent_completions[out] >= good_fraction for out in percent_completions]):
wfi.sendLog('completor', "all is above %s \n%s"%( good_fraction,
json.dumps( percent_completions, indent=2 )
))
elif is_stuck and (all([percent_completions[out] >= truncate_fraction for out in percent_completions])):
wfi.sendLog('completor', "all is above %s truncation level, and the input is stuck\n%s"%( truncate_fraction,
json.dumps( percent_completions, indent=2 ) ) )
sendEmail('workflow for truncation','check on %s'%wfo.name)
##continue ## because you do not want to get this one right now
else:
long_lasting[wfo.name].update({
'completion': sum(percent_completions.values()) / len(percent_completions),
'completions' : percent_completions
})
## do something about the agents this workflow is in
long_lasting[wfo.name]['agents'] = wfi.getAgents()
wfi.sendLog('completor', "%s not over bound %.3f (complete) %.3f truncate \n%s"%(percent_completions.values(),
good_fraction, truncate_fraction,
json.dumps( long_lasting[wfo.name]['agents'], indent=2) ))
continue
if all([percent_completions[out] >= ignore_fraction for out in percent_completions]):
print "all is done, just wait a bit"
continue
for output in percent_completions:
completions[output]['injected'] = then
#further check on delays
cpuh = wfi.getComputingTime(unit='d')
ran_at = wfi.request['SiteWhitelist']
wfi.sendLog('completor',"Required %s, time spend %s"%( cpuh, delay))
##### WILL FORCE COMPLETE BELOW
# only really force complete after n days
if delay <= allowed_delay: continue
## find ACDCs that might be running
if max_force>0:
forceComplete(url, wfi )
set_force_complete.add( wfo.name )
print "going for force-complete of",wfo.name
wfi.sendLog('completor','going for force completing')
wfi.notifyRequestor("The workflow %s was force completed for running too long"% wfo.name)
max_force -=1
else:
wfi.sendLog('completor',"too many completion this round, cannot force complete")
## do it once only for testing
#break
if delay >= 40:
print wfo.name,"has been running for",dealy,"days"
## bumping up the priority?
if set_force_complete:
sendLog('completor','The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
open('%s/completions.json'%monitor_dir,'w').write( json.dumps( completions , indent=2))
text="These have been running for long"
open('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 ))
for wf,info in sorted(long_lasting.items(), key=lambda tp:tp[1]['delay'], reverse=True):
delay = info['delay']
text += "\n %s : %s days"% (wf, delay)
if 'completion' in info:
text += " %d%%"%( info['completion']*100 )
print text
def closor(url, specific=None):
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
#LI = lockInfo()
all_late_files = []
check_fullcopy_to_announce = UC.get('check_fullcopy_to_announce')
## manually closed-out workflows should get to close with checkor
if specific:
wfs = session.query(Workflow).filter(Workflow.name.contains(specific)).all()
else:
wfs = session.query(Workflow).filter(Workflow.status=='close').all()
held = set()
for wfo in wfs:
if specific and not specific in wfo.name: continue
## what is the expected #lumis
wfi = workflowInfo(url, wfo.name )
wfo.wm_status = wfi.request['RequestStatus']
if wfi.request['RequestStatus'] in ['announced','normal-archived']:
## manually announced ??
wfo.status = 'done'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog('closor','%s is announced already : %s'%( wfo.name,wfo.wm_status))
session.commit()
expected_lumis = 1
if not 'TotalInputLumis' in wfi.request:
print wfo.name,"has not been assigned yet, or the database is corrupted"
else:
expected_lumis = wfi.request['TotalInputLumis']
## what are the outputs
outputs = wfi.request['OutputDatasets']
## check whether the number of lumis is as expected for each
all_OK = defaultdict(lambda : False)
#print outputs
if len(outputs):
print wfo.name,wfi.request['RequestStatus']
for out in outputs:
event_count,lumi_count = getDatasetEventsAndLumis(dataset=out)
odb = session.query(Output).filter(Output.datasetname==out).first()
if not odb:
print "adding an output object",out
odb = Output( datasetname = out )
odb.workflow = wfo
session.add( odb )
odb.nlumis = lumi_count
odb.nevents = event_count
odb.workfow_id = wfo.id
if odb.expectedlumis < expected_lumis:
odb.expectedlumis = expected_lumis
else:
expected_lumis = odb.expectedlumis
odb.date = time.mktime(time.gmtime())
session.commit()
wfi.sendLog('closor',"\t%60s %d/%d = %3.2f%%"%(out,lumi_count,expected_lumis,lumi_count/float(expected_lumis)*100.))
#print wfo.fraction_for_closing, lumi_count, expected_lumis
#fraction = wfo.fraction_for_closing
#fraction = 0.0
#all_OK.append((float(lumi_count) > float(expected_lumis*fraction)))
all_OK[out] = True
## check for at least one full copy prior to moving on
in_full = {}
for out in outputs:
in_full[out] = []
presence = getDatasetPresence( url, out )
where = [site for site,info in presence.items() if info[0]]
if where:
all_OK[out] = True
print out,"is in full at",",".join(where)
in_full[out] = copy.deepcopy(where)
else:
going_to = wfi.request['NonCustodialSites']+wfi.request['CustodialSites']
wfi.sendLog('closor',"%s is not in full anywhere. send to %s"%(out, ",".join(sorted(going_to))))
at_destination = dict([(k,v) for (k,v) in presence.items() if k in going_to])
else_where = dict([(k,v) for (k,v) in presence.items() if not k in going_to])
print json.dumps( at_destination )
print json.dumps( else_where, indent=2 )
## do the full stuck transfer study, missing files and shit !
for there in going_to:
late_info = findLateFiles(url, out, going_to = there )
for l in late_info:
l.update({"workflow":wfo.name,"dataset":out})
all_late_files.extend( late_info )
if check_fullcopy_to_announce:
## only set this false if the check is relevant
all_OK[out] = False
## verify if we have to do harvesting
(OK, requests) = spawn_harvesting(url, wfi, in_full)
all_OK.update( OK )
## only that status can let me go into announced
if all(all_OK.values()) and wfi.request['RequestStatus'] in ['closed-out']:
print wfo.name,"to be announced"
results=[]#'dummy']
if not results:
for out in outputs:
if all_OK[out]:
results.append(setDatasetStatus(out, 'VALID'))
tier = out.split('/')[-1]
campaign = None
try:
campaign = out.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request and wfi.request['Campaign']:
campaign = wfi.request['Campaign']
to_DDM = False
## campaign override
if campaign and campaign in CI.campaigns and 'toDDM' in CI.campaigns[campaign] and tier in CI.campaigns[campaign]['toDDM']:
to_DDM = True
## by typical enabling
if tier in UC.get("tiers_to_DDM"):
to_DDM = True
## check for unitarity
if not tier in UC.get("tiers_no_DDM")+UC.get("tiers_to_DDM"):
print "tier",tier,"neither TO or NO DDM for",out
results.append('Not recognitized tier %s'%tier)
sendEmail("failed DDM injection","could not recognize %s for injecting in DDM"% out)
continue
n_copies = 2
destinations=[]
if to_DDM and campaign and campaign in CI.campaigns and 'DDMcopies' in CI.campaigns[campaign]:
ddm_instructions = CI.campaigns[campaign]['DDMcopies']
if type(ddm_instructions) == int:
n_copies = CI.campaigns[campaign]['DDMcopies']
elif type(ddm_instructions) == dict:
## a more fancy configuration
for ddmtier,indication in ddm_instructions.items():
if ddmtier==tier or ddmtier in ['*','all']:
## this is for us
if 'N' in indication:
n_copies = indication['N']
if 'host' in indication:
destinations = indication['host']
destination_spec = ""
if destinations:
destination_spec = "--destination="+",".join( destinations )
## inject to DDM when necessary
if to_DDM:
#print "Sending",out," to DDM"
p = os.popen('python assignDatasetToSite.py --nCopies=%d --dataset=%s %s --exec'%(n_copies, out,destination_spec))
print p.read()
status = p.close()
if status!=None:
print "Failed DDM, retrying a second time"
p = os.popen('python assignDatasetToSite.py --nCopies=%d --dataset=%s %s --exec'%(n_copies, out,destination_spec))
print p.read()
status = p.close()
if status!=None:
sendEmail("failed DDM injection","could not add "+out+" to DDM pool. check closor logs.")
results.append( status )
if status == None:
wfi.sendLog('closor','%s is send to AnalysisOps DDM pool in %s copies %s'%( n_copies, out,destination_spec))
else:
print wfo.name,"no stats for announcing",out
results.append('No Stats')
if all(map(lambda result : result in ['None',None,True],results)):
## only announce if all previous are fine
res = reqMgrClient.announceWorkflowCascade(url, wfo.name)
if not res in ['None',None]:
## check the status again, it might well have toggled
wl_bis = workflowInfo(url, wfo.name)
wfo.wm_status = wl_bis.request['RequestStatus']
session.commit()
if wl_bis.request['RequestStatus'] in ['announced','normal-archived']:
res = None
else:
## retry ?
res = reqMgrClient.announceWorkflowCascade(url, wfo.name)
results.append( res )
#print results
if all(map(lambda result : result in ['None',None,True],results)):
wfo.status = 'done'
session.commit()
wfi.sendLog('closor',"workflow is announced")
else:
print "ERROR with ",wfo.name,"to be announced",json.dumps( results )
else:
print wfo.name,"not good for announcing:",wfi.request['RequestStatus']
wfi.sendLog('closor',"cannot be announced")
held.add( wfo.name )
days_late = 0.
retries_late = 10
really_late_files = [info for info in all_late_files if info['retries']>=retries_late]
really_late_files = [info for info in really_late_files if info['delay']/(60*60*24.)>=days_late]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s'%(len(really_late_files), days_late, retries_late, json.dumps( really_late_files , indent=2) )
sendEmail('waiting for files to announce', subject)
sendLog('closor',subject)
print subject
open('%s/stuck_files.json'%monitor_dir,'w').write( json.dumps( really_late_files , indent=2))
if held:
sendEmail("held from announcing","the workflows below are held up, please check the logs https://cmst2.web.cern.ch/cmst2/unified/logs/closor/last.log \n%s"%("\n".join( held )))
def assignor(url, specific=None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
#SI = siteInfo()
SI = global_SI()
#NLI = newLockInfo()
#if not NLI.free() and not options.go: return
LI = lockInfo()
if not LI.free() and not options.go: return
n_assigned = 0
n_stalled = 0
wfos = []
fetch_from = []
if specific or options.early:
fetch_from.extend(['considered', 'staging'])
if specific:
fetch_from.extend(['considered-tried'])
if options.early:
print "Option Early is on"
fetch_from.extend(['staged'])
if options.from_status:
fetch_from = options.from_status.split(',')
print "Overriding to read from", fetch_from
for status in fetch_from:
print "getting wf in", status
wfos.extend(
session.query(Workflow).filter(Workflow.status == status).all())
print len(wfos)
## in case of partial, go for fetching a list from json ?
#if options.partial and not specific:
# pass
dataset_endpoints = json.loads(
open('%s/dataset_endpoints.json' % monitor_dir).read())
aaa_mapping = json.loads(
open('%s/equalizor.json' % monitor_pub_dir).read())['mapping']
all_stuck = set()
all_stuck.update(
json.loads(open('%s/stuck_transfers.json' % monitor_pub_dir).read()))
all_stuck.update(getAllStuckDataset())
max_per_round = UC.get('max_per_round').get('assignor', None)
max_cpuh_block = UC.get('max_cpuh_block')
##order by priority instead of random
if options.early:
cache = sorted(getWorkflows(url, 'assignment-approved', details=True),
key=lambda r: r['RequestPriority'])
cache = [r['RequestName'] for r in cache]
def rank(wfn):
return cache.index(wfn) if wfn in cache else 0
wfos = sorted(wfos, key=lambda wfo: rank(wfo.name), reverse=True)
print "10 first", [wfo.name for wfo in wfos[:10]]
print "10 last", [wfo.name for wfo in wfos[-10:]]
else:
random.shuffle(wfos)
for wfo in wfos:
if options.limit and (n_stalled + n_assigned) > options.limit:
break
if max_per_round and (n_stalled + n_assigned) > max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name, specific.split(','))):
continue
#if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo(url, wfo.name)
if wfh.request['RequestStatus'] in [
'rejected', 'aborted', 'aborted-completed', 'aborted-archived',
'rejected-archived'
] and wfh.isRelval():
wfo.status = 'forget'
session.commit()
n_stalled += 1
continue
if options.priority and int(
wfh.request['RequestPriority']) < options.priority:
continue
options_text = ""
if options.early: options_text += ", early option is ON"
if options.partial:
options_text += ", partial option is ON"
options_text += ", good fraction is %.2f" % options.good_enough
wfh.sendLog('assignor',
"%s to be assigned%s" % (wfo.name, options_text))
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput, primary, parent, secondary,
sites_allowed) = wfh.getSiteWhiteList()
output_tiers = list(
set([o.split('/')[-1] for o in wfh.request['OutputDatasets']]))
if not output_tiers:
n_stalled += 1
wfh.sendLog('assignor', 'There is no output at all')
sendLog('assignor',
'Workflow %s has no output at all' % (wfo.name),
level='critical')
continue
is_stuck = (all_stuck & primary)
if is_stuck:
wfh.sendLog('assignor',
"%s are stuck input" % (','.join(is_stuck)))
## check if by configuration we gave it a GO
no_go = False
if not wfh.go(log=True) and not options.go:
no_go = True
allowed_secondary = {}
assign_parameters = {}
check_secondary = (not wfh.isRelval())
for campaign in wfh.getCampaigns():
if campaign in CI.campaigns:
assign_parameters.update(CI.campaigns[campaign])
if campaign in CI.campaigns and 'secondaries' in CI.campaigns[
campaign]:
if CI.campaigns[campaign]['secondaries']:
allowed_secondary.update(
CI.campaigns[campaign]['secondaries'])
check_secondary = True
if campaign in CI.campaigns and 'banned_tier' in CI.campaigns[
campaign]:
banned_tier = list(
set(CI.campaigns[campaign]['banned_tier'])
& set(output_tiers))
if banned_tier:
no_go = True
wfh.sendLog(
'assignor', 'These data tiers %s are not allowed' %
(','.join(banned_tier)))
sendLog('assignor',
'These data tiers %s are not allowed' %
(','.join(banned_tier)),
level='critical')
if secondary and check_secondary:
if (set(secondary) & set(allowed_secondary.keys()) !=
set(secondary)):
wfh.sendLog(
'assignor', '%s is not an allowed secondary' %
(', '.join(set(secondary) -
set(allowed_secondary.keys()))))
sendLog(
'assignor',
'%s is not an allowed secondary' %
(', '.join(set(secondary) -
set(allowed_secondary.keys()))),
level='critical')
if not options.go:
no_go = True
## then get whether there is something more to be done by secondary
for sec in secondary:
if sec in allowed_secondary: # and 'parameters' in allowed_secondary[sec]:
assign_parameters.update(allowed_secondary[sec])
if no_go:
n_stalled += 1
## make a very loud noise if >100k priority stalled
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] != 'assignment-approved':
if not options.test:
wfh.sendLog('assignor',
"setting %s away and skipping" % wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name, wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version = wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor', "cannot decide on version number")
n_stalled += 1
wfo.status = 'trouble'
session.commit()
continue
original_sites_allowed = copy.deepcopy(sites_allowed)
wfh.sendLog('assignor', "Site white list %s" % sorted(sites_allowed))
override_sec_location = CI.get(wfh.request['Campaign'],
'SecondaryLocation', [])
blocks = wfh.getBlockWhiteList()
rwl = wfh.getRunWhiteList()
if rwl:
## augment with run white list
for dataset in primary:
blocks = list(set(blocks +
getDatasetBlocks(dataset, runs=rwl)))
lwl = wfh.getLumiWhiteList()
if lwl:
## augment with lumi white list
for dataset in primary:
blocks = list(
set(blocks + getDatasetBlocks(dataset, lumis=lwl)))
wfh.sendLog('assignor', "Allowed %s" % sorted(sites_allowed))
secondary_locations = None
primary_aaa = options.primary_aaa
secondary_aaa = options.secondary_aaa
do_partial = False #options.good_enough if options.partial else 0
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns:
assign_parameters.update(CI.campaigns[wfh.request['Campaign']])
if 'primary_AAA' in assign_parameters:
primary_aaa = primary_aaa or assign_parameters['primary_AAA']
if 'secondary_AAA' in assign_parameters:
secondary_aaa = secondary_aaa or assign_parameters['secondary_AAA']
if 'partial_copy' in assign_parameters:
## can this only work if there is a stuck input ? maybe not
## this is a number. 0 means no
print "Could do partial disk copy assignment"
if is_stuck or options.partial:
do_partial = assign_parameters['partial_copy']
wfh.sendLog(
'assignor',
"Overiding partial copy assignment to %.2f fraction" %
do_partial)
#sendEmail('stuck input to assignment','%s is stuck for assigning %s and going fractional'%(','.join( is_stuck), wfo.name))
do_partial = options.good_enough if options.partial else do_partial
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
#sendEmail("tempting to pass sec location check","but we cannot yet IMO")
#pass
presence = getDatasetPresence(url, sec)
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [
site for (site, (there, frac)) in presence.items()
if frac > 98.
]
if secondary_aaa:
if not one_secondary_locations:
sec_availability = getDatasetBlocksFraction(url, sec)
if sec_availability >= 1. and options.go:
## there is at least one copy of each block on disk. We should go ahead and let it go.
wfh.sendLog(
'assignor',
"The secondary %s is available %s times on disk, and usable"
% (sec, sec_availability))
else:
## not even a copy on disk anywhere !!!!
sites_allowed = [] ## will block the assignment
wfh.sendLog(
'assignor',
"The secondary %s is nowhere on disk" % sec)
#just continue without checking
continue
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations == None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(
set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [
site for site in sites_allowed
if SI.CE_to_SE(site) in one_secondary_locations
]
wfh.sendLog(
'assignor', "From/after secondary requirement, now Allowed%s" %
sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy(
sites_allowed
) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy(sites_allowed)
sites_with_data = copy.deepcopy(sites_allowed)
sites_with_any_data = copy.deepcopy(sites_allowed)
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
endpoints = set()
for prim in list(primary):
if prim in dataset_endpoints:
print "endpoints from stagor", dataset_endpoints[prim]
endpoints.update(dataset_endpoints[prim])
set_lfn = getLFNbase(prim)
presence = getDatasetPresence(url, prim, only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(
url,
prim,
sites=[SI.CE_to_SE(site) for site in sites_allowed],
only_blocks=blocks)
if primary_aaa:
available_fractions[prim] = getDatasetBlocksFraction(
url, prim, only_blocks=blocks)
sites_all_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in [
psite for (psite, (there, frac)) in presence.items()
if there
]
]
if primary_aaa:
sites_all_data = list(
set([
SI.SE_to_CE(psite)
for (psite, (there, frac)) in presence.items() if there
]))
sites_with_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in
[psite for (psite, frac) in presence.items() if frac[1] > 90.]
]
sites_with_any_data = [
site for site in sites_with_any_data
if SI.CE_to_SE(site) in presence.keys()
]
if primary_aaa:
sites_with_any_data = list(
set([SI.SE_to_CE(psite) for psite in presence.keys()]))
wfh.sendLog(
'assignor', "Holding the data but not allowed %s" % sorted(
list(
set([
se_site for se_site in presence.keys()
if not SI.SE_to_CE(se_site) in sites_allowed
]))))
if primary_locations == None:
primary_locations = presence.keys()
else:
primary_locations = list(
set(primary_locations) & set(presence.keys()))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites = []
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in
list((set(secondary_locations) & set(primary_locations)) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
elif primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in list(
set(primary_locations) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
else:
opportunistic_sites = []
wfh.sendLog(
'assignor', "We could be running in addition at %s" %
sorted(opportunistic_sites))
if any(
[osite in SI.sites_not_ready
for osite in opportunistic_sites]):
wfh.sendLog(
'assignor', "One of the usable site is in downtime %s" % ([
osite for osite in opportunistic_sites
if osite in SI.sites_not_ready
]))
down_time = True
## should this be send back to considered ?
## should be 2 but for the time-being let's lower it to get things going
copies_wanted, cpuh = wfh.getNCopies()
wfh.sendLog('assignor', "we need %s CPUh" % cpuh)
if cpuh > max_cpuh_block and not options.go:
#sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog(
'assignor',
'%s requires a large numbr of CPUh %s , not assigning, please check with requester'
% (wfo.name, cpuh),
level='critical')
wfh.sendLog(
'assignor',
"Requiring a large number of CPUh %s, not assigning" % cpuh)
continue
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns and 'maxcopies' in CI.campaigns[
wfh.request['Campaign']]:
copies_needed_from_campaign = CI.campaigns[
wfh.request['Campaign']]['maxcopies']
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(
1, copies_wanted -
less_copies_than_requested) # take one out for the efficiency
else:
## find out whether there is a site in the whitelist, that is lacking jobs and reduce to 1 copy needed to get things going
pass
wfh.sendLog('assignor',
"needed availability fraction %s" % copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
## default back to white list to original white list with any data
wfh.sendLog('assignor', "Allowed sites :%s" % sorted(sites_allowed))
if primary_aaa:
## remove the sites not reachable localy if not in having the data
if not sites_all_data:
wfh.sendLog('assignor',
"Overiding the primary on AAA setting to Off")
primary_aaa = False
else:
aaa_grid = set(sites_all_data)
for site in list(aaa_grid):
aaa_grid.update(aaa_mapping.get(site, []))
sites_allowed = list(set(initial_sites_allowed) & aaa_grid)
wfh.sendLog(
'assignor', "Selected to read primary through xrootd %s" %
sorted(sites_allowed))
isStoreResults = ('StoreResults' == wfh.request.setdefault(
'RequestType', None))
if isStoreResults:
if 'MergedLFNBase' in wfh.request:
set_lfn = wfh.request['MergedLFNBase']
else:
n_stalled += 1
wfh.sendLog(
'assignor',
"Cannot assign StoreResults request because MergedLFN is missing"
)
sendLog(
'assignor',
'Cannot assign StoreResults request because MergedLFN is missing',
level='critical')
continue
if not primary_aaa:
if not isStoreResults:
sites_allowed = sites_with_any_data
else:
## if we are dealing with a StoreResults request, we don't need to check dataset availability and
## should use the SiteWhiteList set in the original request
if 'SiteWhitelist' in wfh.request:
sites_allowed = wfh.request['SiteWhitelist']
else:
wfh.sendLog(
'assignor',
"Cannot assign StoreResults request because SiteWhitelist is missing"
)
sendLog(
'assignor',
'Cannot assign StoreResults request because SiteWhitelist is missing',
level='critical')
n_stalled += 1
continue
available_fractions = {}
wfh.sendLog('assignor',
"Selected for any data %s" % sorted(sites_allowed))
### check on endpoints for on-going transfers
if do_partial:
if endpoints:
end_sites = [SI.SE_to_CE(s) for s in endpoints]
sites_allowed = list(set(sites_allowed + end_sites))
if down_time and not any(osite in SI.sites_not_ready
for osite in end_sites):
print "Flip the status of downtime, since our destinations are good"
down_time = False
print "with added endpoints", sorted(end_sites)
else:
print "Cannot do partial assignment without knowin the endpoints"
n_stalled += 1
continue
#if not len(sites_allowed):
# if not options.early:
# wfh.sendLog('assignor',"cannot be assign with no matched sites")
# sendLog('assignor','%s has no whitelist'% wfo.name, level='critical')
# n_stalled+=1
# continue
low_pressure = SI.sites_low_pressure(0.4)
## if any of the site allowed is low pressure : reduce to 1 copy so that it gets started
allowed_and_low = sorted(set(low_pressure) & set(sites_allowed))
if allowed_and_low:
wfh.sendLog(
'assignor',
"The workflow can run at %s under low pressure currently" %
(','.join(allowed_and_low)))
copies_wanted = max(1., copies_wanted - 1.)
if available_fractions and not all([
available >= copies_wanted
for available in available_fractions.values()
]):
not_even_once = not all([
available >= 1. for available in available_fractions.values()
])
above_good = all([
available >= do_partial
for available in available_fractions.values()
])
wfh.sendLog(
'assignor',
"The input dataset is not available %s times, only %s" %
(copies_wanted, available_fractions.values()))
if down_time and not options.go and not options.early:
wfo.status = 'considered'
session.commit()
wfh.sendLog(
'assignor',
"sending back to considered because of site downtime, instead of waiting"
)
#sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
sendLog(
'assignor',
'%s is not sufficiently available, due to down time of a site in the whitelist. sending back to considered.'
% (wfo.name),
level='delay')
n_stalled += 1
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known and not options.limit and not options.go and not options.early and not (
do_partial and above_good):
wfh.sendLog(
'assignor',
"cannot be assigned, %s is not sufficiently available.\n %s"
% (wfo.name, json.dumps(available_fractions)))
#sendEmail( "cannot be assigned","%s is not sufficiently available.\n %s"%(wfo.name,json.dumps(available_fractions)))
known.append(wfo.name)
open('cannot_assign.json',
'w').write(json.dumps(known, indent=2))
if options.early:
if wfo.status == 'considered':
wfh.sendLog('assignor', "setting considered-tried")
wfo.status = 'considered-tried'
session.commit()
else:
print "tried but status is", wfo.status
if do_partial and above_good:
print "Will move on with partial locations"
else:
n_stalled += 1
continue
if not len(sites_allowed):
if not options.early:
wfh.sendLog('assignor',
"cannot be assign with no matched sites")
sendLog('assignor',
'%s has no whitelist' % wfo.name,
level='critical')
n_stalled += 1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [
SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])
]
wfh.sendLog('assignor', "Placing the output on %s" % sites_out)
parameters = {
'SiteWhitelist': sites_allowed,
'NonCustodialSites': sites_out,
'AutoApproveSubscriptionSites': list(set(sites_out)),
'AcquisitionEra': wfh.acquisitionEra(),
'ProcessingString': wfh.processingString(),
'MergedLFNBase': set_lfn,
'ProcessingVersion': version,
}
if primary_aaa:
parameters['TrustSitelists'] = True
wfh.sendLog(
'assignor',
"Reading primary through xrootd at %s" % sorted(sites_allowed))
if secondary_aaa:
parameters['TrustPUSitelists'] = True
wfh.sendLog(
'assignor', "Reading secondary through xrootd at %s" %
sorted(sites_allowed))
## plain assignment here
team = 'production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
parameters['Team'] = team
if lheinput:
## throttle reading LHE article
wfh.sendLog('assignor',
'Setting the number of events per job to 500k max')
parameters['EventsPerJob'] = 500000
def pick_options(options, parameters):
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v = getattr(options, key)
if v != None:
if type(v) == str and ',' in v:
parameters[key] = filter(None, v.split(','))
else:
parameters[key] = v
def pick_campaign(assign_parameters, parameters):
## pick up campaign specific assignment parameters
parameters.update(assign_parameters.get('parameters', {}))
if options.force_options:
pick_campaign(assign_parameters, parameters)
pick_options(options, parameters)
else:
## campaign parameters update last
pick_options(options, parameters)
pick_campaign(assign_parameters, parameters)
if not options.test:
parameters['execute'] = True
hold_split, split_check = wfh.checkSplitting()
if hold_split and not options.go:
if split_check:
wfh.sendLog(
'assignor', 'Holding on to the change in splitting %s' %
('\n\n'.join([str(i) for i in split_check])))
else:
wfh.sendLog('assignor', 'Change of splitting is on hold')
n_stalled += 1
continue
if split_check == None or split_check == False:
n_stalled += 1
continue
elif split_check:
## operate all recommended changes
reqMgrClient.setWorkflowSplitting(url, wfo.name, split_check)
wfh.sendLog(
'assignor', 'Applying the change in splitting %s' %
('\n\n'.join([str(i) for i in split_check])))
split_check = True ## bypass completely and use the above
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi) / float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents / (reqJobs * 1.4))
lumisPerJob = int(eventsPerJob / eventsPerLumi)
if lumisPerJob == 0:
#sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog('assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob),
level='critical')
wfh.sendLog(
'assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl[
'events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl[
'avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
#sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog('assignor',
"%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob),
level='critical')
wfh.sendLog(
'assignor', "%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
#sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog(
'assignor',
"leaving splitting untouched for %s, please check on %s"
% (pstring, wfo.name),
level='critical')
wfh.sendLog(
'assignor',
"leaving splitting untouched for PU_RD*, please check."
)
if isHEPCloudReady(url) and wfh.isGoodForNERSC():
parameters['Team'] = 'hepcloud'
parameters['SiteWhitelist'] = ['T3_US_NERSC']
if primary:
parameters['TrustSitelists'] = True
if secondary:
parameters['TrustPUSitelists'] = True
sendEmail("sending work to hepcloud",
"pleasse check on %s" % wfh.request['RequestName'],
destination=['*****@*****.**'])
## make sure to autoapprove all NonCustodialSites
parameters['AutoApproveSubscriptionSites'] = list(
set(parameters['NonCustodialSites'] +
parameters.get('AutoApproveSubscriptionSites', [])))
result = reqMgrClient.assignWorkflow(
url, wfo.name, None,
parameters) ## team is not relevant anymore here
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned += 1
wfh.sendLog(
'assignor', "Properly assigned\n%s" %
(json.dumps(parameters, indent=2)))
try:
## refetch information and lock output
new_wfi = workflowInfo(url, wfo.name)
(_, prim, _, sec) = new_wfi.getIO()
for secure in list(prim) + list(
sec) + new_wfi.request['OutputDatasets']:
## lock all outputs
LI.lock(secure, reason='assigning')
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output", str(e))
else:
wfh.sendLog(
'assignor',
"Failed to assign %s.\n%s \n Please check the logs" %
(wfo.name, reqMgrClient.assignWorkflow.errorMessage))
sendLog('assignor',
"Failed to assign %s.\n%s \n Please check the logs" %
(wfo.name, reqMgrClient.assignWorkflow.errorMessage),
level='critical')
print "ERROR could not assign", wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor', "Assigned %d Stalled %s" % (n_assigned, n_stalled))
if n_stalled and not options.go and not options.early:
sendLog('assignor',
"%s workflows cannot be assigned. Please take a look" %
(n_stalled),
level='critical')
def parse_top(url, options=None):
UC = unifiedConfiguration()
top_N = UC.get('full_report_top_N')
diagnose_by_agent_by_site = defaultdict(
lambda: defaultdict(lambda: defaultdict(dict)))
wm = dataCache.get('wmstats')
for wfn in wm.keys():
## filter by runn*
if not wm[wfn]['RequestStatus'] in [
'running-open',
'running-closed',
#'completed',
]:
continue
info = wm[wfn].get('AgentJobInfo', {})
for agent, ai in info.items():
an = agent.split('.')[0]
for task, ti in ai['tasks'].items():
for site, si in ti.get('sites', {}).items():
ssi = condensed(si)
diagnose_by_agent_by_site[task][an][site] = ssi
diagnose = defaultdict(dict) ## the overall picture of the task
diagnose_by_site = defaultdict(lambda: defaultdict(dict))
diagnose_by_agent = defaultdict(lambda: defaultdict(dict))
for task in diagnose_by_agent_by_site:
if '_ACDC' in task: continue
if '_RVCMSSW' in task: continue
for agent in diagnose_by_agent_by_site[task]:
for site in diagnose_by_agent_by_site[task][agent]:
diagnose_by_site[task][site] = add_condensed(
diagnose_by_site[task][site],
diagnose_by_agent_by_site[task][agent][site])
diagnose_by_agent[task][agent] = add_condensed(
diagnose_by_agent[task][agent],
diagnose_by_agent_by_site[task][agent][site])
diagnose[task] = add_condensed(
diagnose[task],
diagnose_by_agent_by_site[task][agent][site])
##include fractions
for t, ti in diagnose.items():
diagnose[t].update(ratios(ti))
for t in diagnose_by_site:
for s, ti in diagnose_by_site[t].items():
diagnose_by_site[t][s].update(ratios(ti))
for t in diagnose_by_agent:
for a, ti in diagnose_by_agent[t].items():
diagnose_by_agent[t][a].update(ratios(ti))
for t in diagnose_by_agent_by_site:
for a in diagnose_by_agent_by_site[t]:
for s, ti in diagnose_by_agent_by_site[t][a].items():
diagnose_by_agent_by_site[t][a][s].update(ratios(ti))
#top_cooloff = dict(sorted([(t,i.get('cooloff',0)) for t,i in diagnose.items()], key = lambda o:o[1], reverse=True)[:top_N])
#top_failure = dict(sorted([(t,i.get('failure',0)) for t,i in diagnose.items()], key = lambda o:o[1], reverse=True)[:top_N])
top_cooloff = dict(
sorted([(t, i.get('cooloff_f', 0)) for t, i in diagnose.items()],
key=lambda o: o[1],
reverse=True)[:top_N])
top_failure = dict(
sorted([(t, i.get('failure_f', 0)) for t, i in diagnose.items()],
key=lambda o: o[1],
reverse=True)[:top_N])
all_bad_wfs = set()
all_bad_wfs.update([t.split('/')[1] for t in top_cooloff.keys()])
all_bad_wfs.update([t.split('/')[1] for t in top_failure.keys()])
print "found", len(all_bad_wfs), "to parse for detailled error report"
parse_those(url, options, all_bad_wfs)
ht = open('%s/toperror.html' % monitor_eos_dir, 'w')
ht.write("""<html>
Report of workflows with top %s error in failure and cooloff<br>
Last updated on %s (GMT)
<table border=1>
<thead><tr><th>Workflow</th><th>Task</th><th>Success</th><th>Failures</th><th>Fail.Frac</th><th>Cooloffs</th><th>Cool.Frac</th><th> Task Error Report </th> </tr></thead>
""" % (top_N, time.asctime(time.gmtime())))
## sort by max errors
tops = defaultdict(int)
for wf in sorted(all_bad_wfs):
for owf, N in top_cooloff.items():
if wf in owf:
tops[wf] += N
for owf, N in top_failure.items():
if wf in owf:
tops[wf] += N
for iw, (wf, count) in enumerate(
sorted(tops.items(), key=lambda o: o[1], reverse=True)):
report = set()
for owf, N in top_cooloff.items():
if wf in owf:
report.add(owf)
for owf, N in top_failure.items():
if wf in owf:
report.add(owf)
lcol = 'bgcolor=%s' % ('white' if iw % 2 == 0 else 'lightblue')
for task in sorted(report):
tdiag = diagnose.get(task, {})
print task
print tdiag
ht.write(
'<tr %s><td>%s</td><td>%s</td><td>%s</td><td>%s</td><td>%s</td><td>%s</td><td>%s</td><td>%s</td></tr>\n'
% (lcol, wf, task.split('/')[-1], tdiag.get('success', '-'),
tdiag.get('failure', '-'), tdiag.get('failure_f', '-'),
tdiag.get('cooloff', '-'), tdiag.get('cooloff_f', '-'),
'<a href=%s/report/%s#%s> report </a>' %
(unified_url, wf, task.split('/')[-1])))
ht.write('</table></html>')
ht.close()
def checkor(url, spec=None, options=None):
if userLock(): return
if duplicateLock() and not options.go: return
fDB = closeoutInfo()
UC = unifiedConfiguration()
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
def time_point(label="",sub_lap=False):
now = time.mktime(time.gmtime())
nows = time.asctime(time.gmtime())
print "Time check (%s) point at : %s"%(label, nows)
print "Since start: %s [s]"% ( now - time_point.start)
if sub_lap:
print "Sub Lap : %s [s]"% ( now - time_point.sub_lap )
time_point.sub_lap = now
else:
print "Lap : %s [s]"% ( now - time_point.lap )
time_point.lap = now
time_point.sub_lap = now
time_point.sub_lap = time_point.lap = time_point.start = time.mktime(time.gmtime())
runnings = session.query(Workflow).filter(Workflow.status == 'away').all()
standings = session.query(Workflow).filter(Workflow.status.startswith('assistance')).all()
## intersect with what is actually in completed status in request manager now
all_completed = set(getWorkflows(url, 'completed' ))
wfs=[]
if options.strict:
## the one which were running and now have completed
print "strict option is on: checking workflows that freshly completed"
wfs.extend( filter(lambda wfo: wfo.name in all_completed , runnings))
if options.update:
print "update option is on: checking workflows that have not completed yet"
wfs.extend( filter(lambda wfo: not wfo.name in all_completed , runnings))
if options.clear:
print "clear option is on: checking workflows that are ready to toggle closed-out"
wfs.extend( filter(lambda wfo: 'custodial' in wfo.status, standings))
if options.review:
print "review option is on: checking the workflows that needed intervention"
wfs.extend( filter(lambda wfo: not 'custodial' in wfo.status, standings))
## what is left out are the wf which were running and ended up aborted/failed/...
custodials = defaultdict(list) #sites : dataset list
transfers = defaultdict(list) #sites : dataset list
invalidations = [] #a list of files
SI = siteInfo()
CI = campaignInfo()
mcm = McMClient(dev=False) if use_mcm else None
def get_campaign(output, wfi):
## this should be a perfect matching of output->task->campaign
campaign = None
era = None
wf_campaign = None
if 'Campaign' in wfi.request: wf_campaign = wfi.request['Campaign']
try:
era = output.split('/')[2].split('-')[0]
except:
era = None
if wfi.isRelval():
campaign = wf_campaign
else:
campaign = era if era else wf_campaign
return campaign
## retrieve bypass and onhold configuration
bypasses = []
forcings = []
overrides = getForceCompletes()
holdings = []
actors = UC.get('allowed_bypass')
for bypassor,email in actors:
bypass_file = '/afs/cern.ch/user/%s/%s/public/ops/bypass.json'%(bypassor[0],bypassor)
if not os.path.isfile(bypass_file):
#sendLog('checkor','no file %s',bypass_file)
continue
try:
bypasses.extend( json.loads(open(bypass_file).read()))
except:
sendLog('checkor',"cannot get by-passes from %s for %s"%(bypass_file ,bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(bypass_file), destination=[email])
holding_file = '/afs/cern.ch/user/%s/%s/public/ops/onhold.json'%(bypassor[0],bypassor)
if not os.path.isfile(holding_file):
#sendLog('checkor',"no file %s"%holding_file)
continue
try:
extending = json.loads(open(holding_file).read())
print bypassor,"is holding",extending
holdings.extend( extending )
except:
sendLog('checkor',"cannot get holdings from %s for %s"%(holding_file, bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(holding_file), destination=[email])
## once this was force-completed, you want to bypass
for rider,email in actors:
rider_file = '/afs/cern.ch/user/%s/%s/public/ops/forcecomplete.json'%(rider[0],rider)
if not os.path.isfile(rider_file):
print "no file",rider_file
#sendLog('checkor',"no file %s"%rider_file)
continue
try:
bypasses.extend( json.loads(open( rider_file ).read() ) )
except:
sendLog('checkor',"cannot get force complete list from %s"%rider)
sendEmail("malformated force complet file","%s is not json readable"%rider_file, destination=[email])
if use_mcm:
forcings = mcm.get('/restapi/requests/forcecomplete')
#if forcings:
# sendEmail('force completing mechanism','please check what checkor is doing with %s'%( ','.join(forcings)))
pattern_fraction_pass = UC.get('pattern_fraction_pass')
total_running_time = 5.*60.
sleep_time = 1
if len(wfs):
sleep_time = min(max(0.5, total_running_time / len(wfs)), 10)
random.shuffle( wfs )
in_manual = 0
## now you have a record of what file was invalidated globally from TT
TMDB_invalid = dataCache.get('file_invalidation')
#try:
# TMDB_invalid = set([row[3] for row in csv.reader( os.popen('curl -s "https://docs.google.com/spreadsheets/d/11fFsDOTLTtRcI4Q3gXw0GNj4ZS8IoXMoQDC3CbOo_2o/export?format=csv"'))])
# TMDB_invalid = map(lambda e : e.split(':')[-1], TMDB_invalid)
# print len(TMDB_invalid),"globally invalidated files"
#except Exception as e:
# print "TMDB not fetched"
# print str(e)
# TMDB_invalid = []
print len(wfs),"to consider, pausing for",sleep_time
max_per_round = UC.get('max_per_round').get('checkor',None)
if options.limit: max_per_round=options.limit
if max_per_round and not spec: wfs = wfs[:max_per_round]
for wfo in wfs:
if spec and not (spec in wfo.name): continue
time.sleep( sleep_time )
time_point("Starting with %s"% wfo.name)
## get info
wfi = workflowInfo(url, wfo.name)
wfi.sendLog('checkor',"checking on %s %s"%( wfo.name,wfo.status))
## make sure the wm status is up to date.
# and send things back/forward if necessary.
wfo.wm_status = wfi.request['RequestStatus']
if wfo.wm_status == 'closed-out':
## manually closed-out
wfi.sendLog('checkor',"%s is already %s, setting close"%( wfo.name , wfo.wm_status))
wfo.status = 'close'
session.commit()
continue
elif wfo.wm_status in ['failed','aborted','aborted-archived','rejected','rejected-archived','aborted-completed']:
## went into trouble
wfo.status = 'trouble'
wfi.sendLog('checkor',"%s is in trouble %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
elif wfo.wm_status in ['assigned','acquired']:
## not worth checking yet
wfi.sendLog('checkor',"%s is not running yet"%wfo.name)
session.commit()
continue
if '-onhold' in wfo.status:
if wfo.name in holdings and wfo.name not in bypasses:
wfi.sendLog('checkor',"%s is on hold"%wfo.name)
continue
if wfo.wm_status != 'completed': #and not wfo.name in bypasses:
## for sure move on with closeout check if in completed
wfi.sendLog('checkor',"no need to check on %s in status %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
if wfo.name in holdings and wfo.name not in bypasses:
wfo.status = 'assistance-onhold'
wfi.sendLog('checkor',"setting %s on hold"%wfo.name)
session.commit()
continue
session.commit()
#sub_assistance="" # if that string is filled, there will be need for manual assistance
existing_assistance_tags = set(wfo.status.split('-')[1:]) #[0] should be assistance
assistance_tags = set()
is_closing = True
## get it from somewhere
bypass_checks = False
for bypass in bypasses:
if bypass in wfo.name:
wfi.sendLog('checkor',"we can bypass checks on %s because of keyword %s "%( wfo.name, bypass))
bypass_checks = True
break
pids = wfi.getPrepIDs()
force_by_mcm = False
force_by_user = False
for force in forcings:
if force in pids:
wfi.sendLog('checkor',"we can bypass checks and force complete %s because of prepid %s "%( wfo.name, force))
bypass_checks = True
force_by_mcm = True
break
for user in overrides:
for force in overrides[user]:
if force in wfo.name:
wfi.sendLog('checkor',"we can bypass checks and force complete %s because of keyword %s of user %s"%( wfo.name, force, user))
bypass_checks = True
force_by_user = True
break
tiers_with_no_check = copy.deepcopy(UC.get('tiers_with_no_check')) # dqm*
vetoed_custodial_tier = copy.deepcopy(UC.get('tiers_with_no_custodial')) #dqm*, reco
to_ddm_tier = copy.deepcopy(UC.get('tiers_to_DDM'))
campaigns = {} ## this mapping of campaign per output dataset assumes era==campaing, which is not true for relval
expected_outputs = copy.deepcopy( wfi.request['OutputDatasets'] )
for out in wfi.request['OutputDatasets']:
c = get_campaign(out, wfi)
campaigns[out] = c
if c in CI.campaigns and 'custodial_override' in CI.campaigns[c]:
vetoed_custodial_tier = list(set(vetoed_custodial_tier) - set(CI.campaigns[c]['custodial_override']))
## add those that we need to check for custodial copy
tiers_with_no_check = list(set(tiers_with_no_check) - set(CI.campaigns[c]['custodial_override'])) ## would remove DQM from the vetoed check
check_output_text = "Initial outputs:"+",".join(sorted(wfi.request['OutputDatasets'] ))
wfi.request['OutputDatasets'] = [ out for out in wfi.request['OutputDatasets'] if not any([out.split('/')[-1] == veto_tier for veto_tier in tiers_with_no_check])]
check_output_text += "\nWill check on:"+",".join(sorted(wfi.request['OutputDatasets'] ))
check_output_text += "\ntiers out:"+",".join( sorted(tiers_with_no_check ))
check_output_text += "\ntiers no custodial:"+",".join( sorted(vetoed_custodial_tier) )
wfi.sendLog('checkor', check_output_text )
## anything running on acdc : getting the real prepid is not worth it
familly = getWorkflowById(url, wfi.request['PrepID'], details=True)
acdc = []
acdc_inactive = []
forced_already=False
acdc_bads = []
true_familly = []
for member in familly:
if member['RequestType'] != 'Resubmission': continue
if member['RequestName'] == wfo.name: continue
if member['RequestDate'] < wfi.request['RequestDate']: continue
if member['PrepID'] != wfi.request['PrepID'] : continue
#if 'OriginalRequestName' in member and (not 'ACDC' in member['OriginalRequestName']) and member['OriginalRequestName'] != wfo.name: continue
if member['RequestStatus'] == None: continue
if not set(member['OutputDatasets']).issubset( set(expected_outputs)):
if not member['RequestStatus'] in ['rejected-archived','rejected','aborted','aborted-archived']:
##this is not good at all
wfi.sendLog('checkor','inconsistent ACDC %s'%member['RequestName'] )
#sendLog('checkor','inconsistent ACDC %s'%member['RequestName'], level='critical')
acdc_bads.append( member['RequestName'] )
is_closing = False
assistance_tags.add('manual')
continue
true_familly.append( member['RequestName'] )
#try:
# parse_one(url, member['RequestName'])
#except:
# print "Could not make error report for",member['RequestName']
if member['RequestStatus'] in ['running-open','running-closed','assigned','acquired']:
print wfo.name,"still has an ACDC running",member['RequestName']
acdc.append( member['RequestName'] )
## cannot be bypassed!
is_closing = False
assistance_tags.add('recovering')
if (force_by_mcm or force_by_user) and not forced_already:
wfi.sendLog('checkor','%s is being forced completed while recovering'%wfo.name)
wfi.notifyRequestor("The workflow %s was force completed"% wfo.name, do_batch=False)
forceComplete(url, wfi)
forced_already=True
else:
acdc_inactive.append( member['RequestName'] )
assistance_tags.add('recovered')
if acdc_bads:
#sendEmail('inconsistent ACDC','for %s, ACDC %s is inconsistent, preventing from closing'%( wfo.name, ','.join(acdc_bads) ))
sendLog('checkor','For %s, ACDC %s is inconsistent, preventing from closing or will create a mess.'%( wfo.name, ','.join(acdc_bads) ), level='critical')
time_point("checked workflow familly", sub_lap=True)
## completion check
percent_completions = {}
if not 'TotalInputEvents' in wfi.request:
event_expected,lumi_expected = 0,0
if not 'recovery' in wfo.status:
#sendEmail("missing member of the request","TotalInputEvents is missing from the workload of %s"% wfo.name, destination=['*****@*****.**'])
sendLog('checkor',"TotalInputEvents is missing from the workload of %s"% wfo.name, level='critical')
else:
event_expected,lumi_expected = wfi.request['TotalInputEvents'],wfi.request['TotalInputLumis']
if 'RequestNumEvents' in wfi.request and int(wfi.request['RequestNumEvents']):
event_expected = int(wfi.request['RequestNumEvents'])
elif 'Task1' in wfi.request and 'RequestNumEvents' in wfi.request['Task1']:
event_expected = wfi.request['Task1']['RequestNumEvents']
for i in range(1,20):
if 'Task%d'%i in wfi.request:
## this is wrong ibsolute
if 'FilterEfficiency' in wfi.request['Task%d'%i]:
event_expected *= float(wfi.request['Task%d'%i]['FilterEfficiency'])
event_expected = int(event_expected)
fractions_pass = {}
events_per_lumi = {}
over_100_pass = False
(lhe,prim,_,_) = wfi.getIO()
if lhe or prim: over_100_pass = False
time_point("execpted statistics", sub_lap=True)
for output in wfi.request['OutputDatasets']:
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
events_per_lumi[output] = event_count/float(lumi_count) if lumi_count else 100
percent_completions[output] = 0.
if lumi_expected:
percent_completions[output] = lumi_count / float( lumi_expected )
if event_expected:
wfi.sendLog('checkor', "event completion real %s expected %s"%(event_count, event_expected ))
percent_completions[output] = max(percent_completions[output], float(event_count) / float( event_expected ) )
default_pass = UC.get('default_fraction_pass')
fractions_pass[output] = default_pass
c = campaigns[output]
if c in CI.campaigns and 'fractionpass' in CI.campaigns[c]:
if type(CI.campaigns[c]['fractionpass']) == dict:
tier = output.split('/')[-1]
priority = str(wfi.request['RequestPriority'])
## defined per tier
fractions_pass[output] = CI.campaigns[c]['fractionpass'].get('all', default_pass)
if tier in CI.campaigns[c]['fractionpass']:
fractions_pass[output] = CI.campaigns[c]['fractionpass'][tier]
if priority in CI.campaigns[c]['fractionpass']:
fractions_pass[output] = CI.campaigns[c]['fractionpass'][priority]
else:
fractions_pass[output] = CI.campaigns[c]['fractionpass']
wfi.sendLog('checkor', "overriding fraction to %s for %s by campaign requirement"%( fractions_pass[output], output))
if options.fractionpass:
fractions_pass[output] = options.fractionpass
print "overriding fraction to",fractions_pass[output],"by command line for",output
for key in pattern_fraction_pass:
if key in output:
fractions_pass[output] = pattern_fraction_pass[key]
print "overriding fraction to",fractions_pass[output],"by dataset key",key
if not all([percent_completions[out] >= fractions_pass[out] for out in fractions_pass]):
possible_recoveries = wfi.getRecoveryDoc()
if possible_recoveries == []:
wfi.sendLog('checkor','%s has missing statistics \n%s \n%s, but nothing is recoverable. passing through to annoucement'%(
wfo.name, json.dumps(percent_completions, indent=2), json.dumps(fractions_pass, indent=2) ))
sendLog('checkor','%s is not completed, but has nothing to be recovered, passing along ?'%wfo.name, level='critical')
#sendEmail('nothing is recoverable','%s is not completed, but has nothing to be recovered, passing along ?'%wfo.name)#,destination=['*****@*****.**'])
## do not bypass for now, until Alan understands why we are loosing ACDC docs
bypass_checks = True
else:
wfi.sendLog('checkor','%s is not completed \n%s \n%s'%(
wfo.name, json.dumps(percent_completions, indent=2), json.dumps(fractions_pass, indent=2) ))
## hook for creating automatically ACDC ?
if not bypass_checks:
assistance_tags.add('recovery')
is_closing = False
if over_100_pass and any([percent_completions[out] >100 for out in fractions_pass]):
print wfo.name,"is over completed"
print json.dumps(percent_completions, indent=2)
if not bypass_checks:
assistance_tags.add('over100')
is_closing = False
time_point("checked output size", sub_lap=True)
## correct lumi < 300 event per lumi
#for output in wfi.request['OutputDatasets']:
#events_per_lumi[output] = getDatasetEventsPerLumi( output )
lumi_upper_limit = {}
for output in wfi.request['OutputDatasets']:
upper_limit = 301.
campaign = campaigns[output]
if campaign in CI.campaigns and 'lumisize' in CI.campaigns[campaign]:
upper_limit = CI.campaigns[campaign]['lumisize']
print "overriding the upper lumi size to",upper_limit,"for",campaign
if options.lumisize:
upper_limit = options.lumisize
print "overriding the upper lumi size to",upper_limit,"by command line"
lumi_upper_limit[output] = upper_limit
if wfi.request['RequestType'] in ['ReDigi','ReReco']: lumi_upper_limit[output] = -1
if any([ (lumi_upper_limit[out]>0 and events_per_lumi[out] >= lumi_upper_limit[out]) for out in events_per_lumi]):
print wfo.name,"has big lumisections"
print json.dumps(events_per_lumi, indent=2)
## hook for rejecting the request ?
if not bypass_checks:
assistance_tags.add('biglumi')
is_closing = False
any_presence = {}
for output in wfi.request['OutputDatasets']:
any_presence[output] = getDatasetPresence(url, output, vetoes=[])
time_point("checked dataset presence", sub_lap=True)
## custodial copy
custodial_locations = {}
custodial_presences = {}
for output in wfi.request['OutputDatasets']:
custodial_presences[output] = [s for s in any_presence[output] if 'MSS' in s]
custodial_locations[output] = phedexClient.getCustodialSubscriptionRequestSite(output)
if not custodial_locations[output]:
custodial_locations[output] = []
time_point("checked custodiality", sub_lap=True)
## presence in phedex
phedex_presence ={}
for output in wfi.request['OutputDatasets']:
phedex_presence[output] = phedexClient.getFileCountDataset(url, output )
time_point("checked phedex count", sub_lap=True)
out_worth_checking = [out for out in custodial_locations.keys() if out.split('/')[-1] not in vetoed_custodial_tier]
size_worth_checking = sum([getDatasetSize(out)/1023. for out in out_worth_checking ]) ## size in TBs of all outputs
size_worht_going_to_ddm = sum([getDatasetSize(out)/1023. for out in out_worth_checking if out.split('/')[-1] in to_ddm_tier ]) ## size in TBs of all outputs
if not all(map( lambda sites : len(sites)!=0, [custodial_locations[out] for out in out_worth_checking])):
print wfo.name,"has not all custodial location"
print json.dumps(custodial_locations, indent=2)
##########
## hook for making a custodial replica ?
custodial = None
## get from other outputs
for output in out_worth_checking:
if len(custodial_locations[output]):
custodial = custodial_locations[output][0]
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the other output custodial:",custodial,"because of limited space"
custodial = None
## try to get it from campaign configuration
if not custodial:
for output in out_worth_checking:
campaign = campaigns[output]
if campaign in CI.campaigns and 'custodial' in CI.campaigns[campaign]:
custodial = CI.campaigns[campaign]['custodial']
print "Setting custodial to",custodial,"from campaign configuration"
group = None
if campaign in CI.campaigns and 'phedex_group' in CI.campaigns[campaign]:
group = CI.campaigns[campaign]['phedex_group']
print "using group",group,"for replica"
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the campaign configuration custodial:",custodial,"because of limited space"
custodial = None
## get from the parent
pick_custodial = True
use_parent_custodial = UC.get('use_parent_custodial')
tape_size_limit = options.tape_size_limit if options.tape_size_limit else UC.get("tape_size_limit")
_,prim,_,_ = wfi.getIO()
if not custodial and prim and use_parent_custodial:
parent_dataset = prim.pop()
## this is terribly dangerous to assume only
parents_custodial = phedexClient.getCustodialSubscriptionRequestSite( parent_dataset )
###parents_custodial = findCustodialLocation(url, parent_dataset)
if not parents_custodial:
parents_custodial = []
if len(parents_custodial):
custodial = parents_custodial[0]
else:
print "the input dataset",parent_dataset,"does not have custodial in the first place. abort"
#sendEmail( "dataset has no custodial location", "Please take a look at %s in the logs of checkor"%parent_dataset)
## does not work for RAWOADSIM
sendLog('checkor',"Please take a look at %s for missing custodial location"% parent_dataset)
## cannot be bypassed, this is an issue to fix
is_closing = False
pick_custodial = False
assistance_tags.add('parentcustodial')
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the parent custodial:",custodial,"because of limited space"
custodial = None
if not custodial and pick_custodial:
## pick one at random
custodial = SI.pick_SE(size=size_worth_checking)
if custodial and size_worht_going_to_ddm > tape_size_limit:
print wfi.sendLog('checkor',"The total output size (%s TB) is too large for the limit set (%s TB)"%( size_worth_checking, tape_size_limit))
custodial = None
if not custodial:
print "cannot find a custodial for",wfo.name
wfi.sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking))
sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking), level='critical')
if custodial and (is_closing or bypass_checks):
print "picked",custodial,"for tape copy"
## remember how much you added this round already ; this stays locally
SI.storage[custodial] -= size_worth_checking
## register the custodial request, if there are no other big issues
for output in out_worth_checking:
if not len(custodial_locations[output]):
if phedex_presence[output]>=1:
wfi.sendLog('checkor','Using %s as a tape destination for %s'%(custodial, output))
custodials[custodial].append( output )
if group: custodials[custodial][-1]+='@%s'%group
## let's wait and see if that's needed
assistance_tags.add('custodial')
else:
print "no file in phedex for",output," not good to add to custodial requests"
#cannot be bypassed
is_closing = False
time_point("determined tape location", sub_lap=True)
## disk copy
disk_copies = {}
for output in wfi.request['OutputDatasets']:
disk_copies[output] = [s for s in any_presence[output] if (not 'MSS' in s) and (not 'Buffer' in s)]
if not all(map( lambda sites : len(sites)!=0, disk_copies.values())):
print wfo.name,"has not all output on disk"
print json.dumps(disk_copies, indent=2)
## presence in dbs
dbs_presence = {}
dbs_invalid = {}
for output in wfi.request['OutputDatasets']:
dbs_presence[output] = dbs3Client.getFileCountDataset( output )
dbs_invalid[output] = dbs3Client.getFileCountDataset( output, onlyInvalid=True)
time_point("dbs file count", sub_lap=True)
if not all([dbs_presence[out] == (dbs_invalid[out]+phedex_presence[out]) for out in wfi.request['OutputDatasets']]) and not options.ignorefiles:
mismatch_notice = wfo.name+" has a dbs,phedex mismatch\n"
mismatch_notice += "in dbs\n"+json.dumps(dbs_presence, indent=2) +"\n"
mismatch_notice += "invalide in dbs\n"+json.dumps(dbs_invalid, indent=2) +"\n"
mismatch_notice += "in phedex\n"+json.dumps(phedex_presence, indent=2) +"\n"
wfi.sendLog('checkor',mismatch_notice)
if not 'recovering' in assistance_tags:
assistance_tags.add('filemismatch')
#print this for show and tell if no recovery on-going
for out in dbs_presence:
_,_,missing_phedex,missing_dbs = getDatasetFiles(url, out)
if missing_phedex:
wfi.sendLog('checkor',"These %d files are missing in phedex\n%s"%(len(missing_phedex),
"\n".join( missing_phedex )))
were_invalidated = sorted(set(missing_phedex) & set(TMDB_invalid ))
if were_invalidated:
wfi.sendLog('checkor',"These %d files were invalidated globally\n%s"%(len(were_invalidated),
"\n".join(were_invalidated)))
sendLog('checkor',"These %d files were invalidated globally\n%s\nand are invalidated in dbs"%(len(were_invalidated),
"\n".join(were_invalidated)), level='critical')
dbs3Client.setFileStatus( were_invalidated, newstatus=0 )
if missing_dbs:
wfi.sendLog('checkor',"These %d files are missing in dbs\n%s"%(len(missing_dbs),
"\n".join( missing_dbs )))
were_invalidated = sorted(set(missing_dbs) & set(TMDB_invalid ))
if were_invalidated:
wfi.sendLog('checkor',"These %d files were invalidated globally\n%s"%(len(were_invalidated),
"\n".join(were_invalidated)))
#if not bypass_checks:
## I don't think we can by pass this
is_closing = False
time_point("checked file count", sub_lap=True)
fraction_invalid = 0.20
if not all([(dbs_invalid[out] <= int(fraction_invalid*dbs_presence[out])) for out in wfi.request['OutputDatasets']]) and not options.ignoreinvalid:
print wfo.name,"has a dbs invalid file level too high"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
## need to be going and taking an eye
assistance_tags.add('invalidfiles')
if not bypass_checks:
#sub_assistance+="-invalidfiles"
is_closing = False
## put that heavy part at the end
## duplication check
duplications = {}
files_per_rl = {}
for output in wfi.request['OutputDatasets']:
duplications[output] = "skiped"
files_per_rl[output] = "skiped"
time_point("checked invalidation", sub_lap=True)
if (is_closing or bypass_checks) and (not options.ignoreduplicates):
print "starting duplicate checker for",wfo.name
for output in wfi.request['OutputDatasets']:
print "\tchecking",output
duplications[output] = True
try:
duplications[output],files_per_rl[output] = dbs3Client.duplicateRunLumiFiles( output , skipInvalid=True, verbose=True)
except:
try:
duplications[output],files_per_rl[output] = dbs3Client.duplicateRunLumiFiles( output , skipInvalid=True, verbose=True)
except Exception as e:
wfi.sendLog('checkor','Not possible to check on duplicate lumi count on %s'%(output))
sendLog('checkor','Not possible to check on duplicate lumi count on %s\n%s'%(output,str(e)),level='critical')
is_closing=False
if is_closing and any(duplications.values()) and not options.ignoreduplicates:
duplicate_notice = ""
duplicate_notice += "%s has duplicates\n"%wfo.name
duplicate_notice += json.dumps( duplications,indent=2)
duplicate_notice += '\n'
duplicate_notice += json.dumps( files_per_rl, indent=2)
wfi.sendLog('checkor',duplicate_notice)
## hook for making file invalidation ?
## it shouldn't be allowed to bypass it
assistance_tags.add('duplicates')
is_closing = False
time_point("checked duplicates", sub_lap=True)
time_point("done with %s"%wfo.name)
## for visualization later on
if not wfo.name in fDB.record:
#print "adding",wfo.name,"to close out record"
fDB.record[wfo.name] = {
'datasets' :{},
'name' : wfo.name,
'closeOutWorkflow' : None,
}
fDB.record[wfo.name]['closeOutWorkflow'] = is_closing
fDB.record[wfo.name]['priority'] = wfi.request['RequestPriority']
fDB.record[wfo.name]['prepid'] = wfi.request['PrepID']
for output in wfi.request['OutputDatasets']:
if not output in fDB.record[wfo.name]['datasets']: fDB.record[wfo.name]['datasets'][output] = {}
rec = fDB.record[wfo.name]['datasets'][output]
#rec['percentage'] = float('%.2f'%(percent_completions[output]*100))
rec['percentage'] = math.floor(percent_completions[output]*10000)/100.## round down
rec['duplicate'] = duplications[output] if output in duplications else 'N/A'
rec['phedexReqs'] = float('%.2f'%any_presence[output][custodial_presences[output][0]][1]) if len(custodial_presences[output])!=0 else 'N/A'
rec['closeOutDataset'] = is_closing
rec['transPerc'] = float('%.2f'%any_presence[output][ disk_copies[output][0]][1]) if len(disk_copies[output])!=0 else 'N/A'
rec['correctLumis'] = int(events_per_lumi[output]) if (events_per_lumi[output] > lumi_upper_limit[output]) else True
rec['missingSubs'] = False if len(custodial_locations[output])==0 else ','.join(list(set(custodial_locations[output])))
rec['dbsFiles'] = dbs_presence[output]
rec['dbsInvFiles'] = dbs_invalid[output]
rec['phedexFiles'] = phedex_presence[output]
rec['acdc'] = "%d / %d"%(len(acdc),len(acdc+acdc_inactive))
rec['familly'] = true_familly
now = time.gmtime()
rec['timestamp'] = time.mktime(now)
rec['updated'] = time.asctime(now)+' (GMT)'
## make the lumi summary
if wfi.request['RequestType'] == 'ReReco':
try:
os.system('python Unified/lumi_summary.py %s 1 > /dev/null'%(wfi.request['PrepID']))
os.system('python Unified/lumi_plot.py %s > /dev/null'%(wfi.request['PrepID']))
wfi.sendLog('checkor','Lumi summary available at %s/datalumi/lumi.%s.html'%(unified_url,wfi.request['PrepID']))
except Exception as e:
print str(e)
## make the error report
## and move on
if is_closing:
## toggle status to closed-out in request manager
wfi.sendLog('checkor',"setting %s closed-out"% wfo.name)
if not options.test:
if wfo.wm_status in ['closed-out','announced','normal-archived']:
print wfo.name,"is already",wfo.wm_status,"not trying to closed-out and assuming it does"
res = None
else:
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
print "close out answer",res
if not res in ["None",None]:
print "try to get the current status again"
wfi_bis = workflowInfo(url, wfo.name)
if wfi_bis.request['RequestStatus'] == 'closed-out':
print "the request did toggle to closed-out"
res = None
if not res in ["None",None]:
print "retrying to closing out"
print res
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
if res in [None,"None"]:
wfo.status = 'close'
session.commit()
if use_mcm and force_by_mcm:
## shoot large on all prepids, on closing the wf
for pid in pids:
mcm.delete('/restapi/requests/forcecomplete/%s'%pid)
else:
print "could not close out",wfo.name,"will try again next time"
else:
if not 'custodial' in assistance_tags or wfi.isRelval():
## do only the report for those
for member in acdc+acdc_inactive+[wfo.name]:
try:
parse_one(url, member)
except:
print "Could not make error report for",member
## full known list
#recovering # has active ACDC
##OUT #recovered #had inactive ACDC
#recovery #not over the pass bar
#over100 # over 100%
#biglumi # has a big lumiblock
#parentcustodial # the parent does not have a valid subscription yet
#custodial # has had the transfer made, is waiting for a valid custodial subscription to appear
#filemismatch # there is a dbs/phedex mismatch
#duplicates #a lumi section is there twice
## manual is not added yet, and should be so by recoveror
print wfo.name,"was tagged with :",list(assistance_tags)
if 'recovering' in assistance_tags:
## if active ACDC, being under threshold, filemismatch do not matter
assistance_tags = assistance_tags - set(['recovery','filemismatch'])
if 'recovery' in assistance_tags and 'recovered' in assistance_tags:
## should not set -recovery to anything that had ACDC already
assistance_tags = assistance_tags - set(['recovery','recovered'])
## straight to manual
assistance_tags.add('manual')
in_manual += 1
if 'recovery' in assistance_tags and 'manual' in assistance_tags:
## this is likely because something bad is happening, so leave it to manual
assistance_tags = assistance_tags - set(['recovery'])
assistance_tags.add('manual')
in_manual += 1
## that means there is something that needs to be done acdc, lumi invalidation, custodial, name it
print wfo.name,"needs assistance with",",".join( assistance_tags )
print wfo.name,"existing conditions",",".join( existing_assistance_tags )
#########################################
##### notification to requester #########
go_notify=False
if assistance_tags and not 'manual' in existing_assistance_tags and existing_assistance_tags != assistance_tags:
go_notify=True
if go_notify:
#if wfo.name in already_notified:
# print "double notification"
# sendEmail('double notification','please take a look at %s'%(wfo.name))
#else:
# already_notified.append( wfo.name )
###detailslink = 'https://cmsweb.cern.ch/reqmgr/view/details/%s'
#detailslink = 'https://cmsweb.cern.ch/reqmgr2/fetch?rid=%s'%(wfo.name)
###perflink = 'https://cmsweb.cern.ch/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s'%(wfo.name)
perflink = '%s/report/%s'%(unified_url,wfo.name)
splitlink = 'https://cmsweb.cern.ch/reqmgr/view/splitting/%s'%(wfo.name)
## notify templates
messages= {
'recovery': 'Samples completed with missing statistics:\n%s\n%s '%( '\n'.join(['%.2f %% complete for %s'%(percent_completions[output]*100, output) for output in wfi.request['OutputDatasets'] ] ), perflink ),
'biglumi': 'Samples completed with large luminosity blocks:\n%s\n%s '%('\n'.join(['%d > %d for %s'%(events_per_lumi[output], lumi_upper_limit[output], output) for output in wfi.request['OutputDatasets'] if (events_per_lumi[output] > lumi_upper_limit[output])]), splitlink),
'duplicates': 'Samples completed with duplicated luminosity blocks:\n%s\n'%( '\n'.join(['%s'%output for output in wfi.request['OutputDatasets'] if output in duplications and duplications[output] ] ) ),
'filemismatch': 'Samples completed with inconsistency in DBS/Phedex',
#'manual' : 'Workflow completed and requires manual checks by Ops',
}
content = "The request PREPID (WORKFLOW) is facing issue in production.\n"
motive = False
for case in messages:
if case in assistance_tags:
content+= "\n"+messages[case]+"\n"
motive = True
content += "You are invited to check, while this is being taken care of by Comp-Ops.\n"
content += "This is an automated message from Comp-Ops.\n"
items_notified = set()
if use_mcm and motive:
wfi.notifyRequestor( content , mcm = mcm)
#########################################
## logic to set the status further
if assistance_tags:
new_status = 'assistance-'+'-'.join(sorted(assistance_tags) )
else:
new_status = 'assistance'
## case where the workflow was in manual from recoveror
if not 'manual' in wfo.status or new_status!='assistance-recovery':
wfo.status = new_status
if not options.test:
wfi.sendLog('checkor','setting %s to %s'%(wfo.name, wfo.status))
session.commit()
else:
print "current status is",wfo.status,"not changing to anything"
#open('already_notifified.json','w').write( json.dumps( already_notified , indent=2))
fDB.html()
if not spec and in_manual!=0:
sendEmail("fresh assistance status available","Fresh status are available at %s/assistance.html"%unified_url,destination=['*****@*****.**'])
#it's a bit annoying
pass
## custodial requests
print "Custodials"
print json.dumps(custodials, indent=2)
for site in custodials:
items_at = defaultdict(set)
for i in custodials[site]:
item, group = i.split('@') if '@' in i else (i,'DataOps')
items_at[group].add( item )
for group,items in items_at.items():
print ','.join(items),'=>',site,'@',group
if not options.test:
result = makeReplicaRequest(url, site, sorted(items) ,"custodial copy at production close-out",custodial='y',priority='low', approve = (site in SI.sites_auto_approve) , group=group)
print result
print "File Invalidation"
print invalidations
def equalizor(url , specific = None, options=None):
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
if not specific:
workflows = getWorkflows(url, status='running-closed', details=True)
workflows.extend(getWorkflows(url, status='running-open', details=True))
## start from scratch
modifications = defaultdict(dict)
## define regionality site => fallback allowed. feed on an ssb metric ??
mapping = defaultdict(list)
reversed_mapping = defaultdict(list)
regions = defaultdict(list)
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
for site in SI.sites_ready:
region = site.split('_')[1]
if not region in ['US','DE','IT']: continue
regions[region] = [region]
def site_in_depletion(s):
return True
if s in SI.sites_pressure:
(m, r, pressure) = SI.sites_pressure[s]
if float(m) < float(r):
print s,m,r,"lacking pressure"
return True
else:
print s,m,r,"pressure"
pass
return False
for site in SI.sites_ready:
region = site.split('_')[1]
## fallback to the region, to site with on-going low pressure
mapping[site] = [fb for fb in SI.sites_ready if any([('_%s_'%(reg) in fb and fb!=site and site_in_depletion(fb))for reg in regions[region]]) ]
use_T0 = ('T0_CH_CERN' in UC.get("site_for_overflow"))
if options.t0: use_T0 = True
#if options.augment : use_T0 = True
use_HLT = ('T2_CH_CERN_HLT' in UC.get("site_for_overflow"))
if options.hlt: use_HLT = True
#if options.augment : use_HLT=True
if use_HLT:
mapping['T2_CH_CERN'].append('T2_CH_CERN_HLT')
if use_T0:
mapping['T2_CH_CERN'].append('T0_CH_CERN')
#mapping['T1_FR_CCIN2P3'].append('T0_CH_CERN')
#mapping['T2_IT_Legnaro'].append('T1_IT_CNAF')
for reg in ['IT','DE','UK']:
mapping['T2_CH_CERN'].extend([fb for fb in SI.sites_ready if '_%s_'%reg in fb])
## make them appear as OK to use
force_sites = []
## overflow CERN to underutilized T1s
upcoming = json.loads( open('%s/GQ.json'%monitor_dir).read())
for possible in SI.sites_T1s:
if not possible in upcoming:
mapping['T2_CH_CERN'].append(possible)
## remove add-hoc sites from overflow mapping
prevent_sites = ['T2_US_Purdue']
for prevent in prevent_sites:
if prevent in mapping: mapping.pop( prevent )
for src in mapping:
for prevent in prevent_sites:
if prevent in mapping[src]:
mapping[src].remove( prevent )
## create the reverse mapping for the condor module
for site,fallbacks in mapping.items():
for fb in fallbacks:
reversed_mapping[fb].append(site)
## this is the fallback mapping
print "Direct mapping : site => overflow"
print json.dumps( mapping, indent=2)
print "Reverse mapping : dest <= from origin"
print json.dumps( reversed_mapping, indent=2)
altered_tasks = set()
def running_idle( wfi , task_name):
gmon = wfi.getGlideMon()
#print gmon
if not gmon: return (0,0)
if not task_name in gmon: return (0,0)
return (gmon[task_name]['Running'], gmon[task_name]['Idle'])
def needs_action( wfi, task, min_idled = 100, pressure = 0.2):
task_name = task.pathName.split('/')[-1]
running, idled = running_idle( wfi, task_name)
go = True
if not idled and not running :
go = False
if idled < 100:
go = False
if (not running and idled) or (running and (idled / float(running) > pressure)):
go = True
else:
go = False
return go, task_name, running, idled
def getPerf( task ):
task = task.split('/')[1]+'/'+task.split('/')[-1]
try:
u = 'http://cms-gwmsmon.cern.ch/prodview/json/history/memoryusage720/%s'%task
print u
perf_data = json.loads(os.popen('curl -s --retry 5 %s'%u).read())
except Exception as e:
print str(e)
return (None,None)
buckets = perf_data['aggregations']["2"]['buckets']
s_m = sum( bucket['key']*bucket['doc_count'] for bucket in buckets)
w_m = sum( bucket['doc_count'] for bucket in buckets)
m_m = max( bucket['key'] for bucket in buckets) if buckets else None
b_m = None
if w_m > 100:
b_m = m_m
try:
perf_data = json.loads(os.popen('curl -s --retry 5 http://cms-gwmsmon.cern.ch/prodview/json/history/runtime720/%s'%task).read())
except Exception as e:
print str(e)
return (b_m,None)
buckets = perf_data['aggregations']["2"]['buckets']
s_t = sum( bucket['key']*bucket['doc_count'] for bucket in buckets)
w_t = sum( bucket['doc_count'] for bucket in buckets)
m_t = max( bucket['key'] for bucket in buckets) if buckets else None
b_t = None
if w_t > 100:
b_t = m_t
return (b_m,b_t)
def getcampaign( task ):
taskname = task.pathName.split('/')[-1]
if hasattr( task, 'prepID'):
return task.prepID.split('-')[1]
elif taskname.count('-')>=1:
return taskname.split('-')[1]
else:
return None
def close( interface ):
open('%s/equalizor.json.new'%monitor_dir,'w').write( json.dumps( interface, indent=2))
os.system('mv %s/equalizor.json.new %s/equalizor.json'%(monitor_dir,monitor_dir))
os.system('cp %s/equalizor.json %s/logs/equalizor/equalizor.%s.json'%(monitor_dir,monitor_dir,time.mktime(time.gmtime())))
interface = {
'reversed_mapping' : reversed_mapping,
'modifications' : {}
}
if options.augment or options.remove:
interface['modifications'] = json.loads( open('%s/equalizor.json'%monitor_dir).read())['modifications']
if options.remove:
if specific in interface['modifications']:
print "poping",specific
interface['modifications'].pop(specific)
close( interface )
return
PU_locations = {}
PU_overflow = {}
LHE_overflow = {}
tune_performance = []
pending_HLT = 0
max_HLT = 60000
pending_T0 = 0
max_T0 = 60000
try:
gmon = json.loads(os.popen('curl -s http://cms-gwmsmon.cern.ch/prodview/json/T2_CH_CERN_HLT').read())
pending_HLT += gmon["Running"]
pending_HLT += gmon["MatchingIdle"]
except:
pass
stay_within_site_whitelist = False
specific_task=None
if specific and ":" in specific:
specific,specific_task = specific.split(':')
if specific:
wfs = session.query(Workflow).filter(Workflow.name.contains(specific)).all()
else:
wfs = session.query(Workflow).filter(Workflow.status == 'away').all()
performance = {}
no_routing = [
]
random.shuffle( wfs )
for wfo in wfs:
if wfo.name in no_routing and not options.augment:
continue
if specific and not specific in wfo.name:
continue
if specific:
wfi = workflowInfo(url, wfo.name)
else:
cached = filter(lambda d : d['RequestName']==wfo.name, workflows)
if not cached : continue
wfi = workflowInfo(url, wfo.name, request = cached[0])
## only running should get re-routed
if not wfi.request['RequestStatus'] in ['running-open','running-closed'] and not specific: continue
tasks_and_campaigns = []
for task in wfi.getWorkTasks():
tasks_and_campaigns.append( (task, getcampaign(task) ) )
_,_,_,sec = wfi.getIO()
## check needs override
needs_overide = False
if not needs_overide and options.augment: needs_overide=True
def overide_from_agent( wfi, needs_overide):
bad_agents = []#'http://cmssrv219.fnal.gov:5984']
if not bad_agents: return needs_overide
if needs_overide: return True
agents = wfi.getAgents()
wqss = ['Running','Acquired']
if any([agent in agents.get(wqs,{}).keys() for wqs,agent in itertools.product( wqss, bad_agents)]):
print "overriding the need for bad agent"
needs_overide = True
return needs_overide
## now parse this for action
for i_task,(task,campaign) in enumerate(tasks_and_campaigns):
if options.augment:
print task.pathName
print campaign
tune = CI.get(campaign,'tune',options.tune)
if tune and not campaign in tune_performance:
tune_performance.append( campaign )
overflow = CI.get(campaign,'overflow',{})
if overflow:
if "PU" in overflow and not campaign in PU_overflow:
PU_overflow[campaign] = copy.deepcopy(overflow['PU'])
print "adding",campaign,"to PU overflow rules"
if "LHE" in overflow and not campaign in LHE_overflow:
print "adding",campaign,"to light input overflow rules"
site_list = overflow['LHE']['site_list']
LHE_overflow[campaign] = copy.deepcopy( getattr(SI,site_list) )
### get the task performance, for further massaging.
if campaign in tune_performance or options.tune:
print "performance",task.taskType,task.pathName
if task.taskType in ['Processing','Production']:
set_memory,set_time = getPerf( task.pathName )
#print "Performance %s GB %s min"%( set_memory,set_time)
wfi.sendLog('equalizor','Performance tuning to %s GB %s min'%( set_memory,set_time))
## get values from gmwsmon
# massage the values : 95% percentile
performance[task.pathName] = {}
if set_memory:
performance[task.pathName]['memory']=set_memory
if set_time and False:
performance[task.pathName]['time'] = set_time
### rule to avoid the issue of taskchain secondary jobs being stuck at sites processing the initial step
if campaign in LHE_overflow:
if task.taskType in ['Processing']:
needs, task_name, running, idled = needs_action(wfi, task)
needs_overide = overide_from_agent( wfi, needs_overide)
extend_to = list(set(copy.deepcopy( LHE_overflow[campaign] )))
if stay_within_site_whitelist:
extend_to = list(set(extend_to) & set(wfi.request['SiteWhitelist'])) ## restrict to stupid-site-whitelist
extend_to = list(set(extend_to) & set(SI.sites_ready + force_sites))
if extend_to and needs or needs_overide:
modifications[wfo.name][task.pathName] = { "ReplaceSiteWhitelist" : extend_to ,"Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : ReplaceSiteWhitelist \n %s'%( task_name,
wfo.name,
running,
idled ,
json.dumps( sorted(modifications[wfo.name][task.pathName]['ReplaceSiteWhitelist']))))
altered_tasks.add( task.pathName )
else:
wfi.sendLog('equalizor','%s of %s is running %d and pending %d'%( task_name, wfo.name, running, idled))
### overflow the 76 digi-reco to the site holding the pileup
if campaign in PU_overflow:
force = PU_overflow[campaign]['force'] if 'force' in PU_overflow[campaign] else False
secondary_locations = set(SI.sites_ready + force_sites)
for s in sec:
if not s in PU_locations:
presence = getDatasetPresence( url, s)
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
PU_locations[s] = one_secondary_locations
print "secondary is at",sorted(PU_locations[s])
secondary_locations = set([SI.SE_to_CE(site) for site in PU_locations[s]]) & secondary_locations
## we should add all sites that hold the secondary input if any
### given that we have the secondary location available, it is not necessary to use the add-hoc list
##secondary_locations = list(set(PU_overflow[campaign]['sites']) & set( SI.sites_ready ))
if any([task.pathName.endswith(finish) for finish in ['_0','StepOneProc','Production']]) :
needs, task_name, running, idled = needs_action(wfi, task)
## removing the ones in the site whitelist already since they encode the primary input location
if stay_within_site_whitelist:
original_site_in_use = set(wfi.request['SiteWhitelist'])
else:
original_site_in_use = set(secondary_locations)
## remove the sites that have already running jobs
gmon = wfi.getGlideMon()
if gmon and task_name in gmon and 'Sites' in gmon[task_name]:
site_in_use = set(gmon[task_name]['Sites'])
print "removing",sorted(site_in_use)
## that determines where you want to run in addition
augment_by = list((set(secondary_locations)- site_in_use) & original_site_in_use)
else:
print "no existing running site"
augment_by = list(original_site_in_use)
needs_overide = overide_from_agent( wfi, needs_overide)
if augment_by and (needs or needs_overide or force) and PU_overflow[campaign]['pending'] < PU_overflow[campaign]['max']:
PU_overflow[campaign]['pending'] += idled
print "raising overflow to",PU_overflow[campaign]['pending'],"for",PU_overflow[campaign]['max']
## the step with an input ought to be the digi part : make this one go anywhere
modifications[wfo.name][task.pathName] = { "AddWhitelist" : augment_by , "Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
altered_tasks.add( task.pathName )
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : AddWhitelist \n %s'%( task_name, wfo.name,
running, idled,
json.dumps( sorted(augment_by), indent=2 )))
else:
print task_name,"of",wfo.name,"running",running,"and pending",idled
### overflow the skims back to multi-core
if campaign in ['Run2015D','Run2015C_25ns'] and task.taskType =='Skim':
original_swl = wfi.request['SiteWhitelist']
needs, task_name, running, idled = needs_action(wfi, task)
if (needs or needs_overide):
modifications[wfo.name][task.pathName] = { 'AddWhitelist' : original_swl,
"Running" : running, "Pending" : idled, "Priority" : wfi.request['RequestPriority']}
altered_tasks.add( task.pathName )
wfi.sendLog('equalizor','%s of %s is running %d and pending %d, taking action : AddWhitelist \n %s'%( task_name, wfo.name,
running, idled,
json.dumps( sorted(original_swl), indent=2 )))
if options.augment:
print sorted(wfi.request['SiteWhitelist']),i_task,use_HLT
### add the HLT at partner of CERN
if 'T2_CH_CERN' in wfi.request['SiteWhitelist'] and i_task in [0,1] and use_HLT:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs=True
needs = True
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide) and pending_HLT < max_HLT:
pending_HLT += idled
if task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
modifications[wfo.name][task.pathName]["AddWhitelist"].append( "T2_CH_CERN_HLT" )
print "\t",wfo.name,"adding addHLT up to",pending_HLT,"for",max_HLT
print task.pathName
## this Replace does not work at all for HLT
#elif task.pathName in modifications[wfo.name] and 'ReplaceSiteWhitelist' in modifications[wfo.name][task.pathName]:
#modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"].append( "T2_CH_CERN_HLT" )
#print "\t",wfo.name,"adding replace HLT up to",pending_HLT,"for",max_HLT
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : ["T2_CH_CERN_HLT"],
"Priority" : wfi.request['RequestPriority'],
"Running" : running,
"Pending" : idled}
wfi.sendLog('equalizor','adding the HLT in whitelist of %s to %d for %d'%( task.pathName, pending_HLT, max_HLT))
if i_task==0 and not sec and use_T0:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs=True
#needs = True
good_type = wfi.request['RequestType'] in ['MonteCarlo','MonteCarloFromGEN']
read_lhe = ((not 'LheInputFiles' in wfi.request) or bool(wfi.request['LheInputFiles']))
good_type &= not read_lhe
if not good_type and not options.augment: needs = False
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide):
pending_T0 += idled
if task.pathName in modifications[wfo.name] and 'AddWhitelist' in modifications[wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][task.pathName]["AddWhitelist"]:
modifications[wfo.name][task.pathName]["AddWhitelist"].append( "T0_CH_CERN" )
wfi,sendLog('equalizor','adding the T0 for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
elif task.pathName in modifications[wfo.name] and 'ReplaceSiteWhitelist' in modifications[wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"]:
modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"].append( "T0_CH_CERN" )
wfi,sendLog('equalizor','adding the T0 to replacement for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
else:
modifications[wfo.name][task.pathName] = { "AddWhitelist" : ["T0_CH_CERN"],
"Priority" : wfi.request['RequestPriority'],
"Running" : running,
"Pending" : idled}
wfi,sendLog('equalizor','adding the T0 for %s to %d for %d'%( task.pathName, pending_T0, max_T0))
interface['modifications'].update( modifications )
### manage the number of core and job resizing
interface['cores']={'T2_CH_CERN_HLT': {'min':4,'max':16}, 'default': {'min':1, 'max':4}}
interface['resizes'] = ['RunIISpring16DR80']
### manage the modification of the memory and target time
interface['time'] = defaultdict(list)
interface['memory'] = defaultdict(list)
max_N_mem = 10
max_N_time = 10
## discretize the memory to 10 at most values
mems = set([o['memory'] for t,o in performance.items() if 'memory' in o])
times = set([o['time'] for t,o in performance.items() if 'time' in o])
if len(mems)>max_N_mem:
mem_step = int((max(mems) - min(mems))/ float(max_N_mem))
for t in performance:
if not 'memory' in performance[t]: continue
(m,r) = divmod(performance[t]['memory'], mem_step)
performance[t]['memory'] = (m+1)*mem_step
if len(times)>max_N_time:
time_step = int((max(times) - min(times))/float(max_N_time))
for t in performance:
if not 'time' in performance[t]: continue
(m,r) = divmod(performance[t]['time'], time_step)
performance[t]['time'] = (m+1)*time_step
for t,o in performance.items():
if 'time' in o:
interface['time'][str(o['time'])] .append( t )
if 'memory' in o:
interface['memory'][str(o['memory'])].append( t )
## close and save
close( interface )
def recoveror(url, specific, options=None):
if userLock('recoveror'): return
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): return
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
use_recoveror = UC.get('use_recoveror')
if not use_recoveror and not options.go:
print "We are told not to run recoveror"
return
def make_int_keys(d):
for code in d:
d[int(code)] = d.pop(code)
error_codes_to_recover = UC.get('error_codes_to_recover')
error_codes_to_block = UC.get('error_codes_to_block')
error_codes_to_notify = UC.get('error_codes_to_notify')
make_int_keys(error_codes_to_recover)
make_int_keys(error_codes_to_block)
make_int_keys(error_codes_to_notify)
#wfs = session.query(Workflow).filter(Workflow.status == 'assistance-recovery').all()
wfs = session.query(Workflow).filter(
Workflow.status.contains('recovery')).all()
if specific:
wfs.extend(
session.query(Workflow).filter(
Workflow.status == 'assistance-manual').all())
for wfo in wfs:
if specific and not specific in wfo.name: continue
if not specific and 'manual' in wfo.status: continue
wfi = workflowInfo(url, wfo.name)
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
all_errors = {}
try:
## this is clearly very truncated and should be changed completely
wfi.getSummary()
all_errors = wfi.summary['errors']
except:
pass
print '-' * 100
print "Looking at", wfo.name, "for recovery options"
recover = True
if not 'MergedLFNBase' in wfi.request:
print "f****d up"
sendEmail('missing lfn', '%s wl cache is screwed up' % wfo.name)
recover = False
if not len(all_errors):
print "\tno error for", wfo.name
recover = False
task_to_recover = defaultdict(list)
message_to_ops = ""
message_to_user = ""
if 'LheInputFilese' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
recover = False
if wfi.request['RequestType'] in ['MonteCarlo', 'ReReco']:
recover = False
if 'Campaign' in wfi.request:
c = wfi.request['Campaign']
if c in CI.campaigns and 'recover' in CI.campaigns[c]:
recover = CI.campaigns[c]['recover']
for task, errors in all_errors.items():
print "\tTask", task
## collect all error codes and #jobs regardless of step at which it occured
all_codes = []
for name, codes in errors.items():
if type(codes) == int: continue
all_codes.extend([
(int(code), info['jobs'], name,
list(set([e['type'] for e in info['errors']])),
list(set([e['details'] for e in info['errors']])))
for code, info in codes.items()
])
all_codes.sort(key=lambda i: i[1], reverse=True)
sum_failed = sum([l[1] for l in all_codes])
for errorCode, njobs, name, types, details in all_codes:
rate = 100 * njobs / float(sum_failed)
#print ("\t\t %10d (%6s%%) failures with error code %10d (%"+str(max_legend)+"s) at stage %s")%(njobs, "%4.2f"%rate, errorCode, legend, name)
print(
"\t\t %10d (%6s%%) failures with error code %10d (%30s) at stage %s"
) % (njobs, "%4.2f" % rate, errorCode, ','.join(types), name)
added_in_recover = False
#if options.go:
# force the recovery of any task with error ?
if errorCode in error_codes_to_recover:
## the error code is registered
for case in error_codes_to_recover[errorCode]:
match = case['details']
matched = (match == None)
if not matched:
matched = False
for detail in details:
if match in detail:
print "[recover] Could find keyword", match, "in"
print 50 * "#"
print detail
print 50 * "#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => we should be able to recover that", case[
'legend']
task_to_recover[task].append((code, case))
added_in_recover = True
message_to_user = ""
else:
print "\t\t recoverable but not frequent enough, needs", case[
'rate']
if errorCode in error_codes_to_block:
for case in error_codes_to_block[errorCode]:
match = case['details']
matched = (match == None)
if not matched:
matched = False
for detail in details:
if match in detail:
print "[block] Could find keyword", match, "in"
print 50 * "#"
print detail
print 50 * "#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => that error means no ACDC on that workflow", case[
'legend']
if not options.go:
message_to_ops += "%s has an error %s blocking an ACDC.\n%s\n " % (
wfo.name, errorCode, '#' * 50)
recover = False
added_in_recover = False
if errorCode in error_codes_to_notify and not added_in_recover:
print "\t\t => we should notify people on this"
message_to_user += "%s has an error %s in processing.\n%s\n" % (
wfo.name, errorCode, '#' * 50)
if message_to_user:
print wfo.name, "to be notified to user(DUMMY)", message_to_user
if message_to_ops:
#sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**'])
sendLog('recoveror', message_to_ops, level='warning')
if len(task_to_recover) != len(all_errors):
print "Should not be doing partial ACDC. skipping"
#sendEmail('recoveror','do not want to make partial acdc on %s'%wfo.name)
sendLog('recoveror',
'do not want to make partial acdc on %s' % wfo.name,
level='warning')
recover = False
if task_to_recover and recover:
print "Initiating recovery"
print ', '.join(task_to_recover.keys()), "to be recovered"
recovering = set()
for task in task_to_recover:
print "Will be making a recovery workflow for", task
## from here you can fetch known solutions, to known error codes
actions = list(
set([
case['solution']
for code, case in task_to_recover[task]
]))
acdc = singleRecovery(url,
task,
wfi.request,
actions,
do=options.do)
if not acdc:
if options.do:
if recovering:
print wfo.name, "has been partially ACDCed. Needs manual attention"
#sendEmail( "failed ACDC partial recovery","%s has had %s/%s recoveries %s only"%( wfo.name, len(recovering), len(task_to_recover), list(recovering)), destination=['*****@*****.**'])
sendLog('recoveror',
"%s has had %s/%s recoveries %s only" %
(wfo.name, len(recovering),
len(task_to_recover), list(recovering)),
level='critical')
continue
else:
print wfo.name, "failed recovery once"
#break
continue
else:
print "no action to take further"
sendLog('recoveror',
"ACDC for %s can be done automatically" %
wfo.name,
level='critical')
continue
## and assign it ?
team = wfi.request['Team']
#assign_to_sites = set(SI.sites_ready) ## that needs to be massaged to prevent assigning to something out.
assign_to_sites = set(SI.all_sites)
parameters = {
#'SiteWhitelist' : wfi.request['SiteWhitelist'],
'SiteWhitelist': sorted(assign_to_sites),
'AcquisitionEra': wfi.acquisitionEra(),
'ProcessingString': wfi.processingString(),
'MergedLFNBase': wfi.request['MergedLFNBase'],
'ProcessingVersion': wfi.request['ProcessingVersion'],
}
## hackery for ACDC merge assignment
if wfi.request[
'RequestType'] == 'TaskChain' and 'Merge' in task.split(
'/')[-1]:
parameters['AcquisitionEra'] = None
parameters['ProcessingString'] = None
## xrootd setttings on primary and secondary
if 'TrustSitelists' in wfi.request and wfi.request[
'TrustSitelists']:
parameters['TrustSitelists'] = True
if 'TrustPUSitelists' in wfi.request and wfi.request[
'TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
if options.ass:
print "really doing the assignment of the ACDC", acdc
parameters['execute'] = True
wfi.sendLog('recoveror',
"%s was assigned for recovery" % acdc)
else:
print "no assignment done with this ACDC", acdc
sendLog('recoveror',
"%s needs to be assigned" % (acdc),
level='critical')
result = reqMgrClient.assignWorkflow(url, acdc, team,
parameters)
if not result:
print acdc, "was not asigned"
sendLog('recoveror',
"%s needs to be assigned" % (acdc),
level='critical')
else:
recovering.add(acdc)
current = None
if recovering:
#if all went well, set the status to -recovering
current = wfo.status
if options.ass:
current = current.replace('recovery', 'recovering')
else:
current = 'assistance-manual'
print 'created ACDC: ' + ', '.join(recovering)
else:
## was set to be recovered, and no acdc was made
current = 'assistance-manual'
if current:
print wfo.name, "setting the status to", current
wfo.status = current
session.commit()
else:
## this workflow should be handled manually at that point
print wfo.name, "needs manual intervention"
wfo.status = 'assistance-manual'
session.commit()
def actor(url, options=None):
if moduleLock(wait=False, silent=True)(): return
if userLock('actor'): return
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
# CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
# Need to look at the actions page https://vocms0113.cern.ch:80/getaction (can add ?days=20) and perform any actions listed
try:
action_list = json.loads(
os.popen(
'curl -s -k https://vocms0113.cern.ch:80/getaction?days=15').
read())
## now we have a list of things that we can take action on
except:
try:
action_list = json.loads(
os.popen(
'curl -s -k https://vocms0113.cern.ch/getaction?days=15').
read())
except:
print "Not able to load action list :("
sendLog('actor', 'Not able to load action list', level='critical')
return
if options.actions:
action_list = json.loads(open(options.actions).read())
print json.dumps(action_list, indent=2)
if not action_list:
print "EMPTY!"
return
wf_list = action_list.keys()
print json.dumps(sorted(wf_list), indent=2)
if options.spec:
wf_list = [wf for wf in wf_list if options.spec in wf]
max_per_round = UC.get('max_per_round').get('actor', None)
if max_per_round:
random.shuffle(wf_list)
wf_list = wf_list[:max_per_round]
for wfname in wf_list:
print '-' * 100
print "Looking at", wfname, "for recovery options"
to_clone = False
to_acdc = False
to_force = False
to_hold = False
something_to_do = False
tasks = action_list[wfname].get('Parameters', None)
to_acdc = action_list[wfname].get('Action', None) == 'acdc'
to_clone = action_list[wfname].get('Action', None) == 'clone'
to_force = action_list[wfname].get(
'Action', None) == 'special' and action_list[wfname].get(
'Parameters', {}).get('action', None) in ['by-pass', 'bypass']
to_hold = action_list[wfname].get(
'Action', None) == 'special' and action_list[wfname].get(
'Parameters', {}).get('action', None) in ['onhold', 'on-hold']
if not to_acdc and not to_clone and not to_force and not to_hold:
sendLog(
'actor',
'Action submitted for something other than acdc, clone, bypass or hold for workflow %s'
% wfname,
level='critical')
print json.dumps(action_list[wfname], indent=2)
continue
if not tasks and to_acdc:
sendLog('actor',
'Empty action submitted for workflow %s' % wfname,
level='critical')
print "Moving on. Parameters is blank for " + wfname
continue
wfi = workflowInfo(url, wfname)
recover = True
message_to_ops = ""
message_to_user = ""
#===========================================================
if to_clone and options.do:
print "Let's try kill and clone: "
wfi.sendLog('actor', 'Going to clone %s' % wfname)
results = []
datasets = set(wfi.request['OutputDatasets'])
comment = ""
if 'comment' in tasks: comment = ", reason: " + tasks['comment']
wfi.sendLog(
'actor',
"invalidating the workflow by traffic controller %s" % comment)
#Reject all workflows in the family
#first reject the original workflow.
reqMgrClient.invalidateWorkflow(
url,
wfi.request['RequestName'],
current_status=wfi.request['RequestStatus'],
cascade=False)
#Then reject any ACDCs associated with that workflow
family = getWorkflowById(url, wfi.request['PrepID'], details=True)
for fwl in family:
print "rejecting", fwl['RequestName'], fwl['RequestStatus']
wfi.sendLog(
'actor', "rejecting %s, previous status %s" %
(fwl['RequestName'], fwl['RequestStatus']))
reqMgrClient.invalidateWorkflow(
url,
fwl['RequestName'],
current_status=fwl['RequestStatus'],
cascade=False)
datasets.update(fwl['OutputDatasets'])
#Invalidate all associated output datasets
for dataset in datasets:
results.append(setDatasetStatus(dataset, 'INVALID'))
if all(map(lambda result: result in ['None', None, True],
results)):
wfi.sendLog('actor', "%s and children are rejected" % wfname)
cloned = None
try:
cloned = singleClone(url, wfname, tasks, comment, options.do)
except Exception as e:
sendLog(
'actor',
'Failed to create clone for %s! Check logs for more information. Action will need to be resubmitted.'
% wfname,
level='critical')
wfi.sendLog('actor', 'Failed to create clone for %s!' % wfname)
print str(e)
##let's not remove the action other the workflow goes to "trouble" and the WTC cannot set the action again
#remove_action(wfname)
if not cloned:
recover = False
wfi.sendLog('actor', 'Failed to create clone for %s!' % wfname)
sendLog('actor',
'Failed to create clone for %s!' % wfname,
level='critical')
else:
wfi.sendLog('actor', "Workflow %s cloned" % wfname)
#===========================================================
elif to_force:
wfi.sendLog('actor',
'Bypassing from workflow traffic controler request')
forcing = json.loads(
open(
'/afs/cern.ch/user/v/vlimant/public/ops/forcecomplete.json'
).read())
forcing.append(wfname)
open('/afs/cern.ch/user/v/vlimant/public/ops/forcecomplete.json',
'w').write(json.dumps(sorted(set(forcing))))
elif to_hold:
wfi.sendLog('actor',
'Holding on workflow traffic controler request')
holding = json.loads(
open('/afs/cern.ch/user/v/vlimant/public/ops/onhold.json').
read())
holding.append(wfname)
open('/afs/cern.ch/user/v/vlimant/public/ops/onhold.json',
'w').write(json.dumps(sorted(set(holding))))
#===========================================================
elif to_acdc:
if 'AllSteps' in tasks:
allTasksDefaults = tasks['AllSteps']
tasks.pop('AllSteps')
for setting in allTasksDefaults:
for task in tasks:
if setting in tasks[task]:
tasks[task][setting] = allTasksDefaults[setting]
else:
tasks[task].append(
{setting: allTasksDefaults[setting]})
print "Tasks is "
print json.dumps(tasks, indent=2)
all_tasks = wfi.getAllTasks()
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
try:
WMErr = wfi.getWMErrors()
# print WMErr
except:
sendLog(
'actor',
'Cannot create ACDCS for %s because WMErr cannot be reached.'
% wfname,
level='critical')
continue
if not WMErr:
wfi.sendLog('actor', 'WMErrors is blank for %s.' % wfname)
print "FYI getWMErrors is blank. Presumably there are only unreported errors"
# continue
try:
where_to_run, missing_to_run, missing_to_run_at = wfi.getRecoveryInfo(
)
print "Where to run = "
print where_to_run
except:
sendLog(
'actor',
'Cannot create ACDCS for %s because recovery info cannot be found.'
% wfname,
level='critical')
print "Moving on. Cannot access recovery info for " + wfname
continue
if not where_to_run:
sendLog(
'actor',
'Cannot create ACDCS for %s because site list cannot be found.'
% wfname,
level='critical')
print "Moving on. where to run is blank"
continue
message_to_ops = ""
message_to_user = ""
num_tasks_to_recover = 0
if WMErr:
for task in WMErr:
if 'LogCollect' in task: continue
if 'Cleanup' in task: continue
if not 'jobfailed' in WMErr[task]:
continue
else:
num_tasks_to_recover += 1
# print "Task to recover: " + task
if not num_tasks_to_recover:
print "\tno error for", wfname
# recover = False
if 'LheInputFiles' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
sendLog(
'actor',
'Cannot create ACDCS for %s because it is a pLHE workflow.'
% wfname,
level='critical')
print "We don't try to recover pLHE. Moving on."
recover = False
# sendEmail('cannot submit action', '%s is a pLHE workflow. We do not try to recover pLHE'%wfname)
# if wfi.request['RequestType'] in ['ReReco']:
# recover= False
# print 'cannot submit action. ReReco'
# sendEmail('cannot submit action', '%s is request type ReReco'%wfname)
recovering = set()
for task in tasks:
assign_to_sites = set()
print "Task names is " + task
fulltaskname = '/' + wfname + '/' + task
# print "Full task name is " + fulltaskname
wrong_task = False
for task_info in all_tasks:
if fulltaskname == task_info.pathName:
if task_info.taskType not in [
'Processing', 'Production', 'Merge'
]:
wrong_task = True
wfi.sendLog(
'actor',
"Skipping task %s because the taskType is %s. Can only ACDC Processing, Production, or Merge tasks"
% (fulltaskname, task_info.taskType))
if wrong_task:
continue
print tasks[task]
actions = tasks[task]
for action in actions:
if action.startswith('sites'):
if type(actions[action]) != list:
assign_to_sites = [SI.SE_to_CE(actions[action])]
else:
assign_to_sites = list(
set([
SI.SE_to_CE(site)
for site in actions[action]
]))
# if action.startswith('mem') and actions[action] != "" and actions[action] != 'Same' and wfi.request['RequestType'] in ['TaskChain']:
# recover = False;
# print "Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname
# wfi.sendLog('actor',"Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname)
if not 'sites' in actions:
assign_to_sites = list(
set([SI.SE_to_CE(site)
for site in where_to_run[task]]))
print "Found", sorted(
assign_to_sites
), "as sites where to run the ACDC at, from the acdc doc of ", wfname
print "Going to run at", sorted(assign_to_sites)
if recover:
print "Initiating recovery"
acdc = singleRecovery(url,
fulltaskname,
wfi.request,
actions,
do=options.do)
if not acdc:
if options.do:
if recovering:
print wfname + " has been partially ACDC'ed. Needs manual attention."
sendLog(
'actor',
"%s has had %s/%s recoveries %s only" %
(wfname, len(recovering),
num_tasks_to_recover, list(recovering)),
level='critical')
wfi.sendLog(
'actor',
"%s has had %s/%s recoveries %s only" %
(wfname, len(recovering),
num_tasks_to_recover, list(recovering)))
break
else:
print wfname + " failed recovery once"
recover = False
break
else:
print "no action to take further"
# sendLog('recoveror', "ACDC for %s can be done automatically"% wfname, level='critical')
continue
else: #ACDC was made correctly. Now we have to assign it.
wfi.sendLog(
'actor',
'ACDC created for task %s. Actions taken \n%s' %
(fulltaskname, json.dumps(actions)))
#team = wfi.request['Teams'][0]
team = 'production'
parameters = {
'SiteWhitelist': sorted(assign_to_sites),
'AcquisitionEra': wfi.acquisitionEra(),
'ProcessingString': wfi.processingString(),
'MergedLFNBase': wfi.request['MergedLFNBase'],
'ProcessingVersion':
wfi.request['ProcessingVersion'],
}
## hackery for ACDC merge assignment
if wfi.request[
'RequestType'] == 'TaskChain' and 'Merge' in task.split(
'/')[-1]:
parameters['AcquisitionEra'] = None
parameters['ProcessingString'] = None
## xrootd setttings on primary and secondary
if 'xrootd' in actions:
if actions['xrootd'] == 'enabled':
print "Going to assign via xrootd"
parameters['TrustSitelists'] = True
elif actions['xrootd'] == 'disabled':
parameters['TrustSitelists'] = False
elif ('TrustSitelists' in wfi.request
and wfi.request['TrustSitelists'] == 'true'):
parameters['TrustSitelists'] = True
else:
parameters['TrustSitelists'] = False
if 'secondary' in actions:
if actions['secondary'] == 'enabled':
print 'Enabling reading the secondary input via xrootd'
parameters['TrustPUSitelists'] = True
elif actions['secondary'] == 'disabled':
parameters['TrustPUSitelists'] = False
#in case secondary is blank or not set to enabled or disabled
elif 'TrustPUSitelists' in wfi.request and wfi.request[
'TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
elif 'TrustPUSitelists' in wfi.request and wfi.request[
'TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
if options.ass:
print "really doing the assignment of the ACDC", acdc
parameters['execute'] = True
wfi.sendLog('actor',
"%s was assigned for recovery" % acdc)
else:
print "no assignment done with this ACDC", acdc
sendLog('actor',
"%s needs to be assigned" % (acdc),
level='critical')
continue
# print parameters
result = reqMgrClient.assignWorkflow(
url, acdc, team, parameters)
if not result:
print acdc, "was not assigned"
sendLog('actor',
"%s needs to be assigned" % (acdc),
level='critical')
else:
recovering.add(acdc)
wfi.sendLog('actor', "ACDCs created for %s" % wfname)
#===========================================================
if recover and options.do:
remove_action(wfname)
if message_to_user:
print wfname, "to be notified to user(DUMMY)", message_to_user
if message_to_ops:
print 'message'
#sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**'])
# sendLog('recoveror',message_to_ops,level='warning')
return
use_mcm = True
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): sys.exit(0)
use_mcm = up.status['mcm']
mcm = None
if use_mcm:
print "mcm interface is up"
mcm = McMClient(dev=False)
statuses = [
'assignment-approved', 'assigned', 'failed', 'acquired', 'running-open',
'running-closed', 'force-complete', 'completed', 'closed-out'
]
UC = unifiedConfiguration()
CI = campaignInfo()
tier_no_custodial = UC.get('tiers_with_no_custodial')
tiers_keep_on_disk = UC.get("tiers_keep_on_disk")
now = time.mktime(time.gmtime())
## can we catch the datasets that actually should go to tape ?
custodial_override = {}
for c in CI.campaigns:
if 'custodial_override' in CI.campaigns[c]:
custodial_override[c] = CI.campaigns[c]['custodial_override']
newly_locking = set()
also_locking_from_reqmgr = set()
def closor(url, specific=None, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
all_late_files = []
check_fullcopy_to_announce = UC.get('check_fullcopy_to_announce')
jump_the_line = options.announce if options else False
if jump_the_line:
print "announce option is on. Checking on things on-going ready to be announced"
wfs = session.query(Workflow).filter(
Workflow.status.contains('announce')).filter(
sqlalchemy.not_(Workflow.status.contains('announced'))).all()
else:
print "regular option. Checking on things done and to be announced"
wfs = session.query(Workflow).filter(Workflow.status == 'close').all()
wfs_n = [w.name for w in wfs]
print "unique names?"
print len(set(wfs_n)) == len(wfs_n)
held = set()
print len(wfs), "closing"
random.shuffle(wfs)
max_per_round = UC.get('max_per_round').get('closor', None)
if options.limit: max_per_round = options.limit
if max_per_round:
## order them by priority
all_closedout = sorted(getWorkflows(url, 'closed-out', details=True),
key=lambda r: r['RequestPriority'])
all_closedout = [r['RequestName'] for r in all_closedout]
def rank(wfn):
return all_closedout.index(wfn) if wfn in all_closedout else 0
wfs = sorted(wfs, key=lambda wfo: rank(wfo.name), reverse=True)
wfs = wfs[:max_per_round]
batch_go = {}
batch_warnings = defaultdict(set)
batch_goodness = UC.get("batch_goodness")
for iwfo, wfo in enumerate(wfs):
if specific and not specific in wfo.name: continue
print "Progress [%d/%d]" % (iwfo, len(wfs))
## what is the expected #lumis
wfi = workflowInfo(url, wfo.name)
wfo.wm_status = wfi.request['RequestStatus']
if wfi.isRelval():
has_batch_go = False
batch_name = wfi.getCampaign()
if not batch_name in batch_go:
## do the esimatation whethere this can be announced : only once per batch
in_batches = getWorkflowByCampaign(url,
batch_name,
details=True)
batch_go[batch_name] = all(
map(
lambda s: not s in [
'completed', 'running-open', 'running-closed',
'acquired', 'assigned', 'assignment-approved'
], [r['RequestStatus'] for r in in_batches]))
## already verified
has_batch_go = batch_go[batch_name]
if not has_batch_go:
wfi.sendLog(
'closor',
'Cannot close for now because the batch <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?campaign=%s>%s</a> is not all close'
% (batch_name, batch_name))
continue
if wfi.request['RequestStatus'] in ['announced', 'normal-archived'
] and not options.force:
## manually announced ??
wfo.status = 'done'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog(
'closor',
'%s is announced already : %s' % (wfo.name, wfo.wm_status))
session.commit()
if jump_the_line:
wfi.sendLog('closor', 'Announcing while completing')
expected_lumis = 1
if not 'TotalInputLumis' in wfi.request:
print wfo.name, "has not been assigned yet, or the database is corrupted"
elif wfi.request['TotalInputLumis'] == 0:
print wfo.name, "is corrupted with 0 expected lumis"
else:
expected_lumis = wfi.request['TotalInputLumis']
## what are the outputs
outputs = wfi.request['OutputDatasets']
## check whether the number of lumis is as expected for each
all_OK = defaultdict(lambda: False)
stats = defaultdict(int)
#print outputs
if len(outputs):
print wfo.name, wfi.request['RequestStatus']
for out in outputs:
event_count, lumi_count = getDatasetEventsAndLumis(dataset=out)
odb = session.query(Output).filter(
Output.datasetname == out).first()
if not odb:
print "adding an output object", out
odb = Output(datasetname=out)
odb.workflow = wfo
session.add(odb)
odb.nlumis = lumi_count
odb.nevents = event_count
odb.workfow_id = wfo.id
if odb.expectedlumis < expected_lumis:
odb.expectedlumis = expected_lumis
else:
expected_lumis = odb.expectedlumis
odb.date = time.mktime(time.gmtime())
session.commit()
fraction = lumi_count / float(expected_lumis) * 100.
completion_line = "%60s %d/%d = %3.2f%%" % (
out, lumi_count, expected_lumis, fraction)
wfi.sendLog('closor', "\t%s" % completion_line)
if wfi.isRelval() and fraction < batch_goodness:
batch_warnings[wfi.getCampaign()].add(completion_line)
stats[out] = lumi_count
all_OK[out] = True
## check for at least one full copy prior to moving on
in_full = {}
for out in outputs:
in_full[out] = []
presence = getDatasetPresence(url, out)
where = [site for site, info in presence.items() if info[0]]
if where:
all_OK[out] = True
print out, "is in full at", ",".join(where)
in_full[out] = copy.deepcopy(where)
else:
going_to = wfi.request['NonCustodialSites'] + wfi.request[
'CustodialSites']
wfi.sendLog(
'closor', "%s is not in full anywhere. send to %s" %
(out, ",".join(sorted(going_to))))
at_destination = dict([(k, v) for (k, v) in presence.items()
if k in going_to])
else_where = dict([(k, v) for (k, v) in presence.items()
if not k in going_to])
print json.dumps(at_destination)
print json.dumps(else_where, indent=2)
## do the full stuck transfer study, missing files and shit !
for there in going_to:
late_info = findLateFiles(url, out, going_to=there)
for l in late_info:
l.update({"workflow": wfo.name, "dataset": out})
all_late_files.extend(late_info)
if check_fullcopy_to_announce:
## only set this false if the check is relevant
all_OK[out] = False
## verify if we have to do harvesting
if not options.no_harvest and not jump_the_line:
(OK, requests) = spawn_harvesting(url, wfi, in_full)
all_OK.update(OK)
## only that status can let me go into announced
if all(all_OK.values()) and (
(wfi.request['RequestStatus'] in ['closed-out']) or options.force
or jump_the_line):
print wfo.name, "to be announced"
results = []
if not results:
for out in outputs:
if out in stats and not stats[out]:
continue
_, dsn, process_string, tier = out.split('/')
if all_OK[out]:
results.append(setDatasetStatus(out, 'VALID'))
if all_OK[out] and wfi.isRelval():
## make the specific relval rules and the replicas
## figure the destination(s) out
destinations = set()
if tier != "RECO" and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if tier == "GEN-SIM":
destinations.add('T1_US_FNAL_Disk')
if tier == "GEN-SIM-DIGI-RAW":
destinations.add('T1_US_FNAL_Disk')
if tier == "GEN-SIM-RECO":
destinations.add('T1_US_FNAL_Disk')
if "RelValTTBar" in dsn and "TkAlMinBias" in process_string and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if "MinimumBias" in dsn and "SiStripCalMinBias" in process_string and tier != "ALCARECO":
destinations.add('T2_CH_CERN')
if destinations:
wfi.sendLog(
'closor', '%s to go to %s' %
(out, ', '.join(sorted(destinations))))
## call to makereplicarequest under relval => done
for site in destinations:
result = makeReplicaRequest(
url,
site, [out],
'Copy for release validation consumption',
priority='normal',
approve=True,
mail=False,
group='RelVal')
try:
request_id = result['phedex'][
'request_created'][0]['id']
results.append(True)
except:
results.append('Failed relval transfer')
elif all_OK[out]:
campaign = None
try:
campaign = out.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request and wfi.request[
'Campaign']:
campaign = wfi.request['Campaign']
to_DDM = False
## campaign override
if campaign and campaign in CI.campaigns and 'toDDM' in CI.campaigns[
campaign] and tier in CI.campaigns[campaign][
'toDDM']:
to_DDM = True
## by typical enabling
if tier in UC.get("tiers_to_DDM"):
to_DDM = True
## check for unitarity
if not tier in UC.get("tiers_no_DDM") + UC.get(
"tiers_to_DDM"):
print "tier", tier, "neither TO or NO DDM for", out
results.append('Not recognitized tier %s' % tier)
#sendEmail("failed DDM injection","could not recognize %s for injecting in DDM"% out)
sendLog(
'closor',
"could not recognize %s for injecting in DDM" %
out,
level='critical')
continue
n_copies = 1
destinations = []
if to_DDM and campaign and campaign in CI.campaigns and 'DDMcopies' in CI.campaigns[
campaign]:
ddm_instructions = CI.campaigns[campaign][
'DDMcopies']
if type(ddm_instructions) == int:
n_copies = CI.campaigns[campaign]['DDMcopies']
elif type(ddm_instructions) == dict:
## a more fancy configuration
for ddmtier, indication in ddm_instructions.items(
):
if ddmtier == tier or ddmtier in [
'*', 'all'
]:
## this is for us
if 'N' in indication:
n_copies = indication['N']
if 'host' in indication:
destinations = indication['host']
destination_spec = ""
if destinations:
destination_spec = "--destination=" + ",".join(
destinations)
group_spec = "" ## not used yet
### should make this a campaign configuration
## inject to DDM when necessary
if to_DDM:
print "Sending", out, " to DDM"
status = pass_to_dynamo(
[out],
N=n_copies,
sites=destinations if destinations else None,
group=group_spec if group_spec else None)
results.append(status)
if status == True:
wfi.sendLog(
'closor',
'%s is send to dynamo in %s copies %s %s' %
(out, n_copies, sorted(destinations),
group_spec))
else:
sendLog('closor',
"could not add " + out +
" to dynamo pool. check closor logs.",
level='critical')
wfi.sendLog(
'closor', "could not add " + out +
" to dynamo pool. check closor logs.")
else:
print wfo.name, "no stats for announcing", out
results.append('No Stats')
if all(
map(lambda result: result in ['None', None, True],
results)):
if not jump_the_line:
## only announce if all previous are fine
res = reqMgrClient.announceWorkflowCascade(
url, wfo.name)
if not res in ['None', None]:
## check the status again, it might well have toggled
wl_bis = workflowInfo(url, wfo.name)
wfo.wm_status = wl_bis.request['RequestStatus']
session.commit()
if wl_bis.request['RequestStatus'] in [
'announced', 'normal-archived'
]:
res = None
else:
## retry ?
res = reqMgrClient.announceWorkflowCascade(
url, wfo.name)
results.append(res)
#print results
if all(map(lambda result: result in ['None', None, True],
results)):
if jump_the_line:
if not 'announced' in wfo.status:
wfo.status = wfo.status.replace(
'announce', 'announced')
else:
wfo.status = 'done'
session.commit()
wfi.sendLog('closor', "workflow outputs are announced")
else:
wfi.sendLog(
'closor', "Error with %s to be announced \n%s" %
(wfo.name, json.dumps(results)))
elif wfi.request['RequestStatus'] in [
'failed', 'aborted', 'aborted-archived', 'rejected',
'rejected-archived', 'aborted-completed'
]:
if wfi.isRelval():
wfo.status = 'forget'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog(
'closor',
"%s is %s, but will not be set in trouble to find a replacement."
% (wfo.name, wfo.wm_status))
else:
wfo.status = 'trouble'
wfo.wm_status = wfi.request['RequestStatus']
session.commit()
else:
print wfo.name, "not good for announcing:", wfi.request[
'RequestStatus']
wfi.sendLog('closor', "cannot be announced")
held.add(wfo.name)
days_late = 0.
retries_late = 10
really_late_files = [
info for info in all_late_files if info['retries'] >= retries_late
]
really_late_files = [
info for info in really_late_files
if info['delay'] / (60 * 60 * 24.) >= days_late
]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s' % (
len(really_late_files), days_late, retries_late,
json.dumps(really_late_files, indent=2))
#sendEmail('waiting for files to announce', subject)
sendLog('closor', subject, level='warning')
sendLog('closor', subject)
print subject
open('%s/stuck_files.json' % monitor_dir,
'w').write(json.dumps(really_late_files, indent=2))
if held:
sendLog('closor',
"the workflows below are held up \n%s" %
("\n".join(sorted(held))),
level='critical')
#batches = json.loads(open('batches.json').read())
for bname, go in batch_go.items():
if go:
subject = "Release Validation Samples Batch %s" % bname
issues = ""
if batch_warnings[bname]:
issues = "The following datasets have outstanding completion (<%d%%) issues:\n\n" % batch_goodness
issues += "\n".join(sorted(batch_warnings[bname]))
issues += "\n\n"
text = """
Dear all,
a batch of release validation workflows has finished.
Batch ID:
%s
Detail of the workflows
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
%s
This is an automated message.
""" % (bname, bname, issues)
to = ['*****@*****.**']
sendEmail(subject, text, destination=to)
def closor(url, specific=None, options=None):
if userLock(): return
mlock = moduleLock()
if mlock() and not options.manual: return
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
BI = batchInfo()
CloseI = closeoutInfo()
all_late_files = []
jump_the_line = options.announce if options else False
if jump_the_line:
print "announce option is on. Checking on things on-going ready to be announced"
wfs = session.query(Workflow).filter(
Workflow.status.contains('announce')).filter(
sqlalchemy.not_(Workflow.status.contains('announced'))).all()
else:
print "regular option. Checking on things done and to be announced"
wfs = session.query(Workflow).filter(Workflow.status == 'close').all()
if specific:
wfs = [wfo for wfo in wfs if specific in wfo.name]
wfs_n = [w.name for w in wfs]
print "unique names?"
print len(set(wfs_n)) == len(wfs_n)
held = set()
print len(wfs), "closing"
random.shuffle(wfs)
max_per_round = UC.get('max_per_round').get('closor', None)
if options.limit: max_per_round = options.limit
if max_per_round:
## order them by priority
all_closedout = sorted(getWorkflows(url, 'closed-out', details=True),
key=lambda r: r['RequestPriority'])
all_closedout = [r['RequestName'] for r in all_closedout]
def rank(wfn):
return all_closedout.index(wfn) if wfn in all_closedout else 0
wfs = sorted(wfs, key=lambda wfo: rank(wfo.name), reverse=True)
wfs = wfs[:max_per_round]
batch_go = {}
batch_warnings = defaultdict(set)
batch_extreme_warnings = defaultdict(set)
batch_goodness = UC.get("batch_goodness")
closers = []
print len(wfs), "closing"
th_start = time.mktime(time.gmtime())
for iwfo, wfo in enumerate(wfs):
if specific and not specific in wfo.name: continue
if not options.manual and (
'cmsunified_task_HIG-RunIIFall17wmLHEGS-05036__v1_T_200712_005621_4159'
.lower() in (wfo.name).lower() or
'pdmvserv_task_HIG-RunIISummer16NanoAODv7-03979__v1_T_200915_013748_1986'
.lower() in (wfo.name).lower()):
continue
closers.append(
CloseBuster(
wfo=wfo,
url=url,
CI=CI,
UC=UC,
jump_the_line=jump_the_line,
batch_goodness=batch_goodness,
batch_go=batch_go,
#stats = stats,
batch_warnings=batch_warnings,
batch_extreme_warnings=batch_extreme_warnings,
all_late_files=all_late_files,
held=held,
))
run_threads = ThreadHandler(threads=closers,
n_threads=options.threads,
sleepy=10,
timeout=None,
verbose=True,
label='closor')
run_threads.start()
## waiting on all to complete
while run_threads.is_alive():
#print "Waiting on closing threads",time.asctime(time.gmtime())
time.sleep(5)
JC = JIRAClient() if up.status.get('jira', False) else None
print len(
run_threads.threads), "finished thread to gather information from"
failed_threads = 0
for to in run_threads.threads:
if to.failed:
failed_threads += 1
continue
if to.outs:
for outO in to.outs:
out = outO.datasetname
odb = session.query(Output).filter(
Output.datasetname == out).first()
if not odb:
print "adding an output object", out
session.add(outO)
else:
odb.date = outO.date
if to.to_status:
to.wfo.status = to.to_status
if JC and to.to_status == "done" and to.wfi:
jiras = JC.find({"prepid": to.wfi.request['PrepID']})
for jira in jiras:
JC.close(jira.key)
if to.to_wm_status:
to.wfo.wm_status = to.to_wm_status
if to.closing:
CloseI.pop(to.wfo.name)
session.commit()
th_stop = time.mktime(time.gmtime())
if wfs:
time_spend_per_workflow = (th_stop - th_start) / float(len(wfs))
print "Average time spend per workflow is", time_spend_per_workflow
if float(failed_threads / run_threads.n_threads) > 0:
sendLog('checkor',
'%d/%d threads have failed, better check this out' %
(failed_threads, run_threads.n_threads),
level='critical')
sendEmail(
'checkor', '%d/%d threads have failed, better check this out' %
(failed_threads, run_threads.n_threads))
days_late = 0.
retries_late = 10
really_late_files = [
info for info in all_late_files if info['retries'] >= retries_late
]
really_late_files = [
info for info in really_late_files
if info['delay'] / (60 * 60 * 24.) >= days_late
]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s' % (
len(really_late_files), days_late, retries_late,
json.dumps(really_late_files, indent=2))
#sendEmail('waiting for files to announce', subject)
sendLog('closor', subject, level='warning')
sendLog('closor', subject)
print subject
open('%s/stuck_files.json' % monitor_dir,
'w').write(json.dumps(really_late_files, indent=2))
if held:
sendLog('closor',
"the workflows below are held up \n%s" %
("\n".join(sorted(held))),
level='critical')
for bname, go in batch_go.items():
if go:
subject = "Release Validation Samples Batch %s" % bname
issues = ""
#if batch_warnings[ bname ]:
# issues="The following datasets have outstanding completion (<%d%%) issues:\n\n"% batch_goodness
# issues+="\n".join( sorted( batch_warnings[ bname ] ))
# issues+="\n\n"
if batch_extreme_warnings[bname]:
subject = "Low Statistics for %s" % bname
issues = "The following datasets have outstanding completion (<50%%) issues:\n\n"
issues += "\n".join(sorted(batch_extreme_warnings[bname]))
issues += "\n\n"
elif batch_warnings[bname]:
issues = "The following datasets have outstanding completion (<%d%%) issues:\n\n" % batch_goodness
issues += "\n".join(sorted(batch_warnings[bname]))
issues += "\n\n"
text = ""
text += "Dear all,\n\n"
text += "A batch of release validation workflows has finished.\n\n"
text += "Batch ID:\n\n"
text += "%s\n\n" % (bname)
text += "Detail of the workflows\n\n"
text += "https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s\n\n" % (
bname)
text += "%s\n\n" % (issues)
text += "This is an automated message.\n\n"
text += ""
to = ['*****@*****.**']
sendEmail(subject, text, destination=to)
## just announced ; take it out now.
BI.pop(bname)
deleteCampaignConfig(bname)
if os.path.isfile('.closor_stop'):
print "The loop on workflows was shortened"
sendEmail('closor',
'Closor loop was shortened artificially using .closor_stop')
os.system('rm -f .closor_stop')
def batchor( url ):
UC = unifiedConfiguration()
## get all workflows in assignment-approved with SubRequestType = relval
all_wfs = []
for user in UC.get("user_relval"):
all_wfs = getWorkflows(url, 'assignment-approved', details=True, user=user, rtype='TaskChain')
wfs = filter( lambda r :r['SubRequestType'] == 'RelVal' if 'SubRequestType' in r else False, all_wfs)
## need a special treatment for those
hi_wfs = filter( lambda r :r['SubRequestType'] == 'HIRelVal' if 'SubRequestType' in r else False, all_wfs)
by_campaign = defaultdict(set)
by_hi_campaign = defaultdict(set)
for wf in wfs:
print "Relval:",wf['RequestName'], wf['Campaign']
by_campaign[wf['Campaign']].add( wf['RequestName'] )
for wf in hi_wfs:
print "HI Relval:",wf['RequestName'], wf['Campaign']
by_hi_campaign[wf['Campaign']].add( wf['RequestName'] )
default_setup = {
"go" :True,
"parameters" : {
"SiteWhitelist": [ "T1_US_FNAL" ],
"MergedLFNBase": "/store/relval",
"Team" : "relval",
"NonCustodialGroup" : "RelVal"
},
"custodial" : "T1_US_FNAL_MSS",
"phedex_group" : "RelVal",
"lumisize" : -1,
"fractionpass" : 0.0,
"maxcopies" : 1
}
default_hi_setup = copy.deepcopy( default_setup )
add_on = {}
batches = json.loads( open('batches.json').read() )
for campaign in by_campaign:
## get a bunch of information
setup = copy.deepcopy( default_setup )
add_on[campaign] = setup
sendLog('batchor','Adding the relval campaigns %s with parameters \n%s'%( campaign, json.dumps( setup, indent=2)),level='critical')
if not campaign in batches: batches[campaign] = []
batches[campaign] = list(set(list(copy.deepcopy( by_campaign[campaign] )) + batches[campaign] ))
for campaign in by_hi_campaign:
## get a bunch of information
setup = copy.deepcopy( default_hi_setup )
hi_site = random.choice(["T1_DE_KIT","T1_FR_CCIN2P3"])
setup["parameters"]["SiteWhitelist"]=[ hi_site ]
add_on[campaign] = setup
sendLog('batchor','Adding the HI relval campaigns %s with parameters \n%s'%( campaign, json.dumps( setup, indent=2)),level='critical')
if not campaign in batches: batches[campaign] = []
batches[campaign] = list(set(list(copy.deepcopy( by_hi_campaign[campaign] )) + batches[campaign] ))
open('batches.json','w').write( json.dumps( batches , indent=2 ) )
## open the campaign configuration
campaigns = json.loads( open('campaigns.relval.json').read() )
## protect for overwriting ??
for new_campaign in list(set(add_on.keys())-set(campaigns.keys())):
## this is new, and can be announced as such
print new_campaign,"is new stuff"
workflows = by_campaign[new_campaign]
requester = list(set([wf.split('_')[0] for wf in workflows]))
subject = "Request of RelVal samples batch %s"% new_campaign
text="""Dear all,
A new batch of relval workflows was requested.
Batch ID:
%s
Requestor:
%s
Details of the workflows:
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
This is an automated message"""%( new_campaign,
', '.join(requester),
new_campaign,
#'\n'.join( sorted(workflows) )
)
print subject
print text
to = ['*****@*****.**']
sendEmail(subject, text, destination=to)
sendLog('batchor',text, level='critical')
## merge all anyways
campaigns.update( add_on )
## write it out for posterity
open('campaigns.json.updated','w').write(json.dumps( campaigns , indent=2))
## read back
rread = json.loads(open('campaigns.json.updated').read())
os.system('mv campaigns.json.updated campaigns.relval.json')
def batchor(url):
UC = unifiedConfiguration()
SI = global_SI()
CI = campaignInfo()
BI = batchInfo()
## get all workflows in assignment-approved with SubRequestType = relval
all_wfs = []
for user in UC.get("user_relval"):
all_wfs.extend(
getWorkflows(url,
'assignment-approved',
details=True,
user=user,
rtype='TaskChain'))
wfs = filter(
lambda r: r['SubRequestType'] == 'RelVal'
if 'SubRequestType' in r else False, all_wfs)
## need a special treatment for those
hi_wfs = filter(
lambda r: r['SubRequestType'] == 'HIRelVal'
if 'SubRequestType' in r else False, all_wfs)
by_campaign = defaultdict(set)
by_hi_campaign = defaultdict(set)
for wf in wfs:
print "Relval:", wf['RequestName'], wf['Campaign']
by_campaign[wf['Campaign']].add(wf['PrepID'])
for wf in hi_wfs:
print "HI Relval:", wf['RequestName'], wf['Campaign']
by_hi_campaign[wf['Campaign']].add(wf['PrepID'])
default_setup = {
"go": True,
"parameters": {
"SiteWhitelist": ["T1_US_FNAL"],
"MergedLFNBase": "/store/relval",
"Team": "relval",
"NonCustodialGroup": "RelVal"
},
"custodial_override": "notape",
"phedex_group": "RelVal",
"lumisize": -1,
"fractionpass": 0.0,
"maxcopies": 1
}
default_hi_setup = copy.deepcopy(default_setup)
add_on = {}
relval_routing = UC.get('relval_routing')
def pick_one_site(p):
## modify the parameters on the spot to have only one site
if "parameters" in p and "SiteWhitelist" in p["parameters"] and len(
p["parameters"]["SiteWhitelist"]) > 1:
choose_from = list(
set(p["parameters"]["SiteWhitelist"]) & set(SI.sites_ready))
picked = random.choice(choose_from)
print "picked", picked, "from", choose_from
p["parameters"]["SiteWhitelist"] = [picked]
batches = BI.all()
for campaign in by_campaign:
if campaign in batches: continue
## get a bunch of information
setup = copy.deepcopy(default_setup)
for key in relval_routing:
if key in campaign:
## augment with the routing information
augment_with = relval_routing[key]
print "Modifying the batch configuration because of keyword", key
print "with", augment_with
setup = deep_update(setup, augment_with)
pick_one_site(setup)
add_on[campaign] = setup
sendLog('batchor',
'Adding the relval campaigns %s with parameters \n%s' %
(campaign, json.dumps(setup, indent=2)),
level='critical')
BI.update(campaign, by_campaign[campaign])
for campaign in by_hi_campaign:
if campaign in batches: continue
## get a bunch of information
setup = copy.deepcopy(default_hi_setup)
possible_sites = set(["T1_DE_KIT", "T1_FR_CCIN2P3"])
hi_site = random.choice(list(possible_sites))
setup["parameters"]["SiteWhitelist"] = [hi_site]
pick_one_site(setup)
add_on[campaign] = setup
sendLog('batchor',
'Adding the HI relval campaigns %s with parameters \n%s' %
(campaign, json.dumps(setup, indent=2)),
level='critical')
BI.update(campaign, by_hi_campaign[campaign])
## only new campaigns in announcement
for new_campaign in list(
set(add_on.keys()) - set(CI.all(c_type='relval'))):
## this is new, and can be announced as such
print new_campaign, "is new stuff"
subject = "Request of RelVal samples batch %s" % new_campaign
text = """Dear all,
A new batch of relval workflows was requested.
Batch ID:
%s
Details of the workflows:
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
This is an automated message""" % (
new_campaign,
new_campaign,
)
print subject
print text
to = ['*****@*****.**']
sendEmail(subject, text, destination=to)
sendLog('batchor', text, level='critical')
## go through all existing campaigns and remove the ones not in use anymore ?
for old_campaign in CI.all(c_type='relval'):
all_in_batch = getWorkflowByCampaign(url, old_campaign, details=True)
if not all_in_batch: continue
is_batch_done = all(
map(
lambda s: not s in [
'completed', 'force-complete', 'running-open',
'running-closed', 'acquired', 'assigned',
'assignment-approved'
], [wf['RequestStatus'] for wf in all_in_batch]))
## check all statuses
if is_batch_done:
#print "batch",old_campaign,"can be closed or removed if necessary"
#campaigns[old_campaign]['go'] = False ## disable
CI.pop(old_campaign) ## or just drop it all together ?
BI.pop(old_campaign)
print "batch", old_campaign, " configuration was removed"
## merge all anyways
CI.update(add_on, c_type='relval')
def checkor(url, spec=None, options=None):
fDB = closeoutInfo()
if userLock():
return
if duplicateLock():
return
UC = unifiedConfiguration()
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=["mcm"])
if not up.check():
return
use_mcm = up.status["mcm"]
wfs = []
if options.fetch:
## get all in running and check
wfs.extend(session.query(Workflow).filter(Workflow.status == "away").all())
wfs.extend(session.query(Workflow).filter(Workflow.status == "assistance").all())
if options.nofetch:
## than get all in need for assistance
wfs.extend(session.query(Workflow).filter(Workflow.status.startswith("assistance-")).all())
custodials = defaultdict(list) # sites : dataset list
transfers = defaultdict(list) # sites : dataset list
invalidations = [] # a list of files
SI = global_SI
CI = campaignInfo()
mcm = McMClient(dev=False)
def get_campaign(output, wfi):
campaign = None
try:
campaign = output.split("/")[2].split("-")[0]
except:
if "Campaign" in wfi.request:
campaign = wfi.request["Campaign"]
return campaign
by_passes = []
holdings = []
for bypassor, email in [
("jbadillo", "*****@*****.**"),
("vlimant", "*****@*****.**"),
("jen_a", "*****@*****.**"),
]:
bypass_file = "/afs/cern.ch/user/%s/%s/public/ops/bypass.json" % (bypassor[0], bypassor)
if not os.path.isfile(bypass_file):
print "no file", bypass_file
continue
try:
by_passes.extend(json.loads(open(bypass_file).read()))
except:
print "cannot get by-passes from", bypass_file, "for", bypassor
sendEmail("malformated by-pass information", "%s is not json readable" % (bypass_file), destination=[email])
holding_file = "/afs/cern.ch/user/%s/%s/public/ops/onhold.json" % (bypassor[0], bypassor)
if not os.path.isfile(holding_file):
print "no file", holding_file
continue
try:
holdings.extend(json.loads(open(holding_file).read()))
except:
print "cannot get holdings from", holding_file, "for", bypassor
sendEmail(
"malformated by-pass information", "%s is not json readable" % (holding_file), destination=[email]
)
total_running_time = 5.0 * 60.0
sleep_time = max(0.5, total_running_time / len(wfs))
for wfo in wfs:
if spec and not (spec in wfo.name):
continue
time.sleep(sleep_time)
print "checking on", wfo.name
## get info
wfi = workflowInfo(url, wfo.name)
## make sure the wm status is up to date.
# and send things back/forward if necessary.
wfo.wm_status = wfi.request["RequestStatus"]
if wfo.wm_status == "closed-out":
## manually closed-out
print wfo.name, "is already", wfo.wm_status
wfo.status = "close"
session.commit()
continue
elif wfo.wm_status in [
"failed",
"aborted",
"aborted-archived",
"rejected",
"rejected-archived",
"aborted-completed",
]:
## went into trouble
wfo.status = "trouble"
print wfo.name, "is in trouble", wfo.wm_status
session.commit()
continue
elif wfo.wm_status in ["assigned", "acquired"]:
## not worth checking yet
print wfo.name, "not running yet"
session.commit()
continue
if "-onhold" in wfo.status:
if wfo.name in holdings and wfo.name not in by_passes:
print wfo.name, "on hold"
continue
if wfo.name in holdings and wfo.name not in by_passes:
wfo.status = "assistance-onhold"
print "setting", wfo.name, "on hold"
session.commit()
continue
if wfo.wm_status != "completed" and not wfo.name in by_passes:
## for sure move on with closeout check if in completed
print "no need to check on", wfo.name, "in status", wfo.wm_status
session.commit()
continue
session.commit()
sub_assistance = "" # if that string is filled, there will be need for manual assistance
is_closing = True
## get it from somewhere
by_pass_checks = False
if wfo.name in by_passes:
print "we can bypass checks on", wfo.name
by_pass_checks = True
for bypass in by_passes:
if bypass in wfo.name:
print "we can bypass", wfo.name, "because of keyword", bypass
by_pass_checks = True
break
if not CI.go(wfi.request["Campaign"]) and not by_pass_checks:
print "No go for", wfo.name
continue
# tuck out DQMIO/DQM
wfi.request["OutputDatasets"] = [out for out in wfi.request["OutputDatasets"] if not "/DQM" in out]
## anything running on acdc
familly = getWorkflowById(url, wfi.request["PrepID"], details=True)
acdc = []
acdc_inactive = []
has_recovery_going = False
had_any_recovery = False
for member in familly:
if member["RequestType"] != "Resubmission":
continue
if member["RequestName"] == wfo.name:
continue
if member["RequestDate"] < wfi.request["RequestDate"]:
continue
if member["RequestStatus"] in [
"running-open",
"running-closed",
"assignment-approved",
"assigned",
"acquired",
]:
print wfo.name, "still has an ACDC running", member["RequestName"]
acdc.append(member["RequestName"])
# print json.dumps(member,indent=2)
## hook for just waiting ...
is_closing = False
has_recovery_going = True
elif member["RequestStatus"] == None:
print member["RequestName"], "is not real"
pass
else:
acdc_inactive.append(member["RequestName"])
had_any_recovery = True
## completion check
percent_completions = {}
# print "let's see who is crashing", wfo.name
# print wfi.request['TotalInputEvents'],wfi.request['TotalInputLumis']
if not "TotalInputEvents" in wfi.request:
event_expected, lumi_expected = 0, 0
if not "recovery" in wfo.status:
sendEmail(
"missing member of the request",
"TotalInputEvents is missing from the workload of %s" % wfo.name,
destination=["*****@*****.**"],
)
else:
event_expected, lumi_expected = wfi.request["TotalInputEvents"], wfi.request["TotalInputLumis"]
if "RequestNumEvents" in wfi.request:
event_expected = int(wfi.request["RequestNumEvents"])
elif "Task1" in wfi.request and "RequestNumEvents" in wfi.request["Task1"]:
event_expected = int(wfi.request["Task1"]["RequestNumEvents"])
fractions_pass = {}
for output in wfi.request["OutputDatasets"]:
event_count, lumi_count = getDatasetEventsAndLumis(dataset=output)
percent_completions[output] = 0.0
if lumi_expected:
percent_completions[output] = lumi_count / float(lumi_expected)
if event_expected:
percent_completions[output] = max(percent_completions[output], event_count / float(event_expected))
fractions_pass[output] = 0.95
c = get_campaign(output, wfi)
if c in CI.campaigns and "fractionpass" in CI.campaigns[c]:
fractions_pass[output] = CI.campaigns[c]["fractionpass"]
print "overriding fraction to", fractions_pass[output], "for", output
if options.fractionpass:
fractions_pass[output] = options.fractionpass
print "overriding fraction to", fractions_pass[output], "by command line for", output
if not all([percent_completions[out] >= fractions_pass[out] for out in fractions_pass]):
print wfo.name, "is not completed"
print json.dumps(percent_completions, indent=2)
print json.dumps(fractions_pass, indent=2)
## hook for creating automatically ACDC ?
if has_recovery_going:
sub_assistance += "-recovering"
elif had_any_recovery:
## we want to have this looked at
sub_assistance += "-manual"
else:
sub_assistance += "-recovery"
is_closing = False
## correct lumi < 300 event per lumi
events_per_lumi = {}
for output in wfi.request["OutputDatasets"]:
events_per_lumi[output] = getDatasetEventsPerLumi(output)
lumi_upper_limit = {}
for output in wfi.request["OutputDatasets"]:
upper_limit = 301.0
campaign = get_campaign(output, wfi)
# if 'EventsPerLumi' in wfi.request and 'FilterEfficiency' in wfi.request:
# upper_limit = 1.5*wfi.request['EventsPerLumi']*wfi.request['FilterEfficiency']
# print "setting the upper limit of lumisize to",upper_limit,"by request configuration"
if campaign in CI.campaigns and "lumisize" in CI.campaigns[campaign]:
upper_limit = CI.campaigns[campaign]["lumisize"]
print "overriding the upper lumi size to", upper_limit, "for", campaign
if options.lumisize:
upper_limit = options.lumisize
print "overriding the upper lumi size to", upper_limit, "by command line"
lumi_upper_limit[output] = upper_limit
if any([events_per_lumi[out] >= lumi_upper_limit[out] for out in events_per_lumi]):
print wfo.name, "has big lumisections"
print json.dumps(events_per_lumi, indent=2)
## hook for rejecting the request ?
sub_assistance += "-biglumi"
is_closing = False
any_presence = {}
for output in wfi.request["OutputDatasets"]:
any_presence[output] = getDatasetPresence(url, output, vetoes=[])
## custodial copy
custodial_locations = {}
custodial_presences = {}
for output in wfi.request["OutputDatasets"]:
custodial_presences[output] = [s for s in any_presence[output] if "MSS" in s]
custodial_locations[output] = phedexClient.getCustodialSubscriptionRequestSite(output)
if not custodial_locations[output]:
custodial_locations[output] = []
## presence in phedex
phedex_presence = {}
for output in wfi.request["OutputDatasets"]:
phedex_presence[output] = phedexClient.getFileCountDataset(url, output)
vetoed_custodial_tier = UC.get("tiers_with_no_custodial")
out_worth_checking = [
out for out in custodial_locations.keys() if out.split("/")[-1] not in vetoed_custodial_tier
]
size_worth_checking = sum(
[getDatasetSize(out) / 1023.0 for out in out_worth_checking]
) ## size in TBs of all outputs
if not all(map(lambda sites: len(sites) != 0, [custodial_locations[out] for out in out_worth_checking])):
print wfo.name, "has not all custodial location"
print json.dumps(custodial_locations, indent=2)
##########
## hook for making a custodial replica ?
custodial = None
## get from other outputs
for output in out_worth_checking:
if len(custodial_locations[output]):
custodial = custodial_locations[output][0]
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the other output custodial:", custodial, "because of limited space"
custodial = None
## try to get it from campaign configuration
if not custodial:
for output in out_worth_checking:
campaign = get_campaign(output, wfi)
if campaign in CI.campaigns and "custodial" in CI.campaigns[campaign]:
custodial = CI.campaigns[campaign]["custodial"]
print "Setting custodial to", custodial, "from campaign configuration"
break
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the unified configuration custodial:", custodial, "because of limited space"
custodial = None
## get from the parent
pick_custodial = True
if not custodial and "InputDataset" in wfi.request:
## this is terribly dangerous to assume only
parents_custodial = phedexClient.getCustodialSubscriptionRequestSite(wfi.request["InputDataset"])
###parents_custodial = findCustodialLocation(url, wfi.request['InputDataset'])
if not parents_custodial:
parents_custodial = []
if len(parents_custodial):
custodial = parents_custodial[0]
else:
print "the input dataset", wfi.request[
"InputDataset"
], "does not have custodial in the first place. abort"
sendEmail(
"dataset has no custodial location",
"Please take a look at %s in the logs of checkor" % wfi.request["InputDataset"],
)
is_closing = False
pick_custodial = False
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the parent custodial:", custodial, "because of limited space"
custodial = None
if not custodial and pick_custodial:
## pick one at random
custodial = SI.pick_SE(size=size_worth_checking)
if not custodial:
print "cannot find a custodial for", wfo.name
sendEmail(
"cannot find a custodial",
"cannot find a custodial for %s probably because of the total output size %d"
% (wfo.name, size_worth_checking),
)
if custodial and ((not sub_assistance and not acdc) or by_pass_checks):
## remember how much you added this round already ; this stays locally
SI.storage[custodial] -= size_worth_checking
## register the custodial request, if there are no other big issues
for output in out_worth_checking:
if not len(custodial_locations[output]):
if phedex_presence[output] >= 1:
custodials[custodial].append(output)
else:
print "no file in phedex for", output, " not good to add to custodial requests"
is_closing = False
## disk copy
disk_copies = {}
for output in wfi.request["OutputDatasets"]:
disk_copies[output] = [s for s in any_presence[output] if (not "MSS" in s) and (not "Buffer" in s)]
if not all(map(lambda sites: len(sites) != 0, disk_copies.values())):
print wfo.name, "has not all output on disk"
print json.dumps(disk_copies, indent=2)
## presence in dbs
dbs_presence = {}
dbs_invalid = {}
for output in wfi.request["OutputDatasets"]:
dbs_presence[output] = dbs3Client.getFileCountDataset(output)
dbs_invalid[output] = dbs3Client.getFileCountDataset(output, onlyInvalid=True)
fraction_invalid = 0.01
if (
not all(
[
dbs_presence[out] == (dbs_invalid[out] + phedex_presence[out])
for out in wfi.request["OutputDatasets"]
]
)
and not options.ignorefiles
):
print wfo.name, "has a dbs,phedex mismatch"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
## hook for just waiting ...
is_closing = False
if (
not all(
[
(dbs_invalid[out] <= int(fraction_invalid * dbs_presence[out]))
for out in wfi.request["OutputDatasets"]
]
)
and not options.ignorefiles
):
print wfo.name, "has a dbs invalid file level too high"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
## need to be going and taking an eye
sub_assistance += "-invalidfiles"
is_closing = False
## put that heavy part at the end
## duplication check
duplications = {}
if is_closing:
print "starting duplicate checker for", wfo.name
for output in wfi.request["OutputDatasets"]:
print "\tchecking", output
duplications[output] = True
try:
duplications[output] = dbs3Client.duplicateRunLumi(output)
except:
try:
duplications[output] = dbs3Client.duplicateRunLumi(output)
except:
print "was not possible to get the duplicate count for", output
is_closing = False
if any(duplications.values()) and not options.ignoreduplicates:
print wfo.name, "has duplicates"
print json.dumps(duplications, indent=2)
## hook for making file invalidation ?
sub_assistance += "-duplicates"
is_closing = False
## for visualization later on
if not wfo.name in fDB.record:
# print "adding",wfo.name,"to close out record"
fDB.record[wfo.name] = {"datasets": {}, "name": wfo.name, "closeOutWorkflow": None}
fDB.record[wfo.name]["closeOutWorkflow"] = is_closing
for output in wfi.request["OutputDatasets"]:
if not output in fDB.record[wfo.name]["datasets"]:
fDB.record[wfo.name]["datasets"][output] = {}
rec = fDB.record[wfo.name]["datasets"][output]
rec["percentage"] = float("%.2f" % (percent_completions[output] * 100))
rec["duplicate"] = duplications[output] if output in duplications else "N/A"
rec["phedexReqs"] = (
float("%.2f" % any_presence[output][custodial_presences[output][0]][1])
if len(custodial_presences[output]) != 0
else "N/A"
)
rec["closeOutDataset"] = is_closing
rec["transPerc"] = (
float("%.2f" % any_presence[output][disk_copies[output][0]][1])
if len(disk_copies[output]) != 0
else "N/A"
)
rec["correctLumis"] = (
int(events_per_lumi[output]) if (events_per_lumi[output] > lumi_upper_limit[output]) else True
)
rec["missingSubs"] = (
False if len(custodial_locations[output]) == 0 else ",".join(list(set(custodial_locations[output])))
)
rec["dbsFiles"] = dbs_presence[output]
rec["dbsInvFiles"] = dbs_invalid[output]
rec["phedexFiles"] = phedex_presence[output]
rec["acdc"] = "%d / %d" % (len(acdc), len(acdc + acdc_inactive))
if by_pass_checks:
## force closing
is_closing = True
## and move on
if is_closing:
## toggle status to closed-out in request manager
print "setting", wfo.name, "closed-out"
if not options.test:
if wfo.wm_status in ["closed-out", "announced", "normal-archived"]:
print wfo.name, "is already", wfo.wm_status, "not trying to closed-out and assuming it does"
res = None
else:
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
print "close out answer", res
if not res in ["None", None]:
print "retrying to closing out"
print res
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
if res in [None, "None"]:
wfo.status = "close"
session.commit()
else:
print "could not close out", wfo.name, "will try again next time"
else:
## that means there is something that needs to be done acdc, lumi invalidation, custodial, name it
new_status = "assistance" + sub_assistance
print wfo.name, "needs assistance with", new_status
if sub_assistance and wfo.status != new_status and "PrepID" in wfi.request and not "manual" in wfo.status:
pid = wfi.getPrepIDs()[0].replace("task_", "")
# pid = wfi.request['PrepID'].replace('task_','')
## notify
messages = {
"recovery": "Samples completed with missing statistics:\n%s "
% (
"\n".join(
[
"%.2f %% complete for %s" % (percent_completions[output] * 100, output)
for output in wfi.request["OutputDatasets"]
]
)
),
"biglumi": "Samples completed with large luminosity blocks:\n%s "
% (
"\n".join(
[
"%d > %d for %s" % (events_per_lumi[output], lumi_upper_limit[output], output)
for output in wfi.request["OutputDatasets"]
if (events_per_lumi[output] > lumi_upper_limit[output])
]
)
),
"duplicate": "Samples completed with duplicated luminosity blocks:\n%s"
% (
"\n".join(
[
"%s" % output
for output in wfi.request["OutputDatasets"]
if output in duplications and duplications[output]
]
)
),
}
text = "The request %s (%s) is facing issue in production.\n" % (pid, wfo.name)
content = ""
for case in messages:
if case in new_status:
content += "\n" + messages[case] + "\n"
text += content
text += "You are invited to check, while this is being taken care of by Ops.\n"
text += "This is an automated message."
if use_mcm and content:
print "Sending notification back to requestor"
print text
batches = mcm.getA("batches", query="contains=%s&status=announced" % pid)
if len(batches):
## go notify the batch
bid = batches[-1]["prepid"]
print "batch nofication to", bid
mcm.put("/restapi/batches/notify", {"notes": text, "prepid": bid})
## go notify the request
print "request notification to", pid
mcm.put("/restapi/requests/notify", {"message": text, "prepids": [pid]})
## case where the workflow was in manual from recoveror
if not "manual" in wfo.status or new_status != "assistance-recovery":
wfo.status = new_status
if not options.test:
print "setting", wfo.name, "to", wfo.status
session.commit()
else:
print "current status is", wfo.status, "not changing to anything"
fDB.html()
## custodial requests
print "Custodials"
print json.dumps(custodials, indent=2)
for site in custodials:
print ",".join(custodials[site]), "=>", site
if not options.test:
result = makeReplicaRequest(
url,
site,
list(set(custodials[site])),
"custodial copy at production close-out",
custodial="y",
priority="low",
approve=(site in SI.sites_auto_approve),
)
print result
print "Transfers"
print json.dumps(transfers, indent=2)
## replicas requests
for site in transfers:
print ",".join(transfers[site]), "=>", site
if not options.test:
result = None
# result = makeReplicaRequest(url, site, list(set(transfers[site])),"copy to disk at production close-out")
print result
print "File Invalidation"
print invalidations
def mappor(url, options=None):
up = componentInfo(soft=['mcm', 'wtc', 'jira'])
## define regionality site => fallback allowed. feed on an ssb metric ??
mapping = defaultdict(list)
reversed_mapping = defaultdict(list)
regions = defaultdict(list)
UC = unifiedConfiguration()
over_rides = []
use_T0 = ('T0_CH_CERN' in UC.get("site_for_overflow"))
if options.t0: use_T0 = True
if use_T0: over_rides.append('T0_CH_CERN')
use_HLT = ('T2_CH_CERN_HLT' in UC.get("site_for_overflow"))
if options.hlt: use_HLT = True
if use_HLT: over_rides.append('T2_CH_CERN_HLT')
use_CSCS = ('T0_CH_CSCS_HPC' in UC.get("site_for_overflow"))
if options.cscs: use_CSCS = True
if use_CSCS: over_rides.append('T0_CH_CSCS_HPC')
SI = global_SI(over_rides)
#print sorted(SI.all_sites)
#print sorted(SI.sites_T0s)
CI = campaignInfo()
#sites_to_consider = SI.all_sites
sites_to_consider = SI.sites_ready
for site in sites_to_consider:
region = site.split('_')[1]
if not region in [
'US',
'DE',
'IT',
'FR',
'CH',
'ES',
'UK',
'RU' ### latest addition
]:
continue
regions[region] = [region]
def site_in_depletion(s):
return True
if s in SI.sites_pressure:
(m, r, pressure) = SI.sites_pressure[s]
if float(m) < float(r):
print s, m, r, "lacking pressure"
return True
else:
print s, m, r, "pressure"
pass
return False
for site in sites_to_consider:
region = site.split('_')[1]
## fallback to the region, to site with on-going low pressure
within_region = [
fb for fb in sites_to_consider
if any([('_%s_' %
(reg) in fb and fb != site and site_in_depletion(fb))
for reg in regions[region]])
]
#print site,region, within_region
mapping[site] = within_region
for site in sites_to_consider:
if site.split('_')[1] == 'US': ## to all site in the US
## add NERSC
mapping[site].append('T3_US_NERSC')
mapping[site].append('T3_US_SDSC')
mapping[site].append('T3_US_TACC')
mapping[site].append('T3_US_PSC')
## add OSG
mapping[site].append('T3_US_OSG')
#mapping[site].append('T3_US_Colorado')
pass
if use_HLT:
mapping['T2_CH_CERN'].append('T2_CH_CERN_HLT')
if use_T0:
## who can read from T0
mapping['T2_CH_CERN'].append('T0_CH_CERN')
mapping['T1_IT_CNAF'].append('T0_CH_CERN')
mapping['T1_FR_CCIN2P3'].append('T0_CH_CERN')
mapping['T1_DE_KIT'].append('T0_CH_CERN')
if use_CSCS:
## analog config to T0:
mapping['T2_CH_CERN'].append('T0_CH_CSCS_HPC')
mapping['T1_IT_CNAF'].append('T0_CH_CSCS_HPC')
mapping['T1_FR_CCIN2P3'].append('T0_CH_CSCS_HPC')
mapping['T1_DE_KIT'].append('T0_CH_CSCS_HPC')
## temptatively
mapping['T0_CH_CERN'].append('T2_CH_CERN')
## all europ can read from CERN
for reg in ['IT', 'DE', 'UK', 'FR', 'BE', 'ES']:
mapping['T2_CH_CERN'].extend(
[fb for fb in sites_to_consider if '_%s_' % reg in fb])
pass
## all europ T1 among each others
europ_t1 = [
site for site in sites_to_consider if site.startswith('T1')
and any([reg in site for reg in ['IT', 'DE', 'UK', 'FR', 'ES', 'RU']])
]
#print europ_t1
for one in europ_t1:
for two in europ_t1:
if one == two: continue
mapping[one].append(two)
pass
## all EU T1 can read from T0
mapping['T0_CH_CERN'].append(one)
mapping['T0_CH_CERN'].append('T1_US_FNAL')
#mapping['T1_IT_CNAF'].append( 'T1_US_FNAL' )
#mapping['T1_IT_CNAF'].extend( [site for site in SI.sites_ready if '_US_' in site] ) ## all US can read from CNAF
mapping['T1_IT_CNAF'].append('T2_CH_CERN')
mapping['T1_DE_KIT'].append('T2_CH_CERN')
mapping['T2_CH_CERN'].append('T1_IT_CNAF')
mapping['T2_CH_CERN'].append('T1_US_FNAL')
mapping['T2_CH_CERN'].append('T3_CH_CERN_HelixNebula')
mapping['T2_CH_CERN'].append('T3_CH_CERN_HelixNebula_REHA')
for site in sites_to_consider:
if '_US_' in site:
mapping[site].append('T2_CH_CERN')
## make them appear as OK to use
force_sites = []
## overflow CERN to underutilized T1s
upcoming = json.loads(eosRead('%s/GQ.json' % monitor_dir))
for possible in SI.sites_T1s:
if not possible in upcoming:
mapping['T2_CH_CERN'].append(possible)
pass
take_site_out = UC.get('site_out_of_overflow')
for site, fallbacks in mapping.items():
mapping[site] = list(set(fallbacks))
### mapping is a dictionnary where
# key can read from site in values.
### reverserd mapping is a dictionnary where
# key can be read by site in values.
## create the reverse mapping for the condor module
for site, fallbacks in mapping.items():
if site in take_site_out:
print "taking", site, "out of overflow source by unified configuration"
mapping.pop(site)
continue
for fb in fallbacks:
if fb == site:
## remove self
mapping[site].remove(fb)
continue
if fb in take_site_out:
## remove those to be removed
print "taking", fb, "out of overflow destination by unified configuration"
mapping[site].remove(fb)
continue
if not site in reversed_mapping[fb]:
reversed_mapping[fb].append(site)
## write it out and bail
cache = cacheInfo()
cache.store('overflow_mapping', mapping)
cache.store('overflow_reverse_mapping', reversed_mapping)
return
def htmlor( caller = ""):
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
try:
boost = json.loads(open('%s/equalizor.json'%monitor_dir).read())['modifications']
except:
boost = {}
cache = getWorkflows(reqmgr_url,'assignment-approved', details=True)
cache.extend( getWorkflows(reqmgr_url,'acquired', details=True) )
cache.extend( getWorkflows(reqmgr_url,'running-open', details=True) )
cache.extend( getWorkflows(reqmgr_url,'running-closed', details=True) )
def getWL( wfn ):
cached = filter(lambda d : d['RequestName']==wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad(reqmgr_url,wfn)
return wl
def wfl(wf,view=False,p=False,ms=False,within=False,ongoing=False,status=False,update=False):
wfn = wf.name
wfs = wf.wm_status
wl = None
pid = None
wl_pid = None
pids=filter(lambda seg: seg.count('-')==2, wf.name.split('_'))
if len(pids):
pids = pids[:1]
pid=pids[0]
if not pids:
wl = getWL( wf.name )
pids = getPrepIDs( wl )
pid = pids[0]
wl_pid = pid
if 'task' in wf.name:
wl_pid = 'task_'+pid
text=', '.join([
#wfn,
#'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">%s</a> '%(wfn,wfn),
#'<table><tr><td>%s</td></tr></table>'%(wfn),
#'<span>%s</span>'%(wfn),
"%s "%wfn,
'(%s) <br>'%wfs])
text+=', '.join([
'<a href="https://%s/reqmgr2/fetch?rid=%s" target="_blank">dts</a>'%(reqmgr_url,wfn),
'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">dts-req1</a>'%wfn,
#TOFIX '<a href=https://cmsweb.cern.ch/reqmgr/view/showWorkload?requestName=%s target="_blank">wkl</a>'%wfn,
'<a href="https://%s/couchdb/reqmgr_workload_cache/%s" target="_blank">wfc</a>'%(reqmgr_url,wfn),
'<a href="https://%s/reqmgr2/data/request?name=%s" target="_blank">req</a>'%(reqmgr_url,wfn),
#'<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/request?requestName=%s" target="_blank">dwkc</a>'%wfn,
#TOFIX '<a href="https://cmsweb.cern.ch/reqmgr/view/splitting/%s" target="_blank">spl</a>'%wfn,
'<a href="https://cms-pdmv.cern.ch/stats/?RN=%s" target="_blank">vw</a>'%wfn,
'<a href="https://cms-pdmv.cern.ch/stats/restapi/get_one/%s" target="_blank">vwo</a>'%wfn,
'<a href="https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=full&reverse=0&reverse=1&npp=20&subtext=%s&sall=q" target="_blank">elog</a>'%pid,
'<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank">pv</a>'%wfn,
#deprecated '<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/outputDatasetsByRequestName/%s" target="_blank">out</a>'%wfn,
'<a href="closeout.html#%s" target="_blank">clo</a>'%wfn,
'<a href="statuses.html#%s" target="_blank">st</a>'%wfn,
'<a href="https://%s/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s" target="_blank">perf</a>'%(reqmgr_url,wfn)
])
if within and (not view or wfs=='completed'):
wl = getWL( wfn )
dataset =None
if 'InputDataset' in wl:
dataset = wl['InputDataset']
if 'Task1' in wl and 'InputDataset' in wl['Task1']:
dataset = wl['Task1']['InputDataset']
if dataset:
text+=', '.join(['',
'<a href=https://cmsweb.cern.ch/das/request?input=%s target=_blank>input</a>'%dataset,
'<a href=https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions#state=create_since=0;filter=%s target=_blank>sub</a>'%dataset,
'<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/subscriptions?dataset=%s&collapse=n target=_blank>ds</a>'%dataset,
'<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/blockreplicas?dataset=%s target=_blank>rep</a>'%dataset,
])
if p:
cached = filter(lambda d : d['RequestName']==wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad('cmsweb.cern.ch',wfn)
text+=', (%s)'%(wl['RequestPriority'])
pass
if pid:
if ms:
mcm_s = json.loads(os.popen('curl https://cms-pdmv.cern.ch/mcm/public/restapi/requests/get_status/%s --insecure'%pid).read())[pid]
text+=', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm (%s)</a>'%(pid,mcm_s)
else:
text+=', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm</a>'%(pid)
text+=', <a href="https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?prep_id=%s" target="_blank">ac</a>'%(wl_pid)
if status:
if wf.status.startswith('assistance'):
text+=', <a href="assistance.html#%s" target="_blank">assist</a>'%wfn
text+=' : %s '%(wf.status)
if view and wfs!='acquired':
text+='<a href="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" target="_blank"><img src="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" style="height:50px"></a>'%(wfn.replace('_','/'),wfn.replace('_','/'))
if ongoing:
text+='<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a>'%(wfn,wfn)
if ongoing:
date1 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime(time.mktime(time.gmtime())-(15*24*60*60)) )
date2 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime())
text+='<a href="http://dashb-cms-job.cern.ch/dashboard/templates/web-job2/#table=Jobs&date1=%s&date2=%s&sortby=site&task=wmagent_%s">dashb</a>'%( date1, date2, wfn )
if ongoing and wfn in boost:
for task in boost[wfn]:
overflow = boost[wfn][task].get('ReplaceSiteWhitelist',None)
if not overflow:
overflow = boost[wfn][task].get('AddWhitelist',None)
if overflow:
text+=',boost (<a href=equalizor.json>%d</a>)'%len(overflow)
#text+="<hr>"
return text
def phl(phid):
text=', '.join([
str(phid),
'<a href="https://cmsweb.cern.ch/phedex/prod/Request::View?request=%s" target="_blank">vw</a>'%phid,
'<a href="https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions?reqfilter=%s" target="_blank">sub</a>'%phid,
])
return text
def ol(out):
return '<a href="https://cmsweb.cern.ch/das/request?input=%s" target="_blank"> %s</a>'%(out,out)
def lap( comment ):
l = time.mktime(time.gmtime())
spend = l-lap.start
lap.start =l
print "Spend %d [s] for %s"%( spend, comment )
lap.start = time.mktime(time.gmtime())
## start to write it
#html_doc = open('/afs/cern.ch/user/v/vlimant/public/ops/index.html','w')
html_doc = open('%s/index.html.new'%monitor_dir,'w')
print "Updating the status page ..."
UC = unifiedConfiguration()
if not caller:
try:
#caller = sys._getframe(1).f_code.co_name
caller = sys.argv[0].split('/')[-1].replace('.py','')
print "caller is"
print caller
except Exception as es:
caller = 'none found'
print "not getting frame"
print str(es)
html_doc.write("""
<html>
<head>
<META HTTP-EQUIV="refresh" CONTENT="900">
<script type="text/javascript">
function showhide(id) {
var e = document.getElementById(id);
e.style.display = (e.style.display == 'block') ? 'none' : 'block';
}
</script>
</head>
<body>
Last update on %s(CET), %s(GMT)
<br>
<a href=logs/ target=_blank>logs</a> <a href=logs/last.log target=_blank>last</a> <a href=statuses.html>statuses</a> <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/ target=_blank>prod mon</a> <a href=https://%s/wmstats/index.html target=_blank>wmstats</a> <a href=http://t3serv001.mit.edu/~cmsprod/IntelROCCS/Detox/SitesInfo.txt target=_blank>detox</a> <a href=locked.html>space</a> <a href=logs/subscribor/last.log target=_blank>blocks</a> <a href=logs/agents/last.log>agents</a>
<br>
<a href=https://twiki.cern.ch/twiki/bin/view/CMSPublic/CompOpsWorkflowL3Responsibilities#Automatic_Assignment_and_Unified>what am I</a> <a href=data.html>json interfaces</a> <a href=logs/addHoc/last.log>add-hoc op</a> created from <b>%s <a href=logs/last_running>last running</a></b> <object height=20 type="text/html" data="logs/last_running"><p>backup content</p></object>
<br><br>
""" %(time.asctime(time.localtime()),
time.asctime(time.gmtime()),
reqmgr_url,
caller
)
)
text=""
count=0
count_by_campaign=defaultdict(lambda : defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status.startswith('considered')).all():
wl = getWL( wf.name )
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])]+=1
#print wf.name
text+="<li> %s </li> \n"%wfl(wf,p=True)
count+=1
text_by_c=""
for c in count_by_campaign:
text_by_c+="<li> %s (%d) : "%( c, sum(count_by_campaign[c].values()) )
for p in sorted(count_by_campaign[c].keys()):
text_by_c+="%d (%d), "%(p,count_by_campaign[c][p])
text_by_c+="</li>"
html_doc.write("""
Worflow next to handle (%d) <a href=https://cms-pdmv.cern.ch/mcm/batches?status=new&page=-1 target="_blank"> batches</a> <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/transferor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('considered')">[Click to show/hide]</a>
<br>
<div id="considered" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('considered_bywf')">[Click to show/hide]</a><div id="considered_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('considered_bycamp')">[Click to show/hide]</a><div id="considered_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""%(count,
count, text,
len(count_by_campaign), text_by_c))
lap( 'done with considered' )
text=""
count=0
count_by_campaign=defaultdict(lambda : defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status=='staging').all():
wl = getWL( wf.name )
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])]+=1
text+="<li> %s </li> \n"%wfl(wf,within=True)
count+=1
text_by_c=""
for c in count_by_campaign:
text_by_c+="<li> %s (%d) : "%( c, sum(count_by_campaign[c].values()) )
for p in sorted(count_by_campaign[c].keys()):
text_by_c+="%d (%d), "%(p,count_by_campaign[c][p])
text_by_c+="</li>"
html_doc.write("""
Worflow waiting in staging (%d) <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staging')">[Click to show/hide]</a>
<br>
<div id="staging" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staging_bywf')">[Click to show/hide]</a><div id="staging_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staging_bycamp')">[Click to show/hide]</a><div id="staging_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""%(count,
count, text,
len(count_by_campaign), text_by_c))
lap ( 'done with staging' )
text="<ul>"
count=0
transfer_per_wf = defaultdict(list)
for ts in session.query(Transfer).filter(Transfer.phedexid>0).all():
hide = True
t_count = 0
stext=""
for pid in ts.workflows_id:
w = session.query(Workflow).get(pid)
hide &= (w.status != 'staging' )
if w.status in ['considered','staging','staged']:
stext += "<li> %s </li>\n"%( wfl(w,status=True))
transfer_per_wf[w].append( ts.phedexid )
t_count +=1
stext = '<li> %s serves %d workflows<br><a href="javascript:showhide(\'%s\')">[show/hide]</a> <div id="%s" style="display:none;"><ul>\n'%( phl(ts.phedexid), t_count, ts.phedexid, ts.phedexid ) + stext
stext+="</ul></li>\n"
if hide:
#text+="<li> %s not needed anymore to start running (does not mean it went through completely)</li>"%phl(ts.phedexid)
pass
else:
count+=1
text+=stext
text+="</ul>"
text_bywf="<ul>"
for wf in transfer_per_wf:
text_bywf += "<li> %s </li>"%(wfl(wf,within=True))
text_bywf += '<a href=javascript:showhide("transfer_%s")>[Click to show/hide] %d transfers</a>'% (wf.name, len(transfer_per_wf[wf]))
text_bywf += '<div id="transfer_%s" style="display:none;">'% wf.name
text_bywf += "<ul>"
for pid in sorted(transfer_per_wf[wf]):
text_bywf += "<li> %s </li>"%(phl(pid))
text_bywf += "</ul></div><hr>"
text_bywf += '</ul>'
html_doc.write("""
Transfer on-going (%d) <a href=http://cmstransferteam.web.cern.ch/cmstransferteam/ target=_blank>dashboard</a> <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('transfer')">[Click to show/hide]</a>
<br>
<div id="transfer" style="display:none;">
<ul>
<li> By Workflow
<a href="javascript:showhide('transfer_bywf')">[Click to show/hide]</a>
<div id="transfer_bywf" style="display:none;">
%s
</div>
</li>
<li> By transfer request
<a href="javascript:showhide('transfer_byreq')">[Click to show/hide]</a>
<div id="transfer_byreq" style="display:none;">
%s
</div>
</li>
</ul>
</div>
"""%(count,
text_bywf,
text))
lap( 'done with transfers' )
text=""
count=0
count_by_campaign=defaultdict(lambda : defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status=='staged').all():
wl = getWL( wf.name )
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])]+=1
text+="<li> %s </li> \n"%wfl(wf,p=True)
count+=1
text_by_c=""
for c in count_by_campaign:
text_by_c+="<li> %s (%d) : "%( c, sum(count_by_campaign[c].values()) )
for p in sorted(count_by_campaign[c].keys()):
text_by_c+="%d (%d), "%(p,count_by_campaign[c][p])
text_by_c+="</li>"
html_doc.write("""Worflow ready for assigning (%d) <a href=logs/stagor/last.log target=_blank>log</a> <a href=logs/assignor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staged')">[Click to show/hide]</a>
<br>
<div id="staged" style="display:none;">
<br>
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staged_bywf')">[Click to show/hide]</a><div id="staged_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staged_bycamp')">[Click to show/hide]</a><div id="staged_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""%(count,
count, text,
len(count_by_campaign), text_by_c))
lap( 'done with staged' )
lines=[]
count_by_campaign=defaultdict(lambda : defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status=='away').all():
wl = getWL( wf.name )
count_by_campaign[wl['Campaign']][int(wl['RequestPriority'])]+=1
lines.append("<li> %s <hr></li>"%wfl(wf,view=True,ongoing=True))
text_by_c=""
for c in count_by_campaign:
text_by_c+="<li> %s (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/campaign.php?campaign=%s>mon</a> <a href=https://cms-pdmv.cern.ch/pmp/historical?r=%s target=_blank>pmp</a> "%( c, sum(count_by_campaign[c].values()),c,c )
for p in sorted(count_by_campaign[c].keys()):
text_by_c+="%d (%d), "%(p,count_by_campaign[c][p])
text_by_c+="</li>"
lines.sort()
html_doc.write("""
Worflow on-going (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests_in_production.php target=_blank>ongoing</a> <a href=https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=summary target=_blank>elog</a> <a href=http://cms-gwmsmon.cern.ch/prodview target=_blank>queues</a> <a href=logs/assignor/last.log target=_blank>log</a> <a href=logs/checkor/last.log target=_blank>postlog</a> <a href=logs/equalizor/last.log target=_blank>equ</a> <a href=logs/completor/last.log target=_blank>comp</a>
<a href="javascript:showhide('away')">[Click to show/hide]</a>
<br>
<div id="away" style="display:none;">
<ul>
<li>By workflow (%d) </li>
<a href="javascript:showhide('away_bywf')">[Click to show/hide]</a><div id="away_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('away_bycamp')">[Click to show/hide]</a><div id="away_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""%(len(lines),
len(lines),
'\n'.join(lines),
len(count_by_campaign),
text_by_c
))
lap ( 'done with away' )
text=""
count=0
#for wf in session.query(Workflow).filter(Workflow.status == 'assistance-custodial').all():
for wf in session.query(Workflow).filter(Workflow.status.startswith('assistance')).filter(Workflow.status.contains('custodial')).all():
text+="<li> %s </li> \n"%wfl(wf,view=True,update=True,status=True)
count+=1
text+="</ul></div>\n"
html_doc.write("""Worflow that are closing (%d)
<a href=closeout.html target=_blank>closeout</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('closing')">[Click to show/hide]</a>
<br>
<div id="closing" style="display:none;">
<br>
<ul>
"""%count)
html_doc.write(text)
lap ( 'done with closing' )
assistance_by_type = defaultdict(list)
text=""
count=0
for wf in session.query(Workflow).filter(Workflow.status.startswith('assistance-')).all():
assistance_by_type[wf.status].append( wf )
count+=1
for assistance_type in assistance_by_type:
text += "<li> %s (%d) <a href=\"javascript:showhide('%s')\">[Click to show/hide]</a><br><div id=\"%s\" style=\"display:none;\"><ul>"%( assistance_type,
len(assistance_by_type[assistance_type]),
assistance_type,
assistance_type,
)
for wf in assistance_by_type[assistance_type]:
text+="<li> %s <hr></li> \n"%wfl(wf,view=True,within=True,status=True,update=True)
text += "</ul></div></li>\n"
html_doc.write("""Worflow which need assistance (%d)
<a href=assistance.html target=_blank>assistance</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/recoveror/last.log target=_blank>postlog</a>
<a href="javascript:showhide('assistance')">[Click to show/hide]</a>
<br>
<div id="assistance" style="display:none;">
<br>
<ul>
%s
</ul>
</div>
"""%(count, text))
lap ( 'done with assistance' )
text=""
count=0
for wf in session.query(Workflow).filter(Workflow.status == 'close').all():
text+="<li> %s </li> \n"%wfl(wf)
count+=1
text+="</ul></div>\n"
html_doc.write("""Worflow ready to close (%d)
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('close')">[Click to show/hide]</a>
<br>
<div id="close" style="display:none;">
<br>
<ul>
"""%count)
html_doc.write(text)
lap ( 'done with annoucing' )
text=""
count=0
for wf in session.query(Workflow).filter(Workflow.status=='trouble').all():
text+="<li> %s </li> \n"%wfl(wf)
count+=1
text+="</ul></div>\n"
html_doc.write("""Worflow with issue (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/injector/last.log target=_blank>postlog</a>
<a href="javascript:showhide('trouble')">[Click to show/hide]</a>
<br>
<div id="trouble" style="display:none;">
<br>
<ul>
"""%count)
html_doc.write(text)
lap ( 'done with trouble' )
text=""
count=0
for wf in session.query(Workflow).filter(Workflow.status=='forget').all():
text+="<li> %s </li> \n"%wfl(wf)
count+=1
text+="</ul></div>\n"
html_doc.write("""
Worflow to forget (%d) <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/lockor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('forget')">[Click to show/hide]</a>
<br>
<div id="forget" style="display:none;">
<br>
<ul>
"""%count)
html_doc.write(text)
lap ( 'done with forget' )
text=""
count=0
for wf in session.query(Workflow).filter(Workflow.status=='done').all():
text+="<li> %s </li> \n"%wfl(wf)#,ms=True)
count+=1
text+="</ul></div>\n"
html_doc.write("""
Worflow through (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/lockor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('done')">[Click to show/hide]</a>
<br>
<div id="done" style="display:none;">
<br>
<ul>
"""%count)
html_doc.write(text)
lap ( 'done with done' )
wfs = session.query(Workflow).filter(Workflow.status.endswith('-unlock')).all()
html_doc.write(" Workflows unlocked : %s <a href=logs/lockor/last.log target=_blank>log</a><br>"%(len(wfs)))
lap ( 'done with unlocked' )
text=""
lines_thisweek=[]
lines_lastweek=[]
now = time.mktime(time.gmtime())
this_week = int(time.strftime("%W",time.gmtime()))
start_time_two_weeks_ago = time.mktime(time.gmtime(now - (20*24*60*60))) # 20
last_week = int(time.strftime("%W",time.gmtime(now - ( 7*24*60*60))))
all_locks = json.loads(open('%s/globallocks.json'%monitor_dir).read())
waiting_custodial = json.loads(open('%s/waiting_custodial.json'%monitor_dir).read())
all_pending_approval_custodial = dict([(k,item) for k,item in waiting_custodial.items() if 'nodes' in item and not any([node['decided'] for node in item['nodes'].values()]) ])
n_pending_approval = len( all_pending_approval_custodial )
#n_pending_approval = len([item for item in waiting_custodial.values() if 'nodes' in item and not any([node['decided'] for node in item['nodes'].values() ])])
missing_approval_custodial = json.loads(open('%s/missing_approval_custodial.json'%monitor_dir).read())
stuck_custudial = json.loads(open('%s/stuck_custodial.json'%monitor_dir).read())
lagging_custudial = json.loads(open('%s/lagging_custodial.json'%monitor_dir).read())
if len(stuck_custudial):
stuck_string = ', <font color=red>%d appear to be <a href=stuck_custodial.json>stuck</a></font>'% len(stuck_custudial)
else:
stuck_string = ''
if len(missing_approval_custodial):
long_approve_string = ', <font color=red>%d more than %d days</font>'%( len(missing_approval_custodial), UC.get('transfer_timeout'))
else:
long_approve_string = ''
output_within_two_weeks=session.query(Output).filter(Output.date>=start_time_two_weeks_ago).all()
waiting_custodial_string=""
waiting_custodial_strings=[]
for ds in waiting_custodial:
out = None
## lots of it will be within two weeks
of = filter(lambda odb: odb.datasetname == ds, output_within_two_weeks)
if of:
out = of[0]
else:
out = session.query(Output).filter(Output.datasetname == ds).first()
if out:
info = waiting_custodial[out.datasetname]
action = 'going'
if out.datasetname in all_pending_approval_custodial:
action = '<font color=red>pending</font>'
try:
size = str(info['size'])
except:
size = "x"
destination = ",".join(info['nodes'].keys())
if not destination:
destination ='<font color=red>NO SITE</font>'
a_waiting_custodial_string = "<li>on week %s : %s %s</li>"%(
time.strftime("%W (%x %X)",time.gmtime(out.date)),
ol(out.datasetname),
' %s [GB] %s to %s on %s (<a href="https://cmsweb.cern.ch/phedex/datasvc/xml/prod/requestlist?dataset=%s&node=T*MSS">%d missing</a>)'%( size, action, destination, time.asctime(time.gmtime(info['checked'])), out.datasetname, info['nmissing'])
)
waiting_custodial_strings.append( (out.date, a_waiting_custodial_string) )
waiting_custodial_strings.sort( key = lambda i:i[0] )
waiting_custodial_string="\n".join( [i[1] for i in waiting_custodial_strings] )
#start_time_two_weeks_ago = time.mktime(time.strptime("15-0-%d"%(this_week-2), "%y-%w-%W"))
for out in output_within_two_weeks:
if not out.workflow:
print "This is a problem with",out.datasetname
continue
if out.workflow.status in ['done-unlock','done','clean','clean-out','clean-unlock']:
custodial=''
if out.datasetname in waiting_custodial:
info = waiting_custodial[out.datasetname]
try:
try:
size = str(info['size'])
except:
size = "x"
destination = ",".join(info['nodes'].keys())
if not destination:
destination ='<font color=red>NO SITE</font>'
action = 'going'
if out.datasetname in all_pending_approval_custodial:
action = '<font color=red>pending</font>'
custodial=' %s [GB] %s to %s on %s (<a href="https://cmsweb.cern.ch/phedex/datasvc/xml/prod/requestlist?dataset=%s&node=T*MSS">%d missing</a>)'%( size, action, destination, time.asctime(time.gmtime(info['checked'])), out.datasetname, info['nmissing'])
except Exception as e:
#print info
#print str(e)
pass
elif out.datasetname in all_locks:
custodial='<font color=green>LOCKED</font>'
out_week = int(time.strftime("%W",time.gmtime(out.date)))
##only show current week, and the previous.
if last_week==out_week:
lines_lastweek.append("<li>on week %s : %s %s</li>"%(
time.strftime("%W (%x %X)",time.gmtime(out.date)),
ol(out.datasetname),
custodial
)
)
if this_week==out_week:
lines_thisweek.append("<li>on week %s : %s %s</li>"%(
time.strftime("%W (%x %X)",time.gmtime(out.date)),
ol(out.datasetname),
custodial
)
)
lines_thisweek.sort()
lines_lastweek.sort()
html_doc.write("""Output produced (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?in_disagreement=1 target=_blank>disagreements</a>
<a href="javascript:showhide('output')">[Click to show/hide]</a>
<br>
<div id="output" style="display:none;">
<br>
<ul>
<li> %d waiting to go to tape</li>
<ul>
<li> %d waiting for tape approval%s</li>
<li> %d are not completed after %d days%s</li>
<li> Full list (%d) <a href="javascript:showhide('waiting-custodial')">[Click to show/hide]</a>
<div id="waiting-custodial" style="display:none;">
<ul>
%s
</ul>
</div>
</li>
</ul>
<li> Last week (%d) </li><a href="javascript:showhide('output_lastweek')">[Click to show/hide]</a><div id="output_lastweek" style="display:none;"><ul>
%s
</ul></div>
<li> This week (%d) </li><a href="javascript:showhide('output_thisweek')">[Click to show/hide]</a><div id="output_thisweek" style="display:none;"><ul>
%s
</ul></div></div>
"""%( len(lines_lastweek)+len(lines_thisweek),
len(waiting_custodial),
n_pending_approval,long_approve_string,
len(lagging_custudial),UC.get('transfer_timeout'),stuck_string,
len(waiting_custodial),waiting_custodial_string,
len(lines_lastweek),
'\n'.join(lines_lastweek),
len(lines_thisweek),
'\n'.join(lines_thisweek))
)
lap ( 'done with output' )
html_doc.write("""Job installed
<a href="javascript:showhide('acron')">[Click to show/hide]</a>
<br>
<div id="acron" style="display:none;">
<br>
<pre>
%s
</pre>
"""%(os.popen('acrontab -l | grep Unified | grep -v \#').read()))
per_module = defaultdict(list)
for t in filter(None,os.popen('cat %s/logs/*/*.time'%monitor_dir).read().split('\n')):
module_name,run_time,spend = t.split(':')
## then do what you want with it !
if 'cleanor' in module_name: continue
per_module[module_name].append( int(spend) )
def display_time( sec ):
m, s = divmod(sec, 60)
h, m = divmod(m, 60)
dis=""
if h:
dis += "%d [h] "%h
if h or m:
dis += "%d [m] "%m
if h or m or s:
dis += "%d [s]"%s
return dis
html_doc.write("Module running time<ul>\n")
for m,spends in per_module.items():
avg = sum(spends)/float(len(spends))
lasttime = spends[-1]
html_doc.write("<li>%s : last %s, avg %s</li>\n"%( m, display_time(lasttime), display_time(avg)))
html_doc.write("</ul>")
html_doc.write("Last running <pre>%s</pre><br>"%( os.popen("tac %s/logs/running | head -5"%monitor_dir).read() ))
html_doc.write("Order in cycle <pre>%s</pre><br>"%( '\n'.join(map(lambda l : l.split('/')[-1].replace('.py',''), filter(lambda l : not l.startswith('#') and 'Unified' in l and 'py' in l.split('/')[-1], open('%s/WmAgentScripts/cycle.sh'%base_dir).read().split('\n')))) ))
html_doc.write("</div>\n")
lap ( 'done with jobs' )
text=""
count=0
for (c,info) in campaignInfo().campaigns.items():
#if 'go' in info and info['go']:
text+="<li>%s <br> <pre>%s</pre> </li>"%( c, json.dumps( info, indent=2))
count+=1
html_doc.write("""Campaign configuration
<a href="javascript:showhide('campaign')">[Click to show/hide]</a>
<br>
<div id="campaign" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""%(text))
text=""
count=0
n_column = 4
SI = siteInfo()
date1 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime(time.mktime(time.gmtime())-(15*24*60*60)) ) ## 15 days
date2 = time.strftime('%Y-%m-%d+%H:%M', time.gmtime())
for t in SI.types():
text+="<li>%s<table border=1>"%t
c=0
for site in getattr(SI,t):
cpu = SI.cpu_pledges[site] if site in SI.cpu_pledges else 'N/A'
disk = SI.disk[SI.CE_to_SE(site)] if SI.CE_to_SE(site) in SI.disk else 'N/A'
if c==0:
text+="<tr>"
if not disk:
ht_disk = '<font color=red>Disk available: %s</font>'%disk
else:
ht_disk = 'Disk available: %s'%disk
text+='<td><a href=http://dashb-ssb.cern.ch/dashboard/templates/sitePendingRunningJobs.html?site=%s>%s</a><br><a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a><br><a href="http://dashb-cms-job.cern.ch/dashboard/templates/web-job2/#user=&refresh=0&table=Jobs&p=1&records=25&activemenu=1&site=%s&submissiontool=wmagent&check=submitted&sortby=activity&scale=linear&bars=20&data1=%s&date2=%s">dashb</a><br>CPU pledge: %s<br>%s</td>'%(site,site,site,site,site,date1,date2,cpu,ht_disk)
if c==n_column:
c=0
else:
c+=1
text+="</table></li>"
text += "<li> Sites in auto-approved transfer<ul>"
for site in sorted(SI.sites_auto_approve):
text+="<li>%s"% site
text += "</ul></li>"
text += "<li> Sites with vetoe transfer<ul>"
for site in sorted(SI.sites_veto_transfer):
text+="<li>%s"% site
text += "</ul></li>"
text += "<li> Sites banned from production<ul>"
for site in sorted(SI.sites_banned):
text+="<li>%s"% site
text += "</ul></li>"
text += "<li> Approximate Free Tape<ul>"
for mss in SI.storage:
waiting = 0
try:
waiting = float(os.popen("grep '%s is pending . Created since' %s/logs/lockor/last.log -B 3 | grep size | awk '{ sum+=$6 ; print sum }' | tail -1" % (mss,monitor_dir)).readline())
except Exception as e:
print str(e)
oldest = ""
os.system('grep pending %s/logs/lockor/last.log | sort -u > %s/logs/pending.log'%(monitor_dir,monitor_dir))
try:
oldest = os.popen("grep '%s is pending . Created since ' %s/logs/lockor/last.log | sort | awk '{print $10, $11, $12, $13, $14 }' | head -1"% (mss,monitor_dir)).readline()
except Exception as e:
print str(e)
waiting /= 1024.
text+="<li>%s : %d [TB]. Waiting for approval %d [TB] since %s </li>"%(mss, SI.storage[mss], waiting, oldest)
text += "</ul></li>"
lap ( 'done with sites' )
open('%s/siteInfo.json'%monitor_dir,'w').write(json.dumps(dict([(t,getattr(SI,t)) for t in SI.types()]),indent=2))
lap ( 'done with sites json' )
chart_data = defaultdict(list)
for site in SI.quota:
chart_data[site].append("""
var data_%s = google.visualization.arrayToDataTable([
['Overall', 'Space in TB'],
//['Quota' , %s],
['Locked' , %s],
['Free' , %s]
]);
"""%( site,
SI.quota[site], SI.locked[site], SI.disk[site],
))
chart_data[site].append("""
var chart_%s = new google.visualization.PieChart(document.getElementById('donutchart_%s'));
chart_%s.draw(data_%s, {title: '%s %s [TB]', pieHole:0.4, slices:{0:{color:'red'},1:{color:'green'}}});
"""%(site,site,
site,site,
site,SI.quota[site]))
chart_data[site].append("""
<div id="donutchart_%s" style="height: 200px;width: 300px"></div>
"""%(site))
## make the locked/available donut chart
donut_html = open('%s/locked.html'%monitor_dir,'w')
tables = "\n".join([info[0] for site,info in chart_data.items()])
draws = "\n".join([info[1] for site,info in chart_data.items()])
divs = "\n".join([info[2] for site,info in chart_data.items()])
divs_table="<table border=0>"
for c,site in enumerate(sorted(chart_data.keys())):
if c%5==0:
divs_table += "<tr>"
divs_table += "<td>%s</td>"%(chart_data[site][2])
divs_table += "</table>"
donut_html.write("""
<html>
<head>
<script type="text/javascript" src="https://www.google.com/jsapi"></script>
<script type="text/javascript">
google.load("visualization", "1", {packages:["corechart"]});
google.setOnLoadCallback(drawChart);
function drawChart() {
%s
%s
}
</script>
</head>
<body>
%s
</body>
</html>
"""%( tables,draws,divs_table )
)
donut_html.close()
html_doc.write("""Site configuration
<a href="javascript:showhide('site')">[Click to show/hide]</a>
<br>
<div id="site" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""%(text))
lap ( 'done with space' )
text = ""
for param in UC.configs:
text +="<li>%s</li><ul>\n"% param
for sub in sorted(UC.configs[param].keys()):
text +="<li> %s : %s </li>\n"%( sub, UC.configs[param][sub] )
text += '</ul>\n'
html_doc.write("""Unified configuration
<a href="javascript:showhide('config')">[Click to show/hide]</a>
<br>
<div id="config" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""%(text))
lap ( 'done with configuration' )
print "... done with status page."
html_doc.write("""
</body>
</html>
""")
html_doc.close()
## and put the file in place
os.system('mv %s/index.html.new %s/index.html'%(monitor_dir,monitor_dir))
statuses = json.loads(open('%s/statusmon.json'%monitor_dir).read())
s_count = defaultdict(int)
now = time.mktime(time.gmtime())
for wf in session.query(Workflow).all():
s_count[wf.status]+=1
statuses[now] = dict( s_count )
## remove old entries
for t in statuses.keys():
if (now-float(t)) > 7*24*60*60:
statuses.pop(t)
open('%s/statusmon.json'%monitor_dir,'w').write( json.dumps( statuses , indent=2))
html_doc = open('%s/statuses.html'%monitor_dir,'w')
html_doc.write(""" <html>
<table border=1>
<thead>
<tr>
<th> workflow </th><th> status </th><th> wm status</th>
</tr>
</thead>
""")
wfs = {}
for wfo in session.query(Workflow).all():
## pass all that is unlocked and considered it gone
wfs[wfo.name] = (wfo.status,wfo.wm_status)
open('%s/statuses.json'%monitor_dir,'w').write(json.dumps( wfs ))
for wfn in sorted(wfs.keys()):
## pass all that is unlocked and considered it gone
if 'unlock' in wfs[wfn][0]: continue
html_doc.write('<tr><td><a id="%s">%s</a></td><td>%s</td><td>%s</td></tr>\n'%( wfn, wfn, wfs[wfn][0], wfs[wfn][1]))
html_doc.write("</table>")
html_doc.write("<br>"*100)
html_doc.write("end of page</html>")
html_doc.close()
def stagor(url,specific =None, options=None):
if not componentInfo().check(): return
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
done_by_wf_id = {}
done_by_input = {}
completion_by_input = {}
good_enough = 100.0
lost_blocks = json.loads(open('%s/lost_blocks_datasets.json'%monitor_dir).read())
lost_files = json.loads(open('%s/lost_files_datasets.json'%monitor_dir).read())
known_lost_blocks = {}
known_lost_files = {}
for dataset in set(lost_blocks.keys()+lost_files.keys()):
b,f = findLostBlocksFiles(url, dataset)
if dataset in lost_blocks and not b:
print dataset,"has no really lost blocks"
else:
known_lost_blocks[dataset] = [i['name'] for i in b]
if dataset in lost_files and not f:
print dataset,"has no really lost files"
else:
known_lost_files[dataset] = [i['name'] for i in f]
try:
cached_transfer_statuses = json.loads(open('cached_transfer_statuses.json').read())
except:
print "inexisting transfer statuses. starting fresh"
cached_transfer_statuses = {}
transfer_statuses = {}
## pop all that are now in negative values
for phedexid in cached_transfer_statuses.keys():
transfers = session.query(Transfer).filter(Transfer.phedexid==int(phedexid)).all()
if not transfers:
print phedexid,"does not look relevant to be in cache anymore. poping"
print cached_transfer_statuses.pop( phedexid )
## collect all datasets that are needed for wf in staging, correcting the status of those that are not really in staging
wfois = []
needs = defaultdict(list)
for wfo in session.query(Workflow).filter(Workflow.status == 'staging').all():
wfi = workflowInfo(url, wfo.name)
if wfi.request['RequestStatus'] in ['running-open','running-closed','completed','assigned','acquired']:
wfi.sendLog('stagor', "is in status %s"%wfi.request['RequestStatus'])
wfi.status='away'
session.commit()
continue
if not wfi.request['RequestStatus'] in ['assignment-approved']:
## should be setting 'away' too
print wfo.name,"is",wfi.request['RequestStatus']
sendEmail("wrong status in staging. debug","%s is in %s, should set away."%(wfo.name,wfi.request['RequestStatus']))
wfois.append( (wfo,wfi) )
_,primaries,_,secondaries = wfi.getIO()
for dataset in list(primaries)+list(secondaries):
needs[wfo.name].append( dataset)
done_by_input[dataset] = {}
completion_by_input[dataset] = {}
wfi.sendLog('stagor', '%s needs %s'%( wfo.name, dataset))
open('%s/dataset_requirements.json'%monitor_dir,'w').write( json.dumps( needs, indent=2))
dataset_endpoints = defaultdict(set)
endpoint_in_downtime = defaultdict(set)
#endpoint_completed = defaultdict(set)
endpoint_incompleted = defaultdict(set)
#endpoint = defaultdict(set)
send_back_to_considered = set()
## phedexid are set negative when not relevant anymore
# probably there is a db schema that would allow much faster and simpler query
for transfer in session.query(Transfer).filter(Transfer.phedexid>0).all():
if specific and str(transfer.phedexid)!=str(specific): continue
skip=True
for wfid in transfer.workflows_id:
tr_wf = session.query(Workflow).get(wfid)
if tr_wf:
if tr_wf.status == 'staging':
sendLog('stagor',"\t%s is staging for %s"%(transfer.phedexid, tr_wf.name))
skip=False
if skip:
sendLog('stagor',"setting %s to negative value"%transfer.phedexid)
transfer.phedexid = -transfer.phedexid
session.commit()
continue
if transfer.phedexid<0: continue
## check the status of transfers
checks = checkTransferApproval(url, transfer.phedexid)
approved = all(checks.values())
if not approved:
sendLog('stagor', "%s is not yet approved"%transfer.phedexid)
approveSubscription(url, transfer.phedexid)
continue
## check on transfer completion
if str(transfer.phedexid) in cached_transfer_statuses:
### use a cache for transfer that already looked done
sendLog('stagor',"read %s from cache"%transfer.phedexid)
checks = cached_transfer_statuses[str(transfer.phedexid)]
else:
checks = checkTransferStatus(url, transfer.phedexid, nocollapse=True)
## just write this out
transfer_statuses[str(transfer.phedexid)] = copy.deepcopy(checks)
if not specific:
for dsname in checks:
if not dsname in done_by_input: done_by_input[dsname]={}
if not dsname in completion_by_input: completion_by_input[dsname] = {}
done_by_input[dsname][transfer.phedexid]=all(map(lambda i:i>=good_enough, checks[dsname].values()))
completion_by_input[dsname][transfer.phedexid]=checks[dsname].values()
if checks:
sendLog('stagor',"Checks for %s are %s"%( transfer.phedexid, [node.values() for node in checks.values()]))
done = all(map(lambda i:i>=good_enough,list(itertools.chain.from_iterable([node.values() for node in checks.values()]))))
else:
## it is empty, is that a sign that all is done and away ?
print "ERROR with the scubscriptions API of ",transfer.phedexid
print "Most likely something else is overiding the transfer request. Need to work on finding the replacement automatically, if the replacement exists"
done = False
## the thing above is NOT giving the right number
#done = False
for wfid in transfer.workflows_id:
tr_wf = session.query(Workflow).get(wfid)
if tr_wf:# and tr_wf.status == 'staging':
if not tr_wf.id in done_by_wf_id: done_by_wf_id[tr_wf.id]={}
done_by_wf_id[tr_wf.id][transfer.phedexid]=done
## for those that are in staging, and the destination site is in drain
#if not done and tr_wf.status == 'staging':
for ds in checks:
for s,v in checks[ds].items():
dataset_endpoints[ds].add( s )
if done:
## transfer.status = 'done'
sendLog('stagor',"%s is done"%transfer.phedexid)
cached_transfer_statuses[str(transfer.phedexid)] = copy.deepcopy(checks)
else:
sendLog('stagor',"%s is not finished %s"%(transfer.phedexid, pprint.pformat( checks )))
pprint.pprint( checks )
## check if the destination is in down-time
for ds in checks:
sites_incomplete = [SI.SE_to_CE(s) for s,v in checks[ds].items() if v<good_enough]
sites_incomplete_down = [s for s in sites_incomplete if not s in SI.sites_ready]
if sites_incomplete_down:
sendLog('stagor',"%s are in downtime, while waiting for %s to get there"%( ",".join(sites_incomplete_down), ds))
#sites_complete = [SI.SE_to_CE(s) for s,v in checks[ds].items() if v>=good_enough]
#endpoint[ds].update( sites_complete )
#endpoint[ds].update( sites_incomplete )
#endpoint_completed[ds].update( sites_complete )
endpoint_incompleted[ds].update( sites_incomplete )
endpoint_in_downtime[ds].update( sites_incomplete_down )
print "End point in down time"
for k in endpoint_in_downtime: endpoint_in_downtime[k] = list(endpoint_in_downtime[k])
for k in dataset_endpoints: dataset_endpoints[k] = list(dataset_endpoints[k])
print json.dumps( endpoint_in_downtime , indent=2)
open('cached_transfer_statuses.json','w').write( json.dumps( cached_transfer_statuses, indent=2))
open('%s/transfer_statuses.json'%monitor_dir,'w').write( json.dumps( transfer_statuses, indent=2))
open('%s/dataset_endpoints.json'%monitor_dir,'w').write( json.dumps(dataset_endpoints, indent=2))
already_stuck = json.loads( open('%s/stuck_transfers.json'%monitor_dir).read() )
missing_in_action = defaultdict(list)
print "-"*10,"Checking on workflows in staging","-"*10
#forget_about = ['/MinBias_TuneCUETP8M1_13TeV-pythia8/RunIISummer15GS-MCRUN2_71_V1-v2/GEN-SIM']
#for what in forget_about:
# if not done_by_input[what]:
# done_by_input[what] = {'fake':True}
## come back to workflows and check if they can go
available_cache = defaultdict(lambda : defaultdict(float))
presence_cache = defaultdict(dict)
for wfo,wfi in wfois:
print "#"*30
## the site white list takes site, campaign, memory and core information
(_,primaries,_,secondaries,sites_allowed) = wfi.getSiteWhiteList(verbose=False)
se_allowed = [SI.CE_to_SE(site) for site in sites_allowed]
se_allowed.sort()
se_allowed_key = ','.join(se_allowed)
readys={}
for need in list(primaries)+list(secondaries):
if not need in done_by_input:
wfi.sendLog('stagor',"missing transfer report for %s"%need)
readys[need] = False
## should warn someone about this !!!
## it cannot happen, by construction
sendEmail('missing transfer report','%s does not have a transfer report'%(need))
continue
if not done_by_input[need] and need in list(secondaries):
wfi.sendLog('stagor',"assuming it is OK for secondary %s to have no attached transfers"% need)
readys[need] = True
done_by_input[need] = { "fake" : True }
continue
if len(done_by_input[need]) and all(done_by_input[need].values()):
wfi.sendLog('stagor',"%s is ready"%need)
print json.dumps( done_by_input[need] , indent=2)
readys[need] = True
else:
wfi.sendLog('stagor',"%s is not ready"%need)
print json.dumps( done_by_input[need] , indent=2)
readys[need] = False
if readys and all(readys.values()):
if wfo.status == 'staging':
wfi.sendLog('stagor',"all needs are fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
wfi.sendLog('stagor',"all needs are fullfilled, already")
print json.dumps( readys, indent=2 )
else:
wfi.sendLog('stagor',"missing requirements")
copies_needed,_ = wfi.getNCopies()
jump_ahead = False
re_transfer = False
## there is missing input let's do something more elaborated
for need in list(primaries):#+list(secondaries):
if endpoint_in_downtime[need] == endpoint_incompleted[need]:
#print need,"is going to an end point in downtime"
wfi.sendLog('stagor',"%s has only incomplete endpoint in downtime"%need)
re_transfer=True
if not se_allowed_key in available_cache[need]:
available_cache[need][se_allowed_key] = getDatasetBlocksFraction( url , need, sites=se_allowed )
if available_cache[need][se_allowed_key] >= copies_needed:
wfi.sendLog('stagor',"assuming it is OK to move on like this already for %s"%need)
jump_ahead = True
## compute a time since staging to filter jump starting ?
# check whether the inputs is already in the stuck list ...
for need in list(primaries)+list(secondaries):
if need in already_stuck:
wfi.sendLog('stagor',"%s is stuck, so try to jump ahead"%need)
jump_ahead = True
if jump_ahead or re_transfer:
details_text = "checking on availability for %s to jump ahead"%wfo.name
details_text += '\n%s wants %s copies'%(wfo.name,copies_needed)
copies_needed = max(1,copies_needed-1)
details_text += '\nlowering by one unit to %s'%copies_needed
wfi.sendLog('stagor', details_text)
all_check = True
prim_where = set()
for need in list(primaries):
if not se_allowed_key in presence_cache[need]:
presence_cache[need][se_allowed_key] = getDatasetPresence( url, need , within_sites=se_allowed)
presence = presence_cache[need][se_allowed_key]
prim_where.update( presence.keys() )
available = available_cache[need][se_allowed_key]
this_check = (available >= copies_needed)
wfi.sendLog('stagor', "%s is available %s times %s"%( need, available, this_check))
all_check &= this_check
if not all_check: break
for need in list(secondaries):
## I do not want to check on the secon
this_check = all(done_by_input[need].values())
wfi.sendLog('stagor',"%s is all transfered %s"%(need, json.dumps(done_by_input[need], indent=2)))
all_check&= this_check
#if not se_allowed_key in presence_cache[need]:
# presence_cache[need][se_allowed_key] = getDatasetPresence( url, need , within_sites=se_allowed)
## restrict to where the primary is
#presence = dict([(k,v) for (k,v) in presence_cache[need][se_allowed_key].items() if k in prim_where])
#this_check = all([there for (there,frac) in presence.values()])
#print need,"is present at all sites:",this_check
#all_check&= this_check
if all_check:
wfi.sendLog('stagor',"needs are sufficiently fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
print wfo.name,"has to wait a bit more"
wfi.sendLog('stagor',"needs to wait a bit more")
else:
wfi.sendLog('stagor',"not checking availability")
if re_transfer:
wfi.sendLog('stagor',"Sending back to considered because of endpoint in downtime")
if wfo.status == 'staging':
wfo.status = 'considered'
session.commit()
send_back_to_considered.add( wfo.name )
if send_back_to_considered:
#sendEmail("transfer to endpoint in downtime","sending back to considered the following workflows \n%s"%('\n'.join( send_back_to_considered)))
sendLog('stagor', "sending back to considered the following workflows \n%s"%('\n'.join( send_back_to_considered)), level='critical')
print "-"*10,"Checking on non-available datasets","-"*10
## now check on those that are not fully available
for dsname in available_cache.keys():
## squash the se_allowed_key key
available_cache[dsname] = min( available_cache[dsname].values() )
for dsname,available in available_cache.items():
using_its = getWorkflowByInput(url, dsname)
#print using_its
using_wfos = []
for using_it in using_its:
wf = session.query(Workflow).filter(Workflow.name == using_it).first()
if wf:
using_wfos.append( wf )
if not len(done_by_input[dsname]):
print "For dataset",dsname,"there are no transfer report. That's an issue."
for wf in using_wfos:
if wf.status == 'staging':
if UC.get("stagor_sends_back"):
print "sending",wf.name,"back to considered"
wf.status = 'considered'
session.commit()
#sendEmail( "send back to considered","%s was send back and might be trouble"% wf.name)
sendLog('stagor', "%s was send back and might be trouble"% wf.name, level='critical')
else:
print "would send",wf.name,"back to considered"
#sendEmail( "subscription lagging behind","susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name)
sendLog('stagor', "susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name, level='critical')
continue
## not compatible with checking on secondary availability
#if all([wf.status != 'staging' for wf in using_wfos]):
# ## means despite all checks that input is not needed
# continue
if available < 1.:
print "incomplete",dsname
## there is a problem in the method below that it does not account for files stuck in T1*Buffer only
lost_blocks,lost_files = findLostBlocksFiles( url, dsname )
lost_block_names = [item['name'] for item in lost_blocks]
lost_file_names = [item['name'] for item in lost_files]
if lost_blocks:
#print json.dumps( lost , indent=2 )
## estimate for how much !
fraction_loss,_,n_missing = getDatasetBlockFraction(dsname, lost_block_names)
print "We have lost",len(lost_block_names),"blocks",lost_block_names,"for %f%%"%(100.*fraction_loss)
if fraction_loss > 0.05: ## 95% completion mark
#sendEmail('we have lost too many blocks','%s is missing %d blocks, for %d events, %f %% loss'%(dsname, len(lost_block_names), n_missing, fraction_loss))
sendLog('stagor', '%s is missing %d blocks, for %d events, %3.2f %% loss'%(dsname, len(lost_block_names), n_missing, 100*fraction_loss), level='warning')
## the workflow should be rejected !
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
#sendEmail('doomed workflow','%s has too much loss on the input dataset %s. please check on stagor logs https://cmst2.web.cern.ch/cmst2/unified/logs/stagor/last.log'%(wf.name, dsname))
sendLog('stagor', '%s has too much loss on the input dataset %s. Missing %d blocks, for %d events, %3.2f %% loss'%(wf.name, dsname, len(lost_block_names), n_missing, 100*fraction_loss), level='critical')
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_blocks:
#sendEmail('we have lost a few blocks', '%s is missing %d blocks, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_block_names), n_missing, fraction_loss, '\n'.join( lost_block_names ) ))
sendLog('stagor', '%s is missing %d blocks, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_block_names), n_missing, fraction_loss, '\n'.join( lost_block_names ) ), level='critical')
known_lost_blocks[dsname] = [i['name'] for i in lost_blocks]
if lost_files:
fraction_loss,_,n_missing = getDatasetFileFraction(dsname, lost_file_names)
print "We have lost",len(lost_file_names),"files",lost_file_names,"for %f%%"%fraction_loss
if fraction_loss > 0.05:
#sendEmail('we have lost too many files','%s is missing %d files, for %d events, %f %% loss'%(dsname, len(lost_file_names),n_missing, fraction_loss))
sendLog('stagor', '%s is missing %d files, for %d events, %f %% loss'%(dsname, len(lost_file_names),n_missing, fraction_loss), level='critical')
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_files:
#sendEmail('we have lost a few files','%s is missing %d files, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_file_names),n_missing, fraction_loss, '\n'.join(lost_file_names)))
sendLog('stagor', '%s is missing %d files, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_file_names),n_missing, fraction_loss, '\n'.join(lost_file_names)), level='critical')
known_lost_files[dsname] = [i['name'] for i in lost_files]
## should the status be change to held-staging and pending on a ticket
missings = [pid for (pid,d) in done_by_input[dsname].items() if d==False]
print "\t",done_by_input[dsname]
print "\tneeds",len(done_by_input[dsname])
print "\tgot",done_by_input[dsname].values().count(True)
print "\tmissing",missings
missing_in_action[dsname].extend( missings )
rr= open('%s/lost_blocks_datasets.json'%monitor_dir,'w')
rr.write( json.dumps( known_lost_blocks, indent=2))
rr.close()
rr= open('%s/lost_files_datasets.json'%monitor_dir,'w')
rr.write( json.dumps( known_lost_files, indent=2))
rr.close()
open('%s/incomplete_transfers.json'%monitor_dir,'w').write( json.dumps(missing_in_action, indent=2) )
print "Stuck transfers and datasets"
print json.dumps( missing_in_action, indent=2 )
print "Going further and make a report of stuck transfers"
datasets_by_phid = defaultdict(set)
for dataset in missing_in_action:
for phid in missing_in_action[dataset]:
#print dataset,"stuck through",phid
datasets_by_phid[phid].add( dataset )
bad_destinations = defaultdict(set)
bad_sources = defaultdict(set)
report = ""
really_stuck_dataset = set()
transfer_timeout = UC.get("transfer_timeout")
transfer_lowrate = UC.get("transfer_lowrate")
for phid,datasets in datasets_by_phid.items():
issues = checkTransferLag( url, phid , datasets=list(datasets) )
for dataset in issues:
for block in issues[dataset]:
for destination in issues[dataset][block]:
(block_size,destination_size,delay,rate,dones) = issues[dataset][block][destination]
## count x_Buffer and x_MSS as one source
redones=[]
for d in dones:
if d.endswith('Buffer') or d.endswith('Export'):
if d.replace('Buffer','MSS').replace('Export','MSS') in dones:
continue
else:
redones.append( d )
else:
redones.append( d )
dones = list(set( redones ))
#dones = filter(lambda s : (s.endswith('Buffer') and not s.replace('Buffer','MSS') in dones) or (not s.endswith('Buffer')) , dones)
if delay>transfer_timeout and rate<transfer_lowrate:
if len(dones)>1:
## its the destination that sucks
bad_destinations[destination].add( block )
else:
dum=[bad_sources[d].add( block ) for d in dones]
really_stuck_dataset.add( dataset )
print "add",dataset,"to really stuck"
report += "%s is not getting to %s, out of %s faster than %f [GB/s] since %f [d]\n"%(block,destination,", ".join(dones), rate, delay)
print "\n"*2
## create tickets right away ?
report+="\nbad sources "+",".join(bad_sources.keys())+"\n"
for site,blocks in bad_sources.items():
report+="\n\n%s:"%site+"\n\t".join(['']+list(blocks))
report+="\nbad destinations "+",".join(bad_destinations.keys())+"\n"
for site,blocks in bad_destinations.items():
report+="\n\n%s:"%site+"\n\t".join(['']+list(blocks))
print '\n'*2,"Datasets really stuck"
print '\n'.join( really_stuck_dataset )
print '\n'*2,"report written at https://cmst2.web.cern.ch/cmst2/unified/logs/incomplete_transfers.log"
print report
stuck_transfers = dict([(k,v) for (k,v) in missing_in_action.items() if k in really_stuck_dataset])
print '\n'*2,'Stuck dataset transfers'
print json.dumps(stuck_transfers , indent=2)
open('%s/stuck_transfers.json'%monitor_dir,'w').write( json.dumps(stuck_transfers , indent=2) )
open('%s/logs/incomplete_transfers.log'%monitor_dir,'w').write( report )
def checkor(url, spec=None, options=None):
if userLock(): return
if duplicateLock(): return
fDB = closeoutInfo()
UC = unifiedConfiguration()
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
wfs=[]
if options.new:
## get all in running and check
## you want to intersect with what is completed !
if options.strict:
completed_wfi = getWorkflows(url, status='completed')
for wfo in session.query(Workflow).filter(Workflow.status == 'away').all():
if wfo.name in completed_wfi:
wfs.append( wfo )
else:
print wfo.name,"is not completed"
sendLog('checkor','%s is not completed'%( wfo.name))
else:
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
if options.current:
## recheck those already there, probably to just pass them along
wfs.extend( session.query(Workflow).filter(Workflow.status== 'assistance').all() )
if options.old:
## than get all in need for assistance
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance-')).all() )
custodials = defaultdict(list) #sites : dataset list
transfers = defaultdict(list) #sites : dataset list
invalidations = [] #a list of files
SI = global_SI
CI = campaignInfo()
mcm = McMClient(dev=False)
def get_campaign(output, wfi):
campaign = None
try:
campaign = output.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request:
campaign = wfi.request['Campaign']
return campaign
## retrieve bypass and onhold configuration
bypasses = []
holdings = []
#try:
# already_notified = json.loads(open('already_notifified.json').read())
#except:
# print "no record of already notified workflow. starting fresh"
# already_notified = []
for bypassor,email in [('vlimant','*****@*****.**'),('jen_a','*****@*****.**')]:
bypass_file = '/afs/cern.ch/user/%s/%s/public/ops/bypass.json'%(bypassor[0],bypassor)
if not os.path.isfile(bypass_file):
#sendLog('checkor','no file %s',bypass_file)
continue
try:
bypasses.extend( json.loads(open(bypass_file).read()))
except:
sendLog('checkor',"cannot get by-passes from %s for %s"%(bypass_file ,bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(bypass_file), destination=[email])
holding_file = '/afs/cern.ch/user/%s/%s/public/ops/onhold.json'%(bypassor[0],bypassor)
if not os.path.isfile(holding_file):
#sendLog('checkor',"no file %s"%holding_file)
continue
try:
holdings.extend( json.loads(open(holding_file).read()))
except:
sendLog('checkor',"cannot get holdings from %s for %s"%(holding_file, bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(holding_file), destination=[email])
## once this was force-completed, you want to bypass
for rider,email in [('vlimant','*****@*****.**'),('jen_a','*****@*****.**'),('srimanob','*****@*****.**')]:
rider_file = '/afs/cern.ch/user/%s/%s/public/ops/forcecomplete.json'%(rider[0],rider)
if not os.path.isfile(rider_file):
print "no file",rider_file
#sendLog('checkor',"no file %s"%rider_file)
continue
try:
bypasses.extend( json.loads(open( rider_file ).read() ) )
except:
sendLog('checkor',"cannot get force complete list from %s"%rider)
sendEmail("malformated force complet file","%s is not json readable"%rider_file, destination=[email])
if use_mcm:
mcm_force = mcm.get('/restapi/requests/forcecomplete')
bypasses.extend( mcm_force )
pattern_fraction_pass = UC.get('pattern_fraction_pass')
total_running_time = 5.*60.
sleep_time = 1
if len(wfs):
sleep_time = min(max(0.5, total_running_time / len(wfs)), 10)
random.shuffle( wfs )
print len(wfs),"to consider, pausing for",sleep_time
for wfo in wfs:
if spec and not (spec in wfo.name): continue
time.sleep( sleep_time )
## get info
wfi = workflowInfo(url, wfo.name)
wfi.sendLog('checkor',"checking on %s %s"%( wfo.name,wfo.status))
## make sure the wm status is up to date.
# and send things back/forward if necessary.
wfo.wm_status = wfi.request['RequestStatus']
if wfo.wm_status == 'closed-out':
## manually closed-out
wfi.sendLog('checkor',"%s is already %s, setting close"%( wfo.name , wfo.wm_status))
wfo.status = 'close'
session.commit()
continue
elif wfo.wm_status in ['failed','aborted','aborted-archived','rejected','rejected-archived','aborted-completed']:
## went into trouble
wfo.status = 'trouble'
wfi.sendLog('checkor',"%s is in trouble %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
elif wfo.wm_status in ['assigned','acquired']:
## not worth checking yet
wfi.sendLog('checkor',"%s is not running yet"%wfo.name)
session.commit()
continue
if '-onhold' in wfo.status:
if wfo.name in holdings and wfo.name not in bypasses:
wfi.sendLog('checkor',"%s is on hold"%wfo.name)
continue
if wfo.wm_status != 'completed': #and not wfo.name in bypasses:
## for sure move on with closeout check if in completed
wfi.sendLog('checkor',"no need to check on %s in status %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
if wfo.name in holdings and wfo.name not in bypasses:
wfo.status = 'assistance-onhold'
wfi.sendLog('checkor',"setting %s on hold"%wfo.name)
session.commit()
continue
session.commit()
#sub_assistance="" # if that string is filled, there will be need for manual assistance
existing_assistance_tags = set(wfo.status.split('-')[1:]) #[0] should be assistance
assistance_tags = set()
is_closing = True
## get it from somewhere
bypass_checks = False
pids = wfi.getPrepIDs()
bypass_by_mcm = False
for bypass in bypasses:
if bypass in wfo.name:
wfi.sendLog('checkor',"we can bypass checks on %s because of keyword %s "%( wfo.name, bypass))
bypass_checks = True
break
if bypass in pids:
wfi.sendLog('checkor',"we can bypass checks on %s because of prepid %s "%( wfo.name, bypass))
bypass_checks = True
bypass_by_mcm = True
break
#if not CI.go( wfi.request['Campaign'] ) and not bypass_checks:
# print "No go for",wfo.name
# wfi.sendLog('checkor',"No go for %s"%wfi.request['Campaign'])
# continue
tiers_with_no_check = copy.deepcopy(UC.get('tiers_with_no_check')) # dqm*
vetoed_custodial_tier = copy.deepcopy(UC.get('tiers_with_no_custodial')) #dqm*, reco
campaigns = {}
for out in wfi.request['OutputDatasets']:
c = get_campaign(out, wfi)
campaigns[out] = c
if c in CI.campaigns and 'custodial_override' in CI.campaigns[c]:
vetoed_custodial_tier = list(set(vetoed_custodial_tier) - set(CI.campaigns[c]['custodial_override']))
## add those that we need to check for custodial copy
tiers_with_no_check = list(set(tiers_with_no_check) - set(CI.campaigns[c]['custodial_override'])) ## would remove DQM from the vetoed check
check_output_text = "Initial outputs:"+",".join(sorted(wfi.request['OutputDatasets'] ))
wfi.request['OutputDatasets'] = [ out for out in wfi.request['OutputDatasets'] if not any([out.split('/')[-1] == veto_tier for veto_tier in tiers_with_no_check])]
check_output_text += "\nWill check on:"+",".join(sorted(wfi.request['OutputDatasets'] ))
check_output_text += "\ntiers out:"+",".join( sorted(tiers_with_no_check ))
check_output_text += "\ntiers no custodial:"+",".join( sorted(vetoed_custodial_tier) )
wfi.sendLog('checkor', check_output_text )
## anything running on acdc : getting the real prepid is not worth it
familly = getWorkflowById(url, wfi.request['PrepID'], details=True)
acdc = []
acdc_inactive = []
for member in familly:
if member['RequestType'] != 'Resubmission': continue
if member['RequestName'] == wfo.name: continue
if member['RequestDate'] < wfi.request['RequestDate']: continue
if member['RequestStatus'] in ['running-open','running-closed','assigned','acquired']:
print wfo.name,"still has an ACDC running",member['RequestName']
acdc.append( member['RequestName'] )
## cannot be bypassed!
is_closing = False
assistance_tags.add('recovering')
elif member['RequestStatus']==None:
print member['RequestName'],"is not real"
pass
else:
acdc_inactive.append( member['RequestName'] )
assistance_tags.add('recovered')
## completion check
percent_completions = {}
if not 'TotalInputEvents' in wfi.request:
event_expected,lumi_expected = 0,0
if not 'recovery' in wfo.status:
#sendEmail("missing member of the request","TotalInputEvents is missing from the workload of %s"% wfo.name, destination=['*****@*****.**'])
sendLog('checkor',"TotalInputEvents is missing from the workload of %s"% wfo.name, level='critical')
else:
event_expected,lumi_expected = wfi.request['TotalInputEvents'],wfi.request['TotalInputLumis']
if 'RequestNumEvents' in wfi.request and int(wfi.request['RequestNumEvents']):
event_expected = int(wfi.request['RequestNumEvents'])
elif 'Task1' in wfi.request and 'RequestNumEvents' in wfi.request['Task1']:
event_expected = int(wfi.request['Task1']['RequestNumEvents'])
fractions_pass = {}
over_100_pass = False
(lhe,prim,_,_) = wfi.getIO()
if lhe or prim: over_100_pass = False
for output in wfi.request['OutputDatasets']:
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
percent_completions[output] = 0.
if lumi_expected:
percent_completions[output] = lumi_count / float( lumi_expected )
if event_expected:
wfi.sendLog('checkor', "event completion real %s expected %s"%(event_count, event_expected ))
percent_completions[output] = max(percent_completions[output], float(event_count) / float( event_expected ) )
fractions_pass[output] = 0.95
c = campaigns[output]
if c in CI.campaigns and 'fractionpass' in CI.campaigns[c]:
fractions_pass[output] = CI.campaigns[c]['fractionpass']
wfi.sendLog('checkor', "overriding fraction to %s for %s by campaign requirement"%( fractions_pass[output], output))
if options.fractionpass:
fractions_pass[output] = options.fractionpass
print "overriding fraction to",fractions_pass[output],"by command line for",output
for key in pattern_fraction_pass:
if key in output:
fractions_pass[output] = pattern_fraction_pass[key]
print "overriding fraction to",fractions_pass[output],"by dataset key",key
if not all([percent_completions[out] >= fractions_pass[out] for out in fractions_pass]):
print wfo.name,"is not completed"
print json.dumps(percent_completions, indent=2)
print json.dumps(fractions_pass, indent=2)
## hook for creating automatically ACDC ?
if not bypass_checks:
assistance_tags.add('recovery')
is_closing = False
if over_100_pass and any([percent_completions[out] >100 for out in fractions_pass]):
print wfo.name,"is over completed"
print json.dumps(percent_completions, indent=2)
if not bypass_checks:
assistance_tags.add('over100')
is_closing = False
## correct lumi < 300 event per lumi
events_per_lumi = {}
for output in wfi.request['OutputDatasets']:
events_per_lumi[output] = getDatasetEventsPerLumi( output )
lumi_upper_limit = {}
for output in wfi.request['OutputDatasets']:
upper_limit = 301.
campaign = campaigns[output]
#if 'EventsPerLumi' in wfi.request and 'FilterEfficiency' in wfi.request:
# upper_limit = 1.5*wfi.request['EventsPerLumi']*wfi.request['FilterEfficiency']
# print "setting the upper limit of lumisize to",upper_limit,"by request configuration"
if campaign in CI.campaigns and 'lumisize' in CI.campaigns[campaign]:
upper_limit = CI.campaigns[campaign]['lumisize']
print "overriding the upper lumi size to",upper_limit,"for",campaign
if options.lumisize:
upper_limit = options.lumisize
print "overriding the upper lumi size to",upper_limit,"by command line"
lumi_upper_limit[output] = upper_limit
if wfi.request['RequestType'] in ['ReDigi']: lumi_upper_limit[output] = -1
if any([ (lumi_upper_limit[out]>0 and events_per_lumi[out] >= lumi_upper_limit[out]) for out in events_per_lumi]):
print wfo.name,"has big lumisections"
print json.dumps(events_per_lumi, indent=2)
## hook for rejecting the request ?
if not bypass_checks:
assistance_tags.add('biglumi')
is_closing = False
any_presence = {}
for output in wfi.request['OutputDatasets']:
any_presence[output] = getDatasetPresence(url, output, vetoes=[])
## custodial copy
custodial_locations = {}
custodial_presences = {}
for output in wfi.request['OutputDatasets']:
custodial_presences[output] = [s for s in any_presence[output] if 'MSS' in s]
custodial_locations[output] = phedexClient.getCustodialSubscriptionRequestSite(output)
if not custodial_locations[output]:
custodial_locations[output] = []
## presence in phedex
phedex_presence ={}
for output in wfi.request['OutputDatasets']:
phedex_presence[output] = phedexClient.getFileCountDataset(url, output )
out_worth_checking = [out for out in custodial_locations.keys() if out.split('/')[-1] not in vetoed_custodial_tier]
size_worth_checking = sum([getDatasetSize(out)/1023. for out in out_worth_checking ]) ## size in TBs of all outputs
if not all(map( lambda sites : len(sites)!=0, [custodial_locations[out] for out in out_worth_checking])):
print wfo.name,"has not all custodial location"
print json.dumps(custodial_locations, indent=2)
##########
## hook for making a custodial replica ?
custodial = None
## get from other outputs
for output in out_worth_checking:
if len(custodial_locations[output]):
custodial = custodial_locations[output][0]
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the other output custodial:",custodial,"because of limited space"
custodial = None
## try to get it from campaign configuration
if not custodial:
for output in out_worth_checking:
campaign = campaigns[output]
if campaign in CI.campaigns and 'custodial' in CI.campaigns[campaign]:
custodial = CI.campaigns[campaign]['custodial']
print "Setting custodial to",custodial,"from campaign configuration"
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the campaign configuration custodial:",custodial,"because of limited space"
custodial = None
## get from the parent
pick_custodial = True
use_parent_custodial = UC.get('use_parent_custodial')
_,prim,_,_ = wfi.getIO()
if not custodial and prim and use_parent_custodial:
parent_dataset = prim.pop()
## this is terribly dangerous to assume only
parents_custodial = phedexClient.getCustodialSubscriptionRequestSite( parent_dataset )
###parents_custodial = findCustodialLocation(url, parent_dataset)
if not parents_custodial:
parents_custodial = []
if len(parents_custodial):
custodial = parents_custodial[0]
else:
print "the input dataset",parent_dataset,"does not have custodial in the first place. abort"
#sendEmail( "dataset has no custodial location", "Please take a look at %s in the logs of checkor"%parent_dataset)
## does not work for RAWOADSIM
sendLog('checkor',"Please take a look at %s for missing custodial location"% parent_dataset)
## cannot be bypassed, this is an issue to fix
is_closing = False
pick_custodial = False
assistance_tags.add('parentcustodial')
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the parent custodial:",custodial,"because of limited space"
custodial = None
if not custodial and pick_custodial:
## pick one at random
custodial = SI.pick_SE(size=size_worth_checking)
if not custodial:
print "cannot find a custodial for",wfo.name
wfi.sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking))
#sendEmail( "cannot find a custodial","cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking))
sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking), level='critical')
if custodial and (is_closing or bypass_checks):
print "picked",custodial,"for tape copy"
## remember how much you added this round already ; this stays locally
SI.storage[custodial] -= size_worth_checking
## register the custodial request, if there are no other big issues
for output in out_worth_checking:
if not len(custodial_locations[output]):
if phedex_presence[output]>=1:
custodials[custodial].append( output )
## let's wait and see if that's needed
assistance_tags.add('custodial')
else:
print "no file in phedex for",output," not good to add to custodial requests"
#cannot be bypassed
is_closing = False
## disk copy
disk_copies = {}
for output in wfi.request['OutputDatasets']:
disk_copies[output] = [s for s in any_presence[output] if (not 'MSS' in s) and (not 'Buffer' in s)]
if not all(map( lambda sites : len(sites)!=0, disk_copies.values())):
print wfo.name,"has not all output on disk"
print json.dumps(disk_copies, indent=2)
## presence in dbs
dbs_presence = {}
dbs_invalid = {}
for output in wfi.request['OutputDatasets']:
dbs_presence[output] = dbs3Client.getFileCountDataset( output )
dbs_invalid[output] = dbs3Client.getFileCountDataset( output, onlyInvalid=True)
fraction_invalid = 0.01
if not all([dbs_presence[out] == (dbs_invalid[out]+phedex_presence[out]) for out in wfi.request['OutputDatasets']]) and not options.ignorefiles:
print wfo.name,"has a dbs,phedex mismatch"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
if not 'recovering' in assistance_tags:
assistance_tags.add('filemismatch')
#print this for show and tell if no recovery on-going
for out in dbs_presence:
_,_,missing_phedex,missing_dbs = getDatasetFiles(url, out)
if missing_phedex:
print "These %d files are missing in phedex"%(len(missing_phedex))
print "\n".join( missing_phedex )
if missing_dbs:
print "These %d files are missing in dbs"%(len(missing_dbs))
print "\n".join( missing_dbs )
#if not bypass_checks:
## I don't think we can by pass this
is_closing = False
if not all([(dbs_invalid[out] <= int(fraction_invalid*dbs_presence[out])) for out in wfi.request['OutputDatasets']]) and not options.ignorefiles:
print wfo.name,"has a dbs invalid file level too high"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
## need to be going and taking an eye
assistance_tags.add('invalidfiles')
if not bypass_checks:
#sub_assistance+="-invalidfiles"
is_closing = False
## put that heavy part at the end
## duplication check
duplications = {}
if is_closing or bypass_checks:
print "starting duplicate checker for",wfo.name
for output in wfi.request['OutputDatasets']:
print "\tchecking",output
duplications[output] = True
try:
duplications[output] = dbs3Client.duplicateRunLumi( output , skipInvalid=True, verbose=True)
except:
try:
duplications[output] = dbs3Client.duplicateRunLumi( output , skipInvalid=True, verbose=True)
except:
print "was not possible to get the duplicate count for",output
is_closing=False
if any(duplications.values()) and not options.ignoreduplicates:
print wfo.name,"has duplicates"
print json.dumps(duplications,indent=2)
## hook for making file invalidation ?
## it shouldn't be allowed to bypass it
assistance_tags.add('duplicates')
is_closing = False
## for visualization later on
if not wfo.name in fDB.record:
#print "adding",wfo.name,"to close out record"
fDB.record[wfo.name] = {
'datasets' :{},
'name' : wfo.name,
'closeOutWorkflow' : None,
}
fDB.record[wfo.name]['closeOutWorkflow'] = is_closing
fDB.record[wfo.name]['priority'] = wfi.request['RequestPriority']
fDB.record[wfo.name]['prepid'] = wfi.request['PrepID']
for output in wfi.request['OutputDatasets']:
if not output in fDB.record[wfo.name]['datasets']: fDB.record[wfo.name]['datasets'][output] = {}
rec = fDB.record[wfo.name]['datasets'][output]
rec['percentage'] = float('%.2f'%(percent_completions[output]*100))
rec['duplicate'] = duplications[output] if output in duplications else 'N/A'
rec['phedexReqs'] = float('%.2f'%any_presence[output][custodial_presences[output][0]][1]) if len(custodial_presences[output])!=0 else 'N/A'
rec['closeOutDataset'] = is_closing
rec['transPerc'] = float('%.2f'%any_presence[output][ disk_copies[output][0]][1]) if len(disk_copies[output])!=0 else 'N/A'
rec['correctLumis'] = int(events_per_lumi[output]) if (events_per_lumi[output] > lumi_upper_limit[output]) else True
rec['missingSubs'] = False if len(custodial_locations[output])==0 else ','.join(list(set(custodial_locations[output])))
rec['dbsFiles'] = dbs_presence[output]
rec['dbsInvFiles'] = dbs_invalid[output]
rec['phedexFiles'] = phedex_presence[output]
rec['acdc'] = "%d / %d"%(len(acdc),len(acdc+acdc_inactive))
## and move on
if is_closing:
## toggle status to closed-out in request manager
print "setting",wfo.name,"closed-out"
if not options.test:
if wfo.wm_status in ['closed-out','announced','normal-archived']:
print wfo.name,"is already",wfo.wm_status,"not trying to closed-out and assuming it does"
res = None
else:
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
print "close out answer",res
if not res in ["None",None]:
print "try to get the current status again"
wfi_bis = workflowInfo(url, wfo.name)
if wfi_bis.request['RequestStatus'] == 'closed-out':
print "the request did toggle to closed-out"
res = None
if not res in ["None",None]:
print "retrying to closing out"
print res
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
if res in [None,"None"]:
wfo.status = 'close'
session.commit()
if use_mcm and bypass_by_mcm:
## shoot large on all prepids
for pid in pids:
mcm.delete('/restapi/requests/forcecomplete/%s'%pid)
else:
print "could not close out",wfo.name,"will try again next time"
else:
## full known list
#recovering # has active ACDC
##OUT #recovered #had inactive ACDC
#recovery #not over the pass bar
#over100 # over 100%
#biglumi # has a big lumiblock
#parentcustodial # the parent does not have a valid subscription yet
#custodial # has had the transfer made, is waiting for a valid custodial subscription to appear
#filemismatch # there is a dbs/phedex mismatch
#duplicates #a lumi section is there twice
## manual is not added yet, and should be so by recoveror
print wfo.name,"was tagged with :",list(assistance_tags)
if 'recovering' in assistance_tags:
## if active ACDC, being under threshold, filemismatch do not matter
assistance_tags = assistance_tags - set(['recovery','filemismatch'])
if 'recovery' in assistance_tags and 'recovered' in assistance_tags:
## should not set -recovery to anything that add ACDC already
assistance_tags = assistance_tags - set(['recovery','recovered'])
## straight to manual
assistance_tags.add('manual')
## that means there is something that needs to be done acdc, lumi invalidation, custodial, name it
print wfo.name,"needs assistance with",",".join( assistance_tags )
print wfo.name,"existing conditions",",".join( existing_assistance_tags )
#########################################
##### notification to requester #########
go_notify=False
if assistance_tags and not 'manual' in existing_assistance_tags and existing_assistance_tags != assistance_tags:
go_notify=True
if go_notify:
#if wfo.name in already_notified:
# print "double notification"
# sendEmail('double notification','please take a look at %s'%(wfo.name))
#else:
# already_notified.append( wfo.name )
detailslink = 'https://cmsweb.cern.ch/reqmgr/view/details/%s'
perflink = 'https://cmsweb.cern.ch/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s'%(wfo.name)
splitlink = 'https://cmsweb.cern.ch/reqmgr/view/splitting/%s'%(wfo.name)
## notify templates
messages= {
'recovery': 'Samples completed with missing statistics:\n%s\n%s '%( '\n'.join(['%.2f %% complete for %s'%(percent_completions[output]*100, output) for output in wfi.request['OutputDatasets'] ] ), perflink ),
'biglumi': 'Samples completed with large luminosity blocks:\n%s\n%s '%('\n'.join(['%d > %d for %s'%(events_per_lumi[output], lumi_upper_limit[output], output) for output in wfi.request['OutputDatasets'] if (events_per_lumi[output] > lumi_upper_limit[output])]), splitlink),
'duplicates': 'Samples completed with duplicated luminosity blocks:\n%s\n'%( '\n'.join(['%s'%output for output in wfi.request['OutputDatasets'] if output in duplications and duplications[output] ] ) ),
'filemismatch': 'Samples completed with inconsistency in DBS/Phedex',
#'manual' : 'Workflow completed and requires manual checks by Ops',
}
content = "The request PREPID (WORKFLOW) is facing issue in production.\n"
motive = False
for case in messages:
if case in assistance_tags:
content+= "\n"+messages[case]+"\n"
motive = True
content += "You are invited to check, while this is being taken care of by Comp-Ops.\n"
content += "This is an automated message from Comp-Ops.\n"
items_notified = set()
if use_mcm and motive:
wfi.notifyRequestor( content , mcm = mcm)
#########################################
## logic to set the status further
if assistance_tags:
new_status = 'assistance-'+'-'.join(sorted(assistance_tags) )
else:
new_status = 'assistance'
## case where the workflow was in manual from recoveror
if not 'manual' in wfo.status or new_status!='assistance-recovery':
wfo.status = new_status
if not options.test:
print "setting",wfo.name,"to",wfo.status
session.commit()
else:
print "current status is",wfo.status,"not changing to anything"
#open('already_notifified.json','w').write( json.dumps( already_notified , indent=2))
fDB.html()
if not spec:
#sendEmail("fresh assistance status available","Fresh status are available at https://cmst2.web.cern.ch/cmst2/unified/assistance.html",destination=['*****@*****.**'])
#it's a bit annoying
pass
## custodial requests
print "Custodials"
print json.dumps(custodials, indent=2)
for site in custodials:
print ','.join(custodials[site]),'=>',site
if not options.test:
result = makeReplicaRequest(url, site, list(set(custodials[site])),"custodial copy at production close-out",custodial='y',priority='low', approve = (site in SI.sites_auto_approve) )
print result
print "Transfers"
print json.dumps(transfers, indent=2)
## replicas requests
for site in transfers:
print ','.join(transfers[site]),'=>',site
if not options.test:
result = None
#result = makeReplicaRequest(url, site, list(set(transfers[site])),"copy to disk at production close-out")
print result
print "File Invalidation"
print invalidations
def closor(url, specific=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
all_late_files = []
check_fullcopy_to_announce = UC.get('check_fullcopy_to_announce')
## manually closed-out workflows should get to close with checkor
if specific:
wfs = session.query(Workflow).filter(
Workflow.name.contains(specific)).all()
else:
wfs = session.query(Workflow).filter(Workflow.status == 'close').all()
held = set()
max_per_round = UC.get('max_per_round').get('closor', None)
random.shuffle(wfs)
if max_per_round: wfs = wfs[:max_per_round]
for wfo in wfs:
if specific and not specific in wfo.name: continue
## what is the expected #lumis
wfi = workflowInfo(url, wfo.name)
wfo.wm_status = wfi.request['RequestStatus']
if wfi.request['RequestStatus'] in ['announced', 'normal-archived']:
## manually announced ??
wfo.status = 'done'
wfo.wm_status = wfi.request['RequestStatus']
wfi.sendLog(
'closor',
'%s is announced already : %s' % (wfo.name, wfo.wm_status))
session.commit()
expected_lumis = 1
if not 'TotalInputLumis' in wfi.request:
print wfo.name, "has not been assigned yet, or the database is corrupted"
else:
expected_lumis = wfi.request['TotalInputLumis']
## what are the outputs
outputs = wfi.request['OutputDatasets']
## check whether the number of lumis is as expected for each
all_OK = defaultdict(lambda: False)
#print outputs
if len(outputs):
print wfo.name, wfi.request['RequestStatus']
for out in outputs:
event_count, lumi_count = getDatasetEventsAndLumis(dataset=out)
odb = session.query(Output).filter(
Output.datasetname == out).first()
if not odb:
print "adding an output object", out
odb = Output(datasetname=out)
odb.workflow = wfo
session.add(odb)
odb.nlumis = lumi_count
odb.nevents = event_count
odb.workfow_id = wfo.id
if odb.expectedlumis < expected_lumis:
odb.expectedlumis = expected_lumis
else:
expected_lumis = odb.expectedlumis
odb.date = time.mktime(time.gmtime())
session.commit()
wfi.sendLog(
'closor', "\t%60s %d/%d = %3.2f%%" %
(out, lumi_count, expected_lumis,
lumi_count / float(expected_lumis) * 100.))
#print wfo.fraction_for_closing, lumi_count, expected_lumis
#fraction = wfo.fraction_for_closing
#fraction = 0.0
#all_OK.append((float(lumi_count) > float(expected_lumis*fraction)))
all_OK[out] = True
## check for at least one full copy prior to moving on
in_full = {}
for out in outputs:
in_full[out] = []
presence = getDatasetPresence(url, out)
where = [site for site, info in presence.items() if info[0]]
if where:
all_OK[out] = True
print out, "is in full at", ",".join(where)
in_full[out] = copy.deepcopy(where)
else:
going_to = wfi.request['NonCustodialSites'] + wfi.request[
'CustodialSites']
wfi.sendLog(
'closor', "%s is not in full anywhere. send to %s" %
(out, ",".join(sorted(going_to))))
at_destination = dict([(k, v) for (k, v) in presence.items()
if k in going_to])
else_where = dict([(k, v) for (k, v) in presence.items()
if not k in going_to])
print json.dumps(at_destination)
print json.dumps(else_where, indent=2)
## do the full stuck transfer study, missing files and shit !
for there in going_to:
late_info = findLateFiles(url, out, going_to=there)
for l in late_info:
l.update({"workflow": wfo.name, "dataset": out})
all_late_files.extend(late_info)
if check_fullcopy_to_announce:
## only set this false if the check is relevant
all_OK[out] = False
## verify if we have to do harvesting
(OK, requests) = spawn_harvesting(url, wfi, in_full)
all_OK.update(OK)
## only that status can let me go into announced
if all(all_OK.values()) and wfi.request['RequestStatus'] in [
'closed-out'
]:
print wfo.name, "to be announced"
results = [] #'dummy']
if not results:
for out in outputs:
if all_OK[out]:
results.append(setDatasetStatus(out, 'VALID'))
tier = out.split('/')[-1]
campaign = None
try:
campaign = out.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request and wfi.request[
'Campaign']:
campaign = wfi.request['Campaign']
to_DDM = False
## campaign override
if campaign and campaign in CI.campaigns and 'toDDM' in CI.campaigns[
campaign] and tier in CI.campaigns[campaign][
'toDDM']:
to_DDM = True
## by typical enabling
if tier in UC.get("tiers_to_DDM"):
to_DDM = True
## check for unitarity
if not tier in UC.get("tiers_no_DDM") + UC.get(
"tiers_to_DDM"):
print "tier", tier, "neither TO or NO DDM for", out
results.append('Not recognitized tier %s' % tier)
#sendEmail("failed DDM injection","could not recognize %s for injecting in DDM"% out)
sendLog(
'closor',
"could not recognize %s for injecting in DDM" %
out,
level='critical')
continue
n_copies = 2
destinations = []
if to_DDM and campaign and campaign in CI.campaigns and 'DDMcopies' in CI.campaigns[
campaign]:
ddm_instructions = CI.campaigns[campaign][
'DDMcopies']
if type(ddm_instructions) == int:
n_copies = CI.campaigns[campaign]['DDMcopies']
elif type(ddm_instructions) == dict:
## a more fancy configuration
for ddmtier, indication in ddm_instructions.items(
):
if ddmtier == tier or ddmtier in [
'*', 'all'
]:
## this is for us
if 'N' in indication:
n_copies = indication['N']
if 'host' in indication:
destinations = indication['host']
destination_spec = ""
if destinations:
destination_spec = "--destination=" + ",".join(
destinations)
## inject to DDM when necessary
if to_DDM:
#print "Sending",out," to DDM"
p = os.popen(
'python assignDatasetToSite.py --nCopies=%d --dataset=%s %s --exec'
% (n_copies, out, destination_spec))
print p.read()
status = p.close()
if status != None:
print "Failed DDM, retrying a second time"
p = os.popen(
'python assignDatasetToSite.py --nCopies=%d --dataset=%s %s --exec'
% (n_copies, out, destination_spec))
print p.read()
status = p.close()
if status != None:
#sendEmail("failed DDM injection","could not add "+out+" to DDM pool. check closor logs.")
sendLog('closor',
"could not add " + out +
" to DDM pool. check closor logs.",
level='critical')
results.append(status)
if status == None:
wfi.sendLog(
'closor',
'%s is send to AnalysisOps DDM pool in %s copies %s'
% (n_copies, out, destination_spec))
else:
print wfo.name, "no stats for announcing", out
results.append('No Stats')
if all(
map(lambda result: result in ['None', None, True],
results)):
## only announce if all previous are fine
res = reqMgrClient.announceWorkflowCascade(url, wfo.name)
if not res in ['None', None]:
## check the status again, it might well have toggled
wl_bis = workflowInfo(url, wfo.name)
wfo.wm_status = wl_bis.request['RequestStatus']
session.commit()
if wl_bis.request['RequestStatus'] in [
'announced', 'normal-archived'
]:
res = None
else:
## retry ?
res = reqMgrClient.announceWorkflowCascade(
url, wfo.name)
results.append(res)
#print results
if all(map(lambda result: result in ['None', None, True],
results)):
wfo.status = 'done'
session.commit()
wfi.sendLog('closor', "workflow is announced")
else:
print "ERROR with ", wfo.name, "to be announced", json.dumps(
results)
else:
print wfo.name, "not good for announcing:", wfi.request[
'RequestStatus']
wfi.sendLog('closor', "cannot be announced")
held.add(wfo.name)
days_late = 0.
retries_late = 10
really_late_files = [
info for info in all_late_files if info['retries'] >= retries_late
]
really_late_files = [
info for info in really_late_files
if info['delay'] / (60 * 60 * 24.) >= days_late
]
if really_late_files:
subject = 'These %d files are lagging for %d days and %d retries announcing dataset \n%s' % (
len(really_late_files), days_late, retries_late,
json.dumps(really_late_files, indent=2))
#sendEmail('waiting for files to announce', subject)
sendLog('closor', subject, level='warning')
sendLog('closor', subject)
print subject
open('%s/stuck_files.json' % monitor_dir,
'w').write(json.dumps(really_late_files, indent=2))
if held:
#sendEmail("held from announcing","the workflows below are held up, please check the logs https://cmst2.web.cern.ch/cmst2/unified/logs/closor/last.log \n%s"%("\n".join( held )))
sendLog('closor',
"the workflows below are held up \n%s" % ("\n".join(held)),
level='critical')
def checkor(url, spec=None, options=None):
if userLock(): return
if duplicateLock(): return
fDB = closeoutInfo()
UC = unifiedConfiguration()
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
wfs=[]
if options.new:
## get all in running and check
## you want to intersect with what is completed !
if options.strict:
completed_wfi = getWorkflows(url, status='completed')
for wfo in session.query(Workflow).filter(Workflow.status == 'away').all():
if wfo.name in completed_wfi:
wfs.append( wfo )
else:
print wfo.name,"is not completed"
sendLog('checkor','%s is not completed'%( wfo.name))
else:
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
if options.current:
## recheck those already there, probably to just pass them along
wfs.extend( session.query(Workflow).filter(Workflow.status== 'assistance').all() )
if options.old:
## than get all in need for assistance
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance-')).all() )
custodials = defaultdict(list) #sites : dataset list
transfers = defaultdict(list) #sites : dataset list
invalidations = [] #a list of files
SI = siteInfo()
CI = campaignInfo()
mcm = McMClient(dev=False)
def get_campaign(output, wfi):
campaign = None
try:
campaign = output.split('/')[2].split('-')[0]
except:
if 'Campaign' in wfi.request:
campaign = wfi.request['Campaign']
return campaign
## retrieve bypass and onhold configuration
bypasses = []
forcings = []
overrides = getForceCompletes()
holdings = []
for bypassor,email in [('vlimant','*****@*****.**'),('jen_a','*****@*****.**'),('prozober','*****@*****.**')]:
bypass_file = '/afs/cern.ch/user/%s/%s/public/ops/bypass.json'%(bypassor[0],bypassor)
if not os.path.isfile(bypass_file):
#sendLog('checkor','no file %s',bypass_file)
continue
try:
bypasses.extend( json.loads(open(bypass_file).read()))
except:
sendLog('checkor',"cannot get by-passes from %s for %s"%(bypass_file ,bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(bypass_file), destination=[email])
holding_file = '/afs/cern.ch/user/%s/%s/public/ops/onhold.json'%(bypassor[0],bypassor)
if not os.path.isfile(holding_file):
#sendLog('checkor',"no file %s"%holding_file)
continue
try:
holdings.extend( json.loads(open(holding_file).read()))
except:
sendLog('checkor',"cannot get holdings from %s for %s"%(holding_file, bypassor))
sendEmail("malformated by-pass information","%s is not json readable"%(holding_file), destination=[email])
## once this was force-completed, you want to bypass
for rider,email in [('vlimant','*****@*****.**'),('jen_a','*****@*****.**'),('srimanob','*****@*****.**')]:
rider_file = '/afs/cern.ch/user/%s/%s/public/ops/forcecomplete.json'%(rider[0],rider)
if not os.path.isfile(rider_file):
print "no file",rider_file
#sendLog('checkor',"no file %s"%rider_file)
continue
try:
bypasses.extend( json.loads(open( rider_file ).read() ) )
except:
sendLog('checkor',"cannot get force complete list from %s"%rider)
sendEmail("malformated force complet file","%s is not json readable"%rider_file, destination=[email])
if use_mcm:
forcings = mcm.get('/restapi/requests/forcecomplete')
if forcings:
sendEmail('force completing mechanism','please check what checkor is doing with %s'%( ','.join(forcings)))
pattern_fraction_pass = UC.get('pattern_fraction_pass')
total_running_time = 5.*60.
sleep_time = 1
if len(wfs):
sleep_time = min(max(0.5, total_running_time / len(wfs)), 10)
random.shuffle( wfs )
print len(wfs),"to consider, pausing for",sleep_time
max_per_round = UC.get('max_per_round').get('checkor',None)
if max_per_round and not spec: wfs = wfs[:max_per_round]
for wfo in wfs:
if spec and not (spec in wfo.name): continue
time.sleep( sleep_time )
## get info
wfi = workflowInfo(url, wfo.name)
wfi.sendLog('checkor',"checking on %s %s"%( wfo.name,wfo.status))
## make sure the wm status is up to date.
# and send things back/forward if necessary.
wfo.wm_status = wfi.request['RequestStatus']
if wfo.wm_status == 'closed-out':
## manually closed-out
wfi.sendLog('checkor',"%s is already %s, setting close"%( wfo.name , wfo.wm_status))
wfo.status = 'close'
session.commit()
continue
elif wfo.wm_status in ['failed','aborted','aborted-archived','rejected','rejected-archived','aborted-completed']:
## went into trouble
wfo.status = 'trouble'
wfi.sendLog('checkor',"%s is in trouble %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
elif wfo.wm_status in ['assigned','acquired']:
## not worth checking yet
wfi.sendLog('checkor',"%s is not running yet"%wfo.name)
session.commit()
continue
if '-onhold' in wfo.status:
if wfo.name in holdings and wfo.name not in bypasses:
wfi.sendLog('checkor',"%s is on hold"%wfo.name)
continue
if wfo.wm_status != 'completed': #and not wfo.name in bypasses:
## for sure move on with closeout check if in completed
wfi.sendLog('checkor',"no need to check on %s in status %s"%(wfo.name, wfo.wm_status))
session.commit()
continue
if wfo.name in holdings and wfo.name not in bypasses:
wfo.status = 'assistance-onhold'
wfi.sendLog('checkor',"setting %s on hold"%wfo.name)
session.commit()
continue
session.commit()
#sub_assistance="" # if that string is filled, there will be need for manual assistance
existing_assistance_tags = set(wfo.status.split('-')[1:]) #[0] should be assistance
assistance_tags = set()
is_closing = True
## get it from somewhere
bypass_checks = False
for bypass in bypasses:
if bypass in wfo.name:
wfi.sendLog('checkor',"we can bypass checks on %s because of keyword %s "%( wfo.name, bypass))
bypass_checks = True
break
pids = wfi.getPrepIDs()
force_by_mcm = False
force_by_user = False
for force in forcings:
if force in pids:
wfi.sendLog('checkor',"we can bypass checks and force complete %s because of prepid %s "%( wfo.name, force))
bypass_checks = True
force_by_mcm = True
break
for user in overrides:
for force in overrides[user]:
if force in wfo.name:
wfi.sendLog('checkor',"we can bypass checks and force complete %s because of keyword %s of user %s"%( wfo.name, force, user))
bypass_checks = True
force_by_user = True
break
tiers_with_no_check = copy.deepcopy(UC.get('tiers_with_no_check')) # dqm*
vetoed_custodial_tier = copy.deepcopy(UC.get('tiers_with_no_custodial')) #dqm*, reco
campaigns = {}
expected_outputs = copy.deepcopy( wfi.request['OutputDatasets'] )
for out in wfi.request['OutputDatasets']:
c = get_campaign(out, wfi)
campaigns[out] = c
if c in CI.campaigns and 'custodial_override' in CI.campaigns[c]:
vetoed_custodial_tier = list(set(vetoed_custodial_tier) - set(CI.campaigns[c]['custodial_override']))
## add those that we need to check for custodial copy
tiers_with_no_check = list(set(tiers_with_no_check) - set(CI.campaigns[c]['custodial_override'])) ## would remove DQM from the vetoed check
check_output_text = "Initial outputs:"+",".join(sorted(wfi.request['OutputDatasets'] ))
wfi.request['OutputDatasets'] = [ out for out in wfi.request['OutputDatasets'] if not any([out.split('/')[-1] == veto_tier for veto_tier in tiers_with_no_check])]
check_output_text += "\nWill check on:"+",".join(sorted(wfi.request['OutputDatasets'] ))
check_output_text += "\ntiers out:"+",".join( sorted(tiers_with_no_check ))
check_output_text += "\ntiers no custodial:"+",".join( sorted(vetoed_custodial_tier) )
wfi.sendLog('checkor', check_output_text )
## anything running on acdc : getting the real prepid is not worth it
familly = getWorkflowById(url, wfi.request['PrepID'], details=True)
acdc = []
acdc_inactive = []
forced_already=False
acdc_bads = []
for member in familly:
if member['RequestType'] != 'Resubmission': continue
if member['RequestName'] == wfo.name: continue
if member['RequestDate'] < wfi.request['RequestDate']: continue
if 'OriginalRequestName' in member and member['OriginalRequestName'] != wfo.name: continue
if member['RequestStatus'] == None: continue
if not set(member['OutputDatasets']).issubset( set(expected_outputs)):
if not member['RequestStatus'] in ['rejected-archived','rejected','aborted','aborted-archived']:
##this is not good at all
wfi.sendLog('checkor','inconsistent ACDC %s'%member['RequestName'] )
acdc_bads.append( member['RequestName'] )
is_closing = False
assistance_tags.add('manual')
continue
if member['RequestStatus'] in ['running-open','running-closed','assigned','acquired']:
print wfo.name,"still has an ACDC running",member['RequestName']
acdc.append( member['RequestName'] )
## cannot be bypassed!
is_closing = False
assistance_tags.add('recovering')
if (force_by_mcm or force_by_user) and not forced_already:
wfi.sendLog('checkor','%s is being forced completed while recovering'%wfo.name)
wfi.notifyRequestor("The workflow %s was force completed"% wfo.name, do_batch=False)
forceComplete(url, wfi)
forced_already=True
else:
acdc_inactive.append( member['RequestName'] )
assistance_tags.add('recovered')
if acdc_bads:
sendEmail('inconsistent ACDC','for %s, ACDC %s is inconsistent, preventing from closing'%( wfo.name, ','.join(acdc_bads) ))
## completion check
percent_completions = {}
if not 'TotalInputEvents' in wfi.request:
event_expected,lumi_expected = 0,0
if not 'recovery' in wfo.status:
#sendEmail("missing member of the request","TotalInputEvents is missing from the workload of %s"% wfo.name, destination=['*****@*****.**'])
sendLog('checkor',"TotalInputEvents is missing from the workload of %s"% wfo.name, level='critical')
else:
event_expected,lumi_expected = wfi.request['TotalInputEvents'],wfi.request['TotalInputLumis']
if 'RequestNumEvents' in wfi.request and int(wfi.request['RequestNumEvents']):
event_expected = int(wfi.request['RequestNumEvents'])
elif 'Task1' in wfi.request and 'RequestNumEvents' in wfi.request['Task1']:
event_expected = wfi.request['Task1']['RequestNumEvents']
for i in range(1,20):
if 'Task%d'%i in wfi.request:
## this is wrong ibsolute
if 'FilterEfficiency' in wfi.request['Task%d'%i]:
event_expected *= float(wfi.request['Task%d'%i]['FilterEfficiency'])
event_expected = int(event_expected)
fractions_pass = {}
over_100_pass = False
(lhe,prim,_,_) = wfi.getIO()
if lhe or prim: over_100_pass = False
for output in wfi.request['OutputDatasets']:
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
percent_completions[output] = 0.
if lumi_expected:
percent_completions[output] = lumi_count / float( lumi_expected )
if event_expected:
wfi.sendLog('checkor', "event completion real %s expected %s"%(event_count, event_expected ))
percent_completions[output] = max(percent_completions[output], float(event_count) / float( event_expected ) )
fractions_pass[output] = 0.95
c = campaigns[output]
if c in CI.campaigns and 'fractionpass' in CI.campaigns[c]:
fractions_pass[output] = CI.campaigns[c]['fractionpass']
wfi.sendLog('checkor', "overriding fraction to %s for %s by campaign requirement"%( fractions_pass[output], output))
if options.fractionpass:
fractions_pass[output] = options.fractionpass
print "overriding fraction to",fractions_pass[output],"by command line for",output
for key in pattern_fraction_pass:
if key in output:
fractions_pass[output] = pattern_fraction_pass[key]
print "overriding fraction to",fractions_pass[output],"by dataset key",key
if not all([percent_completions[out] >= fractions_pass[out] for out in fractions_pass]):
possible_recoveries = wfi.getRecoveryDoc()
if possible_recoveries == []:
wfi.sendLog('checkor','%s has missing statistics \n%s \n%s, but nothing is recoverable. passing through to annoucement'%(
wfo.name, json.dumps(percent_completions, indent=2), json.dumps(fractions_pass, indent=2) ))
sendEmail('nothing is recoverable','%s is not completed, but has nothing to be recovered, passing along ?'%wfo.name)
bypass_checks = True
else:
wfi.sendLog('checkor','%s is not completed \n%s \n%s'%(
wfo.name, json.dumps(percent_completions, indent=2), json.dumps(fractions_pass, indent=2) ))
## hook for creating automatically ACDC ?
if not bypass_checks:
assistance_tags.add('recovery')
is_closing = False
if over_100_pass and any([percent_completions[out] >100 for out in fractions_pass]):
print wfo.name,"is over completed"
print json.dumps(percent_completions, indent=2)
if not bypass_checks:
assistance_tags.add('over100')
is_closing = False
## correct lumi < 300 event per lumi
events_per_lumi = {}
for output in wfi.request['OutputDatasets']:
events_per_lumi[output] = getDatasetEventsPerLumi( output )
lumi_upper_limit = {}
for output in wfi.request['OutputDatasets']:
upper_limit = 301.
campaign = campaigns[output]
#if 'EventsPerLumi' in wfi.request and 'FilterEfficiency' in wfi.request:
# upper_limit = 1.5*wfi.request['EventsPerLumi']*wfi.request['FilterEfficiency']
# print "setting the upper limit of lumisize to",upper_limit,"by request configuration"
if campaign in CI.campaigns and 'lumisize' in CI.campaigns[campaign]:
upper_limit = CI.campaigns[campaign]['lumisize']
print "overriding the upper lumi size to",upper_limit,"for",campaign
if options.lumisize:
upper_limit = options.lumisize
print "overriding the upper lumi size to",upper_limit,"by command line"
lumi_upper_limit[output] = upper_limit
if wfi.request['RequestType'] in ['ReDigi']: lumi_upper_limit[output] = -1
if any([ (lumi_upper_limit[out]>0 and events_per_lumi[out] >= lumi_upper_limit[out]) for out in events_per_lumi]):
print wfo.name,"has big lumisections"
print json.dumps(events_per_lumi, indent=2)
## hook for rejecting the request ?
if not bypass_checks:
assistance_tags.add('biglumi')
is_closing = False
any_presence = {}
for output in wfi.request['OutputDatasets']:
any_presence[output] = getDatasetPresence(url, output, vetoes=[])
## custodial copy
custodial_locations = {}
custodial_presences = {}
for output in wfi.request['OutputDatasets']:
custodial_presences[output] = [s for s in any_presence[output] if 'MSS' in s]
custodial_locations[output] = phedexClient.getCustodialSubscriptionRequestSite(output)
if not custodial_locations[output]:
custodial_locations[output] = []
## presence in phedex
phedex_presence ={}
for output in wfi.request['OutputDatasets']:
phedex_presence[output] = phedexClient.getFileCountDataset(url, output )
out_worth_checking = [out for out in custodial_locations.keys() if out.split('/')[-1] not in vetoed_custodial_tier]
size_worth_checking = sum([getDatasetSize(out)/1023. for out in out_worth_checking ]) ## size in TBs of all outputs
if not all(map( lambda sites : len(sites)!=0, [custodial_locations[out] for out in out_worth_checking])):
print wfo.name,"has not all custodial location"
print json.dumps(custodial_locations, indent=2)
##########
## hook for making a custodial replica ?
custodial = None
## get from other outputs
for output in out_worth_checking:
if len(custodial_locations[output]):
custodial = custodial_locations[output][0]
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the other output custodial:",custodial,"because of limited space"
custodial = None
## try to get it from campaign configuration
if not custodial:
for output in out_worth_checking:
campaign = campaigns[output]
if campaign in CI.campaigns and 'custodial' in CI.campaigns[campaign]:
custodial = CI.campaigns[campaign]['custodial']
print "Setting custodial to",custodial,"from campaign configuration"
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the campaign configuration custodial:",custodial,"because of limited space"
custodial = None
## get from the parent
pick_custodial = True
use_parent_custodial = UC.get('use_parent_custodial')
_,prim,_,_ = wfi.getIO()
if not custodial and prim and use_parent_custodial:
parent_dataset = prim.pop()
## this is terribly dangerous to assume only
parents_custodial = phedexClient.getCustodialSubscriptionRequestSite( parent_dataset )
###parents_custodial = findCustodialLocation(url, parent_dataset)
if not parents_custodial:
parents_custodial = []
if len(parents_custodial):
custodial = parents_custodial[0]
else:
print "the input dataset",parent_dataset,"does not have custodial in the first place. abort"
#sendEmail( "dataset has no custodial location", "Please take a look at %s in the logs of checkor"%parent_dataset)
## does not work for RAWOADSIM
sendLog('checkor',"Please take a look at %s for missing custodial location"% parent_dataset)
## cannot be bypassed, this is an issue to fix
is_closing = False
pick_custodial = False
assistance_tags.add('parentcustodial')
if custodial and float(SI.storage[custodial]) < size_worth_checking:
print "cannot use the parent custodial:",custodial,"because of limited space"
custodial = None
if not custodial and pick_custodial:
## pick one at random
custodial = SI.pick_SE(size=size_worth_checking)
if not custodial:
print "cannot find a custodial for",wfo.name
wfi.sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking))
#sendEmail( "cannot find a custodial","cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking))
sendLog('checkor',"cannot find a custodial for %s probably because of the total output size %d"%( wfo.name, size_worth_checking), level='critical')
if custodial and (is_closing or bypass_checks):
print "picked",custodial,"for tape copy"
## remember how much you added this round already ; this stays locally
SI.storage[custodial] -= size_worth_checking
## register the custodial request, if there are no other big issues
for output in out_worth_checking:
if not len(custodial_locations[output]):
if phedex_presence[output]>=1:
custodials[custodial].append( output )
## let's wait and see if that's needed
assistance_tags.add('custodial')
else:
print "no file in phedex for",output," not good to add to custodial requests"
#cannot be bypassed
is_closing = False
## disk copy
disk_copies = {}
for output in wfi.request['OutputDatasets']:
disk_copies[output] = [s for s in any_presence[output] if (not 'MSS' in s) and (not 'Buffer' in s)]
if not all(map( lambda sites : len(sites)!=0, disk_copies.values())):
print wfo.name,"has not all output on disk"
print json.dumps(disk_copies, indent=2)
## presence in dbs
dbs_presence = {}
dbs_invalid = {}
for output in wfi.request['OutputDatasets']:
dbs_presence[output] = dbs3Client.getFileCountDataset( output )
dbs_invalid[output] = dbs3Client.getFileCountDataset( output, onlyInvalid=True)
fraction_invalid = 0.01
if not all([dbs_presence[out] == (dbs_invalid[out]+phedex_presence[out]) for out in wfi.request['OutputDatasets']]) and not options.ignorefiles:
print wfo.name,"has a dbs,phedex mismatch"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
if not 'recovering' in assistance_tags:
assistance_tags.add('filemismatch')
#print this for show and tell if no recovery on-going
for out in dbs_presence:
_,_,missing_phedex,missing_dbs = getDatasetFiles(url, out)
if missing_phedex:
wfi.sendLog('checkor',"These %d files are missing in phedex\n%s"%(len(missing_phedex),
"\n".join( missing_phedex )))
if missing_dbs:
wfi.sendLog('checkor',"These %d files are missing in dbs\n%s"%(len(missing_dbs),
"\n".join( missing_dbs )))
#if not bypass_checks:
## I don't think we can by pass this
is_closing = False
if not all([(dbs_invalid[out] <= int(fraction_invalid*dbs_presence[out])) for out in wfi.request['OutputDatasets']]) and not options.ignorefiles:
print wfo.name,"has a dbs invalid file level too high"
print json.dumps(dbs_presence, indent=2)
print json.dumps(dbs_invalid, indent=2)
print json.dumps(phedex_presence, indent=2)
## need to be going and taking an eye
assistance_tags.add('invalidfiles')
if not bypass_checks:
#sub_assistance+="-invalidfiles"
is_closing = False
## put that heavy part at the end
## duplication check
duplications = {}
if is_closing or bypass_checks:
print "starting duplicate checker for",wfo.name
for output in wfi.request['OutputDatasets']:
print "\tchecking",output
duplications[output] = True
try:
duplications[output] = dbs3Client.duplicateRunLumi( output , skipInvalid=True, verbose=True)
except:
try:
duplications[output] = dbs3Client.duplicateRunLumi( output , skipInvalid=True, verbose=True)
except:
print "was not possible to get the duplicate count for",output
is_closing=False
if any(duplications.values()) and not options.ignoreduplicates:
print wfo.name,"has duplicates"
print json.dumps(duplications,indent=2)
## hook for making file invalidation ?
## it shouldn't be allowed to bypass it
assistance_tags.add('duplicates')
is_closing = False
## for visualization later on
if not wfo.name in fDB.record:
#print "adding",wfo.name,"to close out record"
fDB.record[wfo.name] = {
'datasets' :{},
'name' : wfo.name,
'closeOutWorkflow' : None,
}
fDB.record[wfo.name]['closeOutWorkflow'] = is_closing
fDB.record[wfo.name]['priority'] = wfi.request['RequestPriority']
fDB.record[wfo.name]['prepid'] = wfi.request['PrepID']
for output in wfi.request['OutputDatasets']:
if not output in fDB.record[wfo.name]['datasets']: fDB.record[wfo.name]['datasets'][output] = {}
rec = fDB.record[wfo.name]['datasets'][output]
rec['percentage'] = float('%.2f'%(percent_completions[output]*100))
rec['duplicate'] = duplications[output] if output in duplications else 'N/A'
rec['phedexReqs'] = float('%.2f'%any_presence[output][custodial_presences[output][0]][1]) if len(custodial_presences[output])!=0 else 'N/A'
rec['closeOutDataset'] = is_closing
rec['transPerc'] = float('%.2f'%any_presence[output][ disk_copies[output][0]][1]) if len(disk_copies[output])!=0 else 'N/A'
rec['correctLumis'] = int(events_per_lumi[output]) if (events_per_lumi[output] > lumi_upper_limit[output]) else True
rec['missingSubs'] = False if len(custodial_locations[output])==0 else ','.join(list(set(custodial_locations[output])))
rec['dbsFiles'] = dbs_presence[output]
rec['dbsInvFiles'] = dbs_invalid[output]
rec['phedexFiles'] = phedex_presence[output]
rec['acdc'] = "%d / %d"%(len(acdc),len(acdc+acdc_inactive))
now = time.gmtime()
rec['timestamp'] = time.mktime(now)
rec['updated'] = time.asctime(now)+' (GMT)'
## and move on
if is_closing:
## toggle status to closed-out in request manager
print "setting",wfo.name,"closed-out"
if not options.test:
if wfo.wm_status in ['closed-out','announced','normal-archived']:
print wfo.name,"is already",wfo.wm_status,"not trying to closed-out and assuming it does"
res = None
else:
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
print "close out answer",res
if not res in ["None",None]:
print "try to get the current status again"
wfi_bis = workflowInfo(url, wfo.name)
if wfi_bis.request['RequestStatus'] == 'closed-out':
print "the request did toggle to closed-out"
res = None
if not res in ["None",None]:
print "retrying to closing out"
print res
res = reqMgrClient.closeOutWorkflowCascade(url, wfo.name)
if res in [None,"None"]:
wfo.status = 'close'
session.commit()
if use_mcm and force_by_mcm:
## shoot large on all prepids, on closing the wf
for pid in pids:
mcm.delete('/restapi/requests/forcecomplete/%s'%pid)
else:
print "could not close out",wfo.name,"will try again next time"
else:
## full known list
#recovering # has active ACDC
##OUT #recovered #had inactive ACDC
#recovery #not over the pass bar
#over100 # over 100%
#biglumi # has a big lumiblock
#parentcustodial # the parent does not have a valid subscription yet
#custodial # has had the transfer made, is waiting for a valid custodial subscription to appear
#filemismatch # there is a dbs/phedex mismatch
#duplicates #a lumi section is there twice
## manual is not added yet, and should be so by recoveror
print wfo.name,"was tagged with :",list(assistance_tags)
if 'recovering' in assistance_tags:
## if active ACDC, being under threshold, filemismatch do not matter
assistance_tags = assistance_tags - set(['recovery','filemismatch'])
if 'recovery' in assistance_tags and 'recovered' in assistance_tags:
## should not set -recovery to anything that add ACDC already
assistance_tags = assistance_tags - set(['recovery','recovered'])
## straight to manual
assistance_tags.add('manual')
## that means there is something that needs to be done acdc, lumi invalidation, custodial, name it
print wfo.name,"needs assistance with",",".join( assistance_tags )
print wfo.name,"existing conditions",",".join( existing_assistance_tags )
#########################################
##### notification to requester #########
go_notify=False
if assistance_tags and not 'manual' in existing_assistance_tags and existing_assistance_tags != assistance_tags:
go_notify=True
if go_notify:
#if wfo.name in already_notified:
# print "double notification"
# sendEmail('double notification','please take a look at %s'%(wfo.name))
#else:
# already_notified.append( wfo.name )
detailslink = 'https://cmsweb.cern.ch/reqmgr/view/details/%s'
perflink = 'https://cmsweb.cern.ch/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s'%(wfo.name)
splitlink = 'https://cmsweb.cern.ch/reqmgr/view/splitting/%s'%(wfo.name)
## notify templates
messages= {
'recovery': 'Samples completed with missing statistics:\n%s\n%s '%( '\n'.join(['%.2f %% complete for %s'%(percent_completions[output]*100, output) for output in wfi.request['OutputDatasets'] ] ), perflink ),
'biglumi': 'Samples completed with large luminosity blocks:\n%s\n%s '%('\n'.join(['%d > %d for %s'%(events_per_lumi[output], lumi_upper_limit[output], output) for output in wfi.request['OutputDatasets'] if (events_per_lumi[output] > lumi_upper_limit[output])]), splitlink),
'duplicates': 'Samples completed with duplicated luminosity blocks:\n%s\n'%( '\n'.join(['%s'%output for output in wfi.request['OutputDatasets'] if output in duplications and duplications[output] ] ) ),
'filemismatch': 'Samples completed with inconsistency in DBS/Phedex',
#'manual' : 'Workflow completed and requires manual checks by Ops',
}
content = "The request PREPID (WORKFLOW) is facing issue in production.\n"
motive = False
for case in messages:
if case in assistance_tags:
content+= "\n"+messages[case]+"\n"
motive = True
content += "You are invited to check, while this is being taken care of by Comp-Ops.\n"
content += "This is an automated message from Comp-Ops.\n"
items_notified = set()
if use_mcm and motive:
wfi.notifyRequestor( content , mcm = mcm)
#########################################
## logic to set the status further
if assistance_tags:
new_status = 'assistance-'+'-'.join(sorted(assistance_tags) )
else:
new_status = 'assistance'
## case where the workflow was in manual from recoveror
if not 'manual' in wfo.status or new_status!='assistance-recovery':
wfo.status = new_status
if not options.test:
print "setting",wfo.name,"to",wfo.status
session.commit()
else:
print "current status is",wfo.status,"not changing to anything"
#open('already_notifified.json','w').write( json.dumps( already_notified , indent=2))
fDB.html()
if not spec:
#sendEmail("fresh assistance status available","Fresh status are available at https://cmst2.web.cern.ch/cmst2/unified/assistance.html",destination=['*****@*****.**'])
#it's a bit annoying
pass
## custodial requests
print "Custodials"
print json.dumps(custodials, indent=2)
for site in custodials:
print ','.join(custodials[site]),'=>',site
if not options.test:
result = makeReplicaRequest(url, site, list(set(custodials[site])),"custodial copy at production close-out",custodial='y',priority='low', approve = (site in SI.sites_auto_approve) )
print result
print "Transfers"
print json.dumps(transfers, indent=2)
## replicas requests
for site in transfers:
print ','.join(transfers[site]),'=>',site
if not options.test:
result = None
#result = makeReplicaRequest(url, site, list(set(transfers[site])),"copy to disk at production close-out")
print result
print "File Invalidation"
print invalidations
def batchor( url ):
UC = unifiedConfiguration()
SI = global_SI()
## get all workflows in assignment-approved with SubRequestType = relval
all_wfs = []
for user in UC.get("user_relval"):
all_wfs.extend( getWorkflows(url, 'assignment-approved', details=True, user=user, rtype='TaskChain') )
wfs = filter( lambda r :r['SubRequestType'] == 'RelVal' if 'SubRequestType' in r else False, all_wfs)
## need a special treatment for those
hi_wfs = filter( lambda r :r['SubRequestType'] == 'HIRelVal' if 'SubRequestType' in r else False, all_wfs)
by_campaign = defaultdict(set)
by_hi_campaign = defaultdict(set)
for wf in wfs:
print "Relval:",wf['RequestName'], wf['Campaign']
#by_campaign[wf['Campaign']].add( wf['RequestName'] )
by_campaign[wf['Campaign']].add( wf['PrepID'] )
for wf in hi_wfs:
print "HI Relval:",wf['RequestName'], wf['Campaign']
#by_hi_campaign[wf['Campaign']].add( wf['RequestName'] )
by_hi_campaign[wf['Campaign']].add( wf['PrepID'] )
default_setup = {
"go" :True,
"parameters" : {
"SiteWhitelist": [ "T1_US_FNAL" ],
"MergedLFNBase": "/store/relval",
"Team" : "relval",
"NonCustodialGroup" : "RelVal"
},
"custodial" : "T1_US_FNAL_MSS",
"custodial_override" : ["DQMIO"],
"phedex_group" : "RelVal",
"lumisize" : -1,
"fractionpass" : 0.0,
"maxcopies" : 1
}
default_hi_setup = copy.deepcopy( default_setup )
add_on = {}
batches = json.loads( open('batches.json').read() )
relval_routing = UC.get('relval_routing')
def pick_one_site( p):
## modify the parameters on the spot to have only one site
if "parameters" in p and "SiteWhitelist" in p["parameters"] and len(p["parameters"]["SiteWhitelist"])>1:
choose_from = list(set(p["parameters"]["SiteWhitelist"]) & set(SI.sites_ready))
picked = random.choice( choose_from )
print "picked",picked,"from",choose_from
p["parameters"]["SiteWhitelist"] = [picked]
for campaign in by_campaign:
if campaign in batches: continue
## get a bunch of information
setup = copy.deepcopy( default_setup )
for key in relval_routing:
if key in campaign:
## augment with the routing information
augment_with = relval_routing[key]
print "Modifying the batch configuration because of keyword",key
print "with",augment_with
setup = deep_update( setup, augment_with )
#if 'cc7' in campaign: setup["parameters"]["SiteWhitelist"] = ["T2_US_Nebraska"]
pick_one_site( setup )
add_on[campaign] = setup
sendLog('batchor','Adding the relval campaigns %s with parameters \n%s'%( campaign, json.dumps( setup, indent=2)),level='critical')
if not campaign in batches: batches[campaign] = []
batches[campaign] = list(set(list(copy.deepcopy( by_campaign[campaign] )) + batches[campaign] ))
for campaign in by_hi_campaign:
if campaign in batches: continue
## get a bunch of information
setup = copy.deepcopy( default_hi_setup )
hi_site = random.choice(["T1_DE_KIT","T1_FR_CCIN2P3"])
setup["parameters"]["SiteWhitelist"]=[ hi_site ]
#setup["parameters"]["SiteWhitelist"]=["T1_DE_KIT","T1_FR_CCIN2P3"]
pick_one_site( setup )
add_on[campaign] = setup
sendLog('batchor','Adding the HI relval campaigns %s with parameters \n%s'%( campaign, json.dumps( setup, indent=2)),level='critical')
if not campaign in batches: batches[campaign] = []
batches[campaign] = list(set(list(copy.deepcopy( by_hi_campaign[campaign] )) + batches[campaign] ))
open('batches.json','w').write( json.dumps( batches , indent=2 ) )
## open the campaign configuration
campaigns = json.loads( open('campaigns.relval.json').read() )
## protect for overwriting ??
for new_campaign in list(set(add_on.keys())-set(campaigns.keys())):
## this is new, and can be announced as such
print new_campaign,"is new stuff"
subject = "Request of RelVal samples batch %s"% new_campaign
text="""Dear all,
A new batch of relval workflows was requested.
Batch ID:
%s
Details of the workflows:
https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?campaign=%s
This is an automated message"""%( new_campaign,
new_campaign,
)
print subject
print text
to = ['*****@*****.**']
sendEmail(subject, text, destination=to)
sendLog('batchor',text, level='critical')
## go through all existing campaigns and remove the ones not in use anymore ?
for old_campaign in campaigns.keys():
all_in_batch = getWorkflowByCampaign(url, old_campaign, details=True)
is_batch_done = all(map(lambda s : not s in ['completed','running-open','running-closed','acquired','assigned','assignment-approved'], [wf['RequestStatus']for wf in all_in_batch]))
## check all statuses
if is_batch_done:
#print "batch",old_campaign,"can be closed or removed if necessary"
#campaigns[old_campaign]['go'] = False ## disable
campaigns.pop( old_campaign ) ## or just drop it all together ?
print "batch",old_campaign," configuration was removed"
## merge all anyways
campaigns.update( add_on )
## write it out for posterity
open('campaigns.json.updated','w').write(json.dumps( campaigns , indent=2))
## read back
rread = json.loads(open('campaigns.json.updated').read())
os.system('mv campaigns.json.updated campaigns.relval.json')
def assignor(url, specific=None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = siteInfo()
NLI = newLockInfo()
n_assigned = 0
n_stalled = 0
wfos = []
if specific or options.early:
wfos.extend(
session.query(Workflow).filter(
Workflow.status == 'considered').all())
wfos.extend(
session.query(Workflow).filter(Workflow.status == 'staging').all())
if specific:
wfos.extend(
session.query(Workflow).filter(
Workflow.status == 'considered-tried').all())
wfos.extend(
session.query(Workflow).filter(Workflow.status == 'staged').all())
dataset_endpoints = json.loads(
open('%s/dataset_endpoints.json' % monitor_dir).read())
max_per_round = UC.get('max_per_round').get('assignor', None)
max_cpuh_block = UC.get('max_cpuh_block')
random.shuffle(wfos)
for wfo in wfos:
if options.limit and (n_stalled + n_assigned) > options.limit:
break
if max_per_round and (n_stalled + n_assigned) > max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name, specific.split(','))):
continue
#if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo(url, wfo.name)
wfh.sendLog('assignor', "%s to be assigned" % wfo.name)
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput, primary, parent, secondary,
sites_allowed) = wfh.getSiteWhiteList()
## check if by configuration we gave it a GO
no_go = False
if not wfh.go(log=True) and not options.go:
n_stalled += 1
no_go = True
allowed_secondary = set()
for campaign in wfh.getCampaigns():
if campaign in CI.campaigns and 'secondaries' in CI.campaigns[
campaign]:
allowed_secondary.update(CI.campaigns[campaign]['secondaries'])
if (secondary and allowed_secondary) and (
set(secondary) & allowed_secondary != set(secondary)):
wfh.sendLog(
'assignor', '%s is not an allowed secondary' %
(', '.join(set(secondary) - allowed_secondary)))
#sendEmail('secondary not allowed','%s is not an allowed secondary'%( ', '.join(set(secondary)-allowed_secondary)))
sendLog('assignor',
'%s is not an allowed secondary' %
(', '.join(set(secondary) - allowed_secondary)),
level='critical')
if not options.go:
n_stalled += 1
no_go = True
if no_go:
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] != 'assignment-approved':
if not options.test:
wfh.sendLog('assignor',
"setting %s away and skipping" % wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name, wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version = wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor', "cannot decide on version number")
n_stalled += 1
wfo.status = 'trouble'
session.commit()
continue
original_sites_allowed = copy.deepcopy(sites_allowed)
wfh.sendLog('assignor', "Site white list %s" % sorted(sites_allowed))
override_sec_location = CI.get(wfh.request['Campaign'],
'SecondaryLocation', [])
blocks = []
if 'BlockWhitelist' in wfh.request:
blocks = wfh.request['BlockWhitelist']
if 'RunWhitelist' in wfh.request and wfh.request['RunWhitelist']:
## augment with run white list
for dataset in primary:
blocks = list(
set(blocks + getDatasetBlocks(
dataset, runs=wfh.request['RunWhitelist'])))
wfh.sendLog('assignor', "Allowed %s" % sorted(sites_allowed))
secondary_locations = None
primary_aaa = options.primary_aaa
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns and 'primary_AAA' in CI.campaigns[
wfh.request['Campaign']]:
primary_aaa = primary_aaa or CI.campaigns[
wfh.request['Campaign']]['primary_AAA']
secondary_aaa = options.secondary_aaa
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns and 'secondary_AAA' in CI.campaigns[
wfh.request['Campaign']]:
secondary_aaa = secondary_aaa or CI.campaigns[
wfh.request['Campaign']]['secondary_AAA']
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
sendEmail("tempting to pass sec location check",
"but we cannot yet IMO")
#pass
if secondary_aaa:
#just continue without checking
continue
presence = getDatasetPresence(url, sec)
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [
site for (site, (there, frac)) in presence.items()
if frac > 98.
]
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations == None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(
set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [
site for site in sites_allowed
if SI.CE_to_SE(site) in one_secondary_locations
]
wfh.sendLog(
'assignor', "From secondary requirement, now Allowed%s" %
sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy(
sites_allowed
) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy(sites_allowed)
sites_with_data = copy.deepcopy(sites_allowed)
sites_with_any_data = copy.deepcopy(sites_allowed)
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
endpoints = set()
for prim in list(primary):
if prim in dataset_endpoints:
print "endpoints from stagor", dataset_endpoints[prim]
endpoints.update(dataset_endpoints[prim])
set_lfn = getLFNbase(prim)
presence = getDatasetPresence(url, prim, only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(
url,
prim,
sites=[SI.CE_to_SE(site) for site in sites_allowed],
only_blocks=blocks)
#sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
#sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in [
psite for (psite, (there, frac)) in presence.items()
if there
]
]
sites_with_data = [
site for site in sites_with_data if SI.CE_to_SE(site) in
[psite for (psite, frac) in presence.items() if frac[1] > 90.]
]
sites_with_any_data = [
site for site in sites_with_any_data
if SI.CE_to_SE(site) in presence.keys()
]
wfh.sendLog(
'assignor', "Holding the data but not allowed %s" % sorted(
list(
set([
se_site for se_site in presence.keys()
if not SI.SE_to_CE(se_site) in sites_allowed
]))))
if primary_locations == None:
primary_locations = presence.keys()
else:
primary_locations = list(
set(primary_locations) & set(presence.keys()))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites = []
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in
list((set(secondary_locations) & set(primary_locations)) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
elif primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site) for site in list(
set(primary_locations) -
set([SI.CE_to_SE(site) for site in sites_allowed]))
]
else:
opportunistic_sites = []
wfh.sendLog(
'assignor', "We could be running in addition at %s" %
sorted(opportunistic_sites))
if any(
[osite in SI.sites_not_ready
for osite in opportunistic_sites]):
wfh.sendLog(
'assignor', "One of the usable site is in downtime %s" % ([
osite in SI.sites_not_ready
for osite in opportunistic_sites
]))
down_time = True
## should this be send back to considered ?
## should be 2 but for the time-being let's lower it to get things going
copies_wanted, cpuh = wfh.getNCopies()
wfh.sendLog('assignor', "we need %s CPUh" % cpuh)
if cpuh > max_cpuh_block and not options.go:
#sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog(
'assignor',
'%s requires a large numbr of CPUh %s , not assigning, please check with requester'
% (wfo.name, cpuh),
level='critical')
wfh.sendLog(
'assignor',
"Requiring a large number of CPUh %s, not assigning" % cpuh)
continue
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns and 'maxcopies' in CI.campaigns[
wfh.request['Campaign']]:
copies_needed_from_campaign = CI.campaigns[
wfh.request['Campaign']]['maxcopies']
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(
1, copies_wanted -
less_copies_than_requested) # take one out for the efficiency
else:
## find out whether there is a site in the whitelist, that is lacking jobs and reduce to 1 copy needed to get things going
pass
wfh.sendLog('assignor',
"needed availability fraction %s" % copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
if available_fractions and not all([
available >= copies_wanted
for available in available_fractions.values()
]):
not_even_once = not all([
available >= 1. for available in available_fractions.values()
])
wfh.sendLog(
'assignor',
"The input dataset is not available %s times, only %s" %
(copies_wanted, available_fractions.values()))
if down_time and not options.go and not options.early:
wfo.status = 'considered'
session.commit()
wfh.sendLog(
'assignor',
"sending back to considered because of site downtime, instead of waiting"
)
#sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
sendLog(
'assignor',
'%s is not sufficiently available, due to down time of a site in the whitelist. sending back to considered.'
% (wfo.name),
level='delay')
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known and not options.limit and not options.go and not options.early and not options.partial:
wfh.sendLog(
'assignor',
"cannot be assigned, %s is not sufficiently available.\n %s"
% (wfo.name, json.dumps(available_fractions)))
sendEmail(
"cannot be assigned",
"%s is not sufficiently available.\n %s" %
(wfo.name, json.dumps(available_fractions)))
known.append(wfo.name)
open('cannot_assign.json',
'w').write(json.dumps(known, indent=2))
n_stalled += 1
if options.early:
if wfo.status == 'considered':
wfh.sendLog('assignor', "setting considered-tried")
wfo.status = 'considered-tried'
session.commit()
else:
print "tried but status is", wfo.status
if options.partial:
print "Will move on with partial locations"
else:
continue
## default back to white list to original white list with any data
print "Allowed", sorted(sites_allowed)
if primary_aaa:
sites_allowed = initial_sites_allowed
options.TrustSitelists = True
wfh.sendLog(
'assignor', "Selected to read primary through xrootd %s" %
sorted(sites_allowed))
else:
sites_allowed = sites_with_any_data
wfh.sendLog('assignor',
"Selected for any data %s" % sorted(sites_allowed))
if secondary_aaa:
options.TrustPUSitelists = True
wfh.sendLog(
'assignor', "Reading secondary through xrootd from %s" %
sorted(sites_allowed))
### check on endpoints for on-going transfers
if endpoints and options.partial:
sites_allowed = list(
set(sites_allowed + [SI.SE_to_CE(s) for s in endpoints]))
print "with added endpoints", sorted(sites_allowed)
if not len(sites_allowed):
wfh.sendLog('assignor', "cannot be assign with no matched sites")
sendLog('assignor',
'%s has no whitelist' % wfo.name,
level='critical')
n_stalled += 1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [
SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])
]
wfh.sendLog('assignor', "Placing the output on %s" % sites_out)
parameters = {
'SiteWhitelist': sites_allowed,
'NonCustodialSites': sites_out,
'AutoApproveSubscriptionSites': list(set(sites_out)),
'AcquisitionEra': wfh.acquisitionEra(),
'ProcessingString': wfh.processingString(),
'MergedLFNBase': set_lfn,
'ProcessingVersion': version,
}
## plain assignment here
team = 'production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
if False and 'T2_CH_CERN' in parameters['SiteWhitelist']:
## add some check on
### the amount pending to HLT
### the size of the request
### the priority of the request (maybe not if we decide to overflow during runs)
parameters['SiteWhitelist'] = ['T2_CH_CERN_HLT']
team = 'hlt'
## reduce the splitting by factor of 4, regardless of type of splitting
sendEmail("sending work to HLT",
"%s was assigned to HLT" % wfo.name)
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v = getattr(options, key)
if v != None:
if type(v) == str and ',' in v:
parameters[key] = filter(None, v.split(','))
else:
parameters[key] = v
if lheinput:
## throttle reading LHE article
wfh.sendLog('assignor',
'Setting the number of events per job to 500k max')
parameters['EventsPerJob'] = 500000
## pick up campaign specific assignment parameters
parameters.update(CI.parameters(wfh.request['Campaign']))
if not options.test:
parameters['execute'] = True
split_check = wfh.checkWorkflowSplitting()
if split_check != True:
parameters.update(split_check)
if 'EventBased' in split_check.values():
wfh.sendLog('assignor', "Falling back to event splitting.")
#sendEmail("Fallback to EventBased","the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"%wfo.name)
sendLog(
'assignor',
'the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting'
% wfo.name,
level='critical')
elif 'EventsPerJob' in split_check.values():
wfh.sendLog('assignor',
"Modifying the number of job per event")
#sendEmail("Modifying the job per events","the workflow %s is too heavy in number of jobs explosion"%wfo.name)
sendLog(
'assignor',
"the workflow %s is too heavy in number of jobs explosion"
% wfo.name,
level='critical')
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi) / float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents / (reqJobs * 1.4))
lumisPerJob = int(eventsPerJob / eventsPerLumi)
if lumisPerJob == 0:
#sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog('assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob),
level='critical')
wfh.sendLog(
'assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl[
'events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl[
'avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
#sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog('assignor',
"%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob),
level='critical')
wfh.sendLog(
'assignor', "%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
#sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog(
'assignor',
"leaving splitting untouched for %s, please check on %s"
% (pstring, wfo.name),
level='critical')
wfh.sendLog(
'assignor',
"leaving splitting untouched for PU_RD*, please check."
)
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned += 1
wfh.sendLog(
'assignor', "Properly assigned\n%s" %
(json.dumps(parameters, indent=2)))
try:
## refetch information and lock output
new_wfi = workflowInfo(url, wfo.name)
(_, prim, _, sec) = new_wfi.getIO()
for secure in list(prim) + list(
sec) + new_wfi.request['OutputDatasets']:
## lock all outputs flat
NLI.lock(secure)
#for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output", str(e))
else:
print "ERROR could not assign", wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor', "Assigned %d Stalled %s" % (n_assigned, n_stalled))
def actor(url,options=None):
if userLock('actor'): return
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
# CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
# Need to look at the actions page https://vocms0113.cern.ch:80/getaction (can add ?days=20) and perform any actions listed
try:
action_list = json.loads(os.popen('curl -s -k https://vocms0113.cern.ch:80/getaction?days=15').read())
## now we have a list of things that we can take action on
except:
print "Not able to load action list :("
sendLog('actor','Not able to load action list', level='critical')
return
print action_list
if not action_list:
print "EMPTY!"
return
for wfname in action_list:
print '-'*100
print "Looking at",wfname,"for recovery options"
to_clone = False
to_acdc = False
for key in action_list[wfname]:
if key == 'Parameters':
tasks = action_list[wfname][key]
elif key == 'Action' and action_list[wfname][key] == 'acdc':
print "Going to create ACDCs for ", wfname
to_acdc = True
elif key == 'Action' and action_list[wfname][key] == 'clone':
print "Going to clone ", wfname
to_clone = True
if not to_acdc and not to_clone:
sendLog('actor','Action submitted for something other than acdc and clone for workflow %s'%wfname,level='critical')
print "Can only do acdcs and clones! Skipping workflow ",wfname
continue
if not tasks:
sendLog('actor','Empty action submitted for workflow %s'%wfname,level='critical')
print "Moving on. Parameters is blank for " + wfname
continue
wfi = workflowInfo(url, wfname)
recover = True
message_to_ops = ""
message_to_user = ""
#===========================================================
if to_clone and options.do:
print "Let's try kill and clone: "
wfi.sendLog('actor','Going to clone %s'%wfname)
results=[]
datasets = set(wfi.request['OutputDatasets'])
comment=""
if 'comment' in tasks: comment = ", reason: "+ tasks['comment']
wfi.sendLog('actor',"invalidating the workflow by traffic controller %s"%comment)
#Reject all workflows in the family
#first reject the original workflow.
reqMgrClient.invalidateWorkflow(url, wfi.request['RequestName'], current_status=wfi.request['RequestStatus'], cascade=False)
#Then reject any ACDCs associated with that workflow
if 'ACDCs' in action_list[wfname]:
children = action_list[wfname]['ACDCs']
for child in children:
wfi.sendLog('actor',"rejecting %s"%child)
wfi_acdc = workflowInfo(url, child)
reqMgrClient.invalidateWorkflow(url, wfi_acdc.request['RequestName'], current_status=wfi_acdc.request['RequestStatus'], cascade=False)
datasets.update( wfi_acdc.request['OutputDatasets'] )
#Invalidate all associated output datasets
for dataset in datasets:
results.append( setDatasetStatus(dataset, 'INVALID') )
if all(map(lambda result : result in ['None',None,True],results)):
wfi.sendLog('actor',"%s and children are rejected"%wfname)
cloned = None
try:
cloned = singleClone(url, wfname, tasks, comment, options.do)
except:
sendLog('actor','Failed to create clone for %s! Check logs for more information. Action will need to be resubmitted.'%wfname,level='critical')
wfi.sendLog('actor','Failed to create clone for %s!'%wfname)
remove_action(wfname)
if not cloned:
recover = False
wfi.sendLog('actor','Failed to create clone for %s!'%wfname)
sendLog('actor','Failed to create clone for %s!'%wfname,level='critical')
else:
wfi.sendLog('actor',"Workflow %s cloned"%wfname)
#===========================================================
elif to_acdc:
if 'AllSteps' in tasks:
allTasksDefaults = tasks['AllSteps']
tasks.pop('AllSteps')
for setting in allTasksDefaults:
for task in tasks:
if setting in tasks[task]:
tasks[task][setting] = allTasksDefaults[setting]
else:
tasks[task].append({setting:allTasksDefaults[setting]})
print "Tasks is "
print tasks
all_tasks = wfi.getAllTasks()
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
try:
WMErr = wfi.getWMErrors()
# print WMErr
except:
sendLog('actor','Cannot create ACDCS for %s because WMErr cannot be reached.'%wfname,level='critical')
continue
if not WMErr:
sendLog('actor','Cannot create ACDCS for %s because WMErr is blank.'%wfname,level='critical')
print "Moving on. WMErr is blank"
continue
try:
where_to_run, missing_to_run,missing_to_run_at = wfi.getRecoveryInfo()
print "Where to run = "
print where_to_run
except:
sendLog('actor','Cannot create ACDCS for %s because recovery info cannot be found.'%wfname,level='critical')
print "Moving on. Cannot access recovery info for " + wfname
continue
if not where_to_run:
sendLog('actor','Cannot create ACDCS for %s because site list cannot be found.'%wfname,level='critical')
print "Moving on. where to run is blank"
continue
message_to_ops = ""
message_to_user = ""
num_tasks_to_recover = 0
for task in WMErr:
if 'LogCollect' in task: continue
if 'Cleanup' in task: continue
if not 'jobfailed' in WMErr[task]:
continue
else:
num_tasks_to_recover += 1
# print "Task to recover: " + task
if not num_tasks_to_recover:
print "\tno error for",wfname
# recover = False
if 'LheInputFiles' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
sendLog('actor','Cannot create ACDCS for %s because it is a pLHE workflow.'%wfname,level='critical')
print "We don't try to recover pLHE. Moving on."
recover = False
# sendEmail('cannot submit action', '%s is a pLHE workflow. We do not try to recover pLHE'%wfname)
# if wfi.request['RequestType'] in ['ReReco']:
# recover= False
# print 'cannot submit action. ReReco'
# sendEmail('cannot submit action', '%s is request type ReReco'%wfname)
recovering = set()
for task in tasks:
assign_to_sites = set()
print "Task names is " + task
fulltaskname = '/' + wfname + '/' + task
# print "Full task name is " + fulltaskname
wrong_task = False
for task_info in all_tasks:
if fulltaskname == task_info.pathName:
if task_info.taskType not in ['Processing','Production','Merge']:
wrong_task=True
wfi.sendLog('actor', "Skipping task %s because the taskType is %s. Can only ACDC Processing, Production, or Merge tasks"%( fulltaskname, task_info.taskType))
if wrong_task:
continue
print tasks[task]
actions = tasks[task]
for action in actions:
if action.startswith('sites'):
if type(actions[action]) != list:
assign_to_sites=[SI.SE_to_CE(actions[action])]
else:
assign_to_sites=list(set([SI.SE_to_CE(site) for site in actions[action]]))
# if action.startswith('mem') and actions[action] != "" and actions[action] != 'Same' and wfi.request['RequestType'] in ['TaskChain']:
# recover = False;
# print "Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname
# wfi.sendLog('actor',"Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname)
if not 'sites' in actions:
assign_to_sites = list(set([SI.SE_to_CE(site) for site in where_to_run[task]]))
print "Found",sorted(assign_to_sites),"as sites where to run the ACDC at, from the acdc doc of ",wfname
print "Going to run at",sorted(assign_to_sites)
if recover:
print "Initiating recovery"
acdc = singleRecovery(url, fulltaskname, wfi.request, actions, do = options.do)
if not acdc:
if options.do:
if recovering:
print wfname + " has been partially ACDC'ed. Needs manual attention."
sendLog('actor', "%s has had %s/%s recoveries %s only"%( wfname, len(recovering), num_tasks_to_recover, list(recovering)), level='critical')
wfi.sendLog('actor', "%s has had %s/%s recoveries %s only"%( wfname, len(recovering), num_tasks_to_recover, list(recovering)))
break
else:
print wfname + " failed recovery once"
recover = False
break
else:
print "no action to take further"
# sendLog('recoveror', "ACDC for %s can be done automatically"% wfname, level='critical')
continue
else: #ACDC was made correctly. Now we have to assign it.
wfi.sendLog('actor','ACDC created for task %s. Actions taken \n%s'%(fulltaskname,list(actions)))
team = wfi.request['Teams'][0]
parameters={
'SiteWhitelist' : sorted(assign_to_sites),
'AcquisitionEra' : wfi.acquisitionEra(),
'ProcessingString' : wfi.processingString(),
'MergedLFNBase' : wfi.request['MergedLFNBase'],
'ProcessingVersion' : wfi.request['ProcessingVersion'],
}
## hackery for ACDC merge assignment
if wfi.request['RequestType'] == 'TaskChain' and 'Merge' in task.split('/')[-1]:
parameters['AcquisitionEra'] = None
parameters['ProcessingString'] = None
## xrootd setttings on primary and secondary
if 'xrootd' in actions:
if actions['xrootd'] == 'enabled':
print "Going to assign via xrootd"
parameters['TrustSitelists'] = True
elif actions['xrootd'] == 'disabled':
parameters['TrustSitelists'] = False
elif ('TrustSitelists' in wfi.request and wfi.request['TrustSitelists']=='true'):
parameters['TrustSitelists'] = True
else:
parameters['TrustSitelists'] = False
if 'TrustPUSitelists' in wfi.request and wfi.request['TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
if options.ass:
print "really doing the assignment of the ACDC",acdc
parameters['execute']=True
wfi.sendLog('actor',"%s was assigned for recovery"% acdc)
else:
print "no assignment done with this ACDC",acdc
sendLog('actor',"%s needs to be assigned"%(acdc), level='critical')
continue
# print parameters
result = reqMgrClient.assignWorkflow(url, acdc, team, parameters)
if not result:
print acdc,"was not assigned"
sendLog('actor',"%s needs to be assigned"%(acdc), level='critical')
else:
recovering.add( acdc )
wfi.sendLog('actor',"ACDCs created for %s"%wfname)
#===========================================================
if recover and options.do:
remove_action(wfname)
if message_to_user:
print wfname,"to be notified to user(DUMMY)",message_to_user
if message_to_ops:
print 'message'
#sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**'])
# sendLog('recoveror',message_to_ops,level='warning')
return
def htmlor(caller=""):
cache = getWorkflows("cmsweb.cern.ch", "assignment-approved", details=True)
cache.extend(getWorkflows("cmsweb.cern.ch", "running-open", details=True))
cache.extend(getWorkflows("cmsweb.cern.ch", "running-closed", details=True))
def getWL(wfn):
cached = filter(lambda d: d["RequestName"] == wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad("cmsweb.cern.ch", wfn)
return wl
def wfl(wf, view=False, p=False, ms=False, within=False, ongoing=False, status=False, update=False):
wfn = wf.name
wfs = wf.wm_status
wl = None
pid = None
pids = filter(lambda seg: seg.count("-") == 2, wf.name.split("_"))
if len(pids):
pids = pids[:1]
pid = pids[0]
if not pids:
wl = getWL(wf.name)
pids = getPrepIDs(wl)
pid = pids[0]
text = ", ".join(
[
# wfn,
'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">%s</a>' % (wfn, wfn),
"(%s) <br>" % wfs,
]
)
text += ", ".join(
[
'<a href="https://cmsweb.cern.ch/reqmgr/view/details/%s" target="_blank">dts</a>' % wfn,
'<a href=https://cmsweb.cern.ch/reqmgr/view/showWorkload?requestName=%s target="_blank">wkl</a>' % wfn,
'<a href="https://cmsweb.cern.ch/couchdb/reqmgr_workload_cache/%s" target="_blank">wfc</a>' % wfn,
'<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/request?requestName=%s" target="_blank">dwkc</a>' % wfn,
'<a href="https://cmsweb.cern.ch/reqmgr/view/splitting/%s" target="_blank">spl</a>' % wfn,
'<a href="https://cms-pdmv.cern.ch/stats/?RN=%s" target="_blank">vw</a>' % wfn,
'<a href="https://cms-pdmv.cern.ch/stats/restapi/get_one/%s" target="_blank">vwo</a>' % wfn,
'<a href="https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=full&reverse=0&reverse=1&npp=20&subtext=%s&sall=q" target="_blank">elog</a>'
% pid,
'<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank">pv</a>' % wfn,
'<a href="https://cmsweb.cern.ch/reqmgr/reqMgr/outputDatasetsByRequestName/%s" target="_blank">out</a>'
% wfn,
'<a href="closeout.html#%s" target="_blank">clo</a>' % wfn,
'<a href="statuses.html#%s" target="_blank">st</a>' % wfn,
'<a href="https://cmsweb.cern.ch/couchdb/workloadsummary/_design/WorkloadSummary/_show/histogramByWorkflow/%s" target="_blank">perf</a>'
% wfn,
]
)
if within and (not view or wfs == "completed"):
wl = getWL(wfn)
dataset = None
if "InputDataset" in wl:
dataset = wl["InputDataset"]
if "Task1" in wl and "InputDataset" in wl["Task1"]:
dataset = wl["Task1"]["InputDataset"]
if dataset:
text += ", ".join(
[
"",
"<a href=https://cmsweb.cern.ch/das/request?input=%s target=_blank>input</a>" % dataset,
"<a href=https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions#state=create_since=0;filter=%s target=_blank>sub</a>"
% dataset,
"<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/subscriptions?dataset=%s&collapse=n target=_blank>ds</a>"
% dataset,
"<a href=https://cmsweb.cern.ch/phedex/datasvc/xml/prod/blockreplicas?dataset=%s target=_blank>rep</a>"
% dataset,
]
)
if p:
cached = filter(lambda d: d["RequestName"] == wfn, cache)
if cached:
wl = cached[0]
else:
wl = getWorkLoad("cmsweb.cern.ch", wfn)
text += ", (%s)" % (wl["RequestPriority"])
pass
if pid:
if ms:
mcm_s = json.loads(
os.popen(
"curl https://cms-pdmv.cern.ch/mcm/public/restapi/requests/get_status/%s --insecure" % pid
).read()
)[pid]
text += ', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm (%s)</a>' % (
pid,
mcm_s,
)
else:
text += ', <a href="https://cms-pdmv.cern.ch/mcm/requests?prepid=%s" target="_blank">mcm</a>' % (pid)
text += (
', <a href="https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?prep_id=%s" target="_blank">ac</a>'
% (pid)
)
if status:
if wf.status.startswith("assistance"):
text += ', <a href="assistance.html#%s" target="_blank">assist</a>' % wfn
text += " : %s " % (wf.status)
if view and wfs != "acquired":
text += (
'<a href="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" target="_blank"><img src="https://cms-pdmv.web.cern.ch/cms-pdmv/stats/growth/%s.gif" style="height:50px"></a>'
% (wfn.replace("_", "/"), wfn.replace("_", "/"))
)
if ongoing:
text += (
'<a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a>'
% (wfn, wfn)
)
if ongoing:
date1 = time.strftime("%Y-%m-%d+%H:%M", time.gmtime(time.mktime(time.gmtime()) - (15 * 24 * 60 * 60)))
date2 = time.strftime("%Y-%m-%d+%H:%M", time.gmtime())
text += (
'<a href="http://dashb-cms-job.cern.ch/dashboard/templates/web-job2/#table=Jobs&date1=%s&date2=%s&sortby=site&task=wmagent_%s">dashb</a>'
% (date1, date2, wfn)
)
text += "<hr>"
return text
def phl(phid):
text = ", ".join(
[
str(phid),
'<a href="https://cmsweb.cern.ch/phedex/prod/Request::View?request=%s" target="_blank">vw</a>' % phid,
'<a href="https://cmsweb.cern.ch/phedex/prod/Data::Subscriptions?reqfilter=%s" target="_blank">sub</a>'
% phid,
]
)
return text
def ol(out):
return '<a href="https://cmsweb.cern.ch/das/request?input=%s" target="_blank"> %s</a>' % (out, out)
def lap(comment):
l = time.mktime(time.gmtime())
spend = l - lap.start
lap.start = l
print "Spend %d [s] for %s" % (spend, comment)
lap.start = time.mktime(time.gmtime())
## start to write it
# html_doc = open('/afs/cern.ch/user/v/vlimant/public/ops/index.html','w')
html_doc = open("/afs/cern.ch/user/c/cmst2/www/unified/index.html", "w")
print "Updating the status page ..."
if not caller:
try:
# caller = sys._getframe(1).f_code.co_name
caller = sys.argv[0].split("/")[-1].replace(".py", "")
print "caller is"
print caller
except Exception as es:
caller = "none found"
print "not getting frame"
print str(es)
html_doc.write(
"""
<html>
<head>
<META HTTP-EQUIV="refresh" CONTENT="900">
<script type="text/javascript">
function showhide(id) {
var e = document.getElementById(id);
e.style.display = (e.style.display == 'block') ? 'none' : 'block';
}
</script>
</head>
<body>
Last update on %s(CET), %s(GMT), <a href=logs/ target=_blank>logs</a> <a href=logs/last.log target=_blank>last</a> <a href=statuses.html>statuses</a> <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/ target=_blank>prod mon</a> <a href=https://cmsweb.cern.ch/wmstats/index.html target=_blank>wmstats</a> <a href=http://t3serv001.mit.edu/~cmsprod/IntelROCCS/Detox/SitesInfo.txt target=_blank>detox</a> <a href=locked.html>space</a> <a href=logs/subscribor/last.log target=_blank>blocks</a> <a href=https://twiki.cern.ch/twiki/bin/view/CMSPublic/CompOpsWorkflowL3Responsibilities#Automatic_Assignment_and_Unified>what am I</a> <a href=logs/addHoc/last.log>add-hoc op</a> created from <b>%s <a href=logs/last_running>last running</a></b><br><br>
"""
% (time.asctime(time.localtime()), time.asctime(time.gmtime()), caller)
)
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status == "considered").all():
wl = getWL(wf.name)
count_by_campaign[wl["Campaign"]][int(wl["RequestPriority"])] += 1
text += "<li> %s </li> \n" % wfl(wf, p=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write(
"""
Worflow next to handle (%d) <a href=https://cms-pdmv.cern.ch/mcm/batches?status=new&page=-1 target="_blank"> batches</a> <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/transferor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('considered')">[Click to show/hide]</a>
<br>
<div id="considered" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('considered_bywf')">[Click to show/hide]</a><div id="considered_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('considered_bycamp')">[Click to show/hide]</a><div id="considered_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""
% (count, count, text, len(count_by_campaign), text_by_c)
)
lap("done with considered")
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status == "staging").all():
wl = getWL(wf.name)
count_by_campaign[wl["Campaign"]][int(wl["RequestPriority"])] += 1
text += "<li> %s </li> \n" % wfl(wf, within=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write(
"""
Worflow waiting in staging (%d) <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staging')">[Click to show/hide]</a>
<br>
<div id="staging" style="display:none;">
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staging_bywf')">[Click to show/hide]</a><div id="staging_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staging_bycamp')">[Click to show/hide]</a><div id="staging_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""
% (count, count, text, len(count_by_campaign), text_by_c)
)
lap("done with staging")
text = ""
count = 0
for ts in session.query(Transfer).all():
stext = (
'<li> %s serves </li><a href="javascript:showhide(\'%s\')">[show/hide] relevant workflows</a> <div id="%s" style="display:none;"><ul>'
% (phl(ts.phedexid), ts.phedexid, ts.phedexid)
)
hide = True
for pid in ts.workflows_id:
w = session.query(Workflow).get(pid)
hide &= w.status != "staging"
if w.status in ["considered", "staging", "staged"]:
stext += "<li> %s </li>\n" % (wfl(w, status=True))
stext += "</ul></div>\n"
if hide:
# text+="<li> %s not needed anymore to start running (does not mean it went through completely)</li>"%phl(ts.phedexid)
pass
else:
count += 1
text += stext
text += "</ul></div>"
html_doc.write(
"""
Transfer on-going (%d) <a href=https://transferteam.web.cern.ch/transferteam/dashboard/ target=_blank>dashboard</a> <a href=logs/transferor/last.log target=_blank>log</a> <a href=logs/stagor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('transfer')">[Click to show/hide]</a>
<br>
<div id="transfer" style="display:none;">
<br>
<ul>"""
% count
)
html_doc.write(text)
lap("done with transfers")
text = ""
count = 0
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status == "staged").all():
wl = getWL(wf.name)
count_by_campaign[wl["Campaign"]][int(wl["RequestPriority"])] += 1
text += "<li> %s </li> \n" % wfl(wf, p=True)
count += 1
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
html_doc.write(
"""Worflow ready for assigning (%d) <a href=logs/stagor/last.log target=_blank>log</a> <a href=logs/assignor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('staged')">[Click to show/hide]</a>
<br>
<div id="staged" style="display:none;">
<br>
<ul>
<li> By workflow (%d) </li><a href="javascript:showhide('staged_bywf')">[Click to show/hide]</a><div id="staged_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('staged_bycamp')">[Click to show/hide]</a><div id="staged_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""
% (count, count, text, len(count_by_campaign), text_by_c)
)
lap("done with staged")
lines = []
count_by_campaign = defaultdict(lambda: defaultdict(int))
for wf in session.query(Workflow).filter(Workflow.status == "away").all():
wl = getWL(wf.name)
count_by_campaign[wl["Campaign"]][int(wl["RequestPriority"])] += 1
lines.append("<li> %s </li>" % wfl(wf, view=True, ongoing=True))
text_by_c = ""
for c in count_by_campaign:
text_by_c += "<li> %s (%d) : " % (c, sum(count_by_campaign[c].values()))
for p in sorted(count_by_campaign[c].keys()):
text_by_c += "%d (%d), " % (p, count_by_campaign[c][p])
text_by_c += "</li>"
lines.sort()
html_doc.write(
"""
Worflow on-going (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests_in_production.php target=_blank>ongoing</a> <a href=https://cms-logbook.cern.ch/elog/Workflow+processing/?mode=summary target=_blank>elog</a> <a href=http://cms-gwmsmon.cern.ch/prodview target=_blank>queues</a> <a href=logs/assignor/last.log target=_blank>log</a> <a href=logs/checkor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('away')">[Click to show/hide]</a>
<br>
<div id="away" style="display:none;">
<ul>
<li>By workflow (%d) </li>
<a href="javascript:showhide('away_bywf')">[Click to show/hide]</a><div id="away_bywf" style="display:none;">
<ul>
%s
</ul></div>
<li> By campaigns (%d) </li><a href="javascript:showhide('away_bycamp')">[Click to show/hide]</a><div id="away_bycamp" style="display:none;">
<ul>
%s
</ul></div>
</ul>
</div>
"""
% (len(lines), len(lines), "\n".join(lines), len(count_by_campaign), text_by_c)
)
lap("done with away")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == "assistance").all():
text += "<li> %s </li> \n" % wfl(wf, view=True, update=True, status=True)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""Worflow that are closing (%d)
<a href=closeout.html target=_blank>closeout</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('closing')">[Click to show/hide]</a>
<br>
<div id="closing" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with closing")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status.startswith("assistance-")).all():
text += "<li> %s </li> \n" % wfl(wf, view=True, within=True, status=True, update=True)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""Worflow which need assistance (%d)
<a href=assistance.html target=_blank>assistance</a>
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/recoveror/last.log target=_blank>postlog</a>
<a href="javascript:showhide('assistance')">[Click to show/hide]</a>
<br>
<div id="assistance" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with assistance")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == "close").all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""Worflow ready to close (%d)
<a href=logs/checkor/last.log target=_blank>log</a> <a href=logs/closor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('close')">[Click to show/hide]</a>
<br>
<div id="close" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with annoucing")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == "trouble").all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""Worflow with issue (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/injector/last.log target=_blank>postlog</a>
<a href="javascript:showhide('trouble')">[Click to show/hide]</a>
<br>
<div id="trouble" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with trouble")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == "forget").all():
text += "<li> %s </li> \n" % wfl(wf)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""
Worflow to forget (%d) <a href=logs/injector/last.log target=_blank>log</a> <a href=logs/outcleanor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('forget')">[Click to show/hide]</a>
<br>
<div id="forget" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with forget")
text = ""
count = 0
for wf in session.query(Workflow).filter(Workflow.status == "done").all():
text += "<li> %s </li> \n" % wfl(wf) # ,ms=True)
count += 1
text += "</ul></div>\n"
html_doc.write(
"""
Worflow through (%d) <a href=logs/closor/last.log target=_blank>log</a> <a href=logs/cleanor/last.log target=_blank>postlog</a>
<a href="javascript:showhide('done')">[Click to show/hide]</a>
<br>
<div id="done" style="display:none;">
<br>
<ul>
"""
% count
)
html_doc.write(text)
lap("done with done")
wfs = session.query(Workflow).filter(Workflow.status.endswith("-unlock")).all()
html_doc.write(" Workflows unlocked : %s <br>" % (len(wfs)))
lap("done with unlocked")
text = ""
lines_thisweek = []
lines_lastweek = []
now = time.mktime(time.gmtime())
this_week = int(time.strftime("%W", time.gmtime()))
start_time_two_weeks_ago = time.mktime(time.strptime("15-0-%d" % (this_week - 2), "%y-%w-%W"))
for out in session.query(Output).filter(Output.date >= start_time_two_weeks_ago).all():
if not out.workflow:
print "This is a problem with", out.datasetname
continue
if out.workflow.status in ["done", "clean", "clean-out", "clean-unlock"]:
out_week = int(time.strftime("%W", time.gmtime(out.date)))
##only show current week, and the previous.
if (this_week - out_week) == 1:
lines_lastweek.append(
"<li>on week %s : %s </li>"
% (time.strftime("%W (%x %X)", time.gmtime(out.date)), ol(out.datasetname))
)
if (this_week - out_week) == 0:
lines_thisweek.append(
"<li>on week %s : %s </li>"
% (time.strftime("%W (%x %X)", time.gmtime(out.date)), ol(out.datasetname))
)
lines_thisweek.sort()
lines_lastweek.sort()
html_doc.write(
"""Output produced (%d) <a href=https://dmytro.web.cern.ch/dmytro/cmsprodmon/requests.php?in_disagreement=1 target=_blank>disagreements</a>
<a href="javascript:showhide('output')">[Click to show/hide]</a>
<br>
<div id="output" style="display:none;">
<br>
<ul>
<li> Last week (%d) </li><a href="javascript:showhide('output_lastweek')">[Click to show/hide]</a><div id="output_lastweek" style="display:none;"><ul>
%s
</ul></div>
<li> This week (%d) </li><a href="javascript:showhide('output_thisweek')">[Click to show/hide]</a><div id="output_thisweek" style="display:none;"><ul>
%s
</ul></div></div>
"""
% (
len(lines_lastweek) + len(lines_thisweek),
len(lines_lastweek),
"\n".join(lines_lastweek),
len(lines_thisweek),
"\n".join(lines_thisweek),
)
)
lap("done with output")
html_doc.write(
"""Job installed
<a href="javascript:showhide('acron')">[Click to show/hide]</a>
<br>
<div id="acron" style="display:none;">
<br>
<pre>
%s
</pre>
"""
% (os.popen("acrontab -l | grep Unified | grep -v \#").read())
)
per_module = defaultdict(list)
for t in filter(None, os.popen("cat /afs/cern.ch/user/c/cmst2/www/unified/logs/*/*.time").read().split("\n")):
module_name, run_time, spend = t.split(":")
## then do what you want with it !
per_module[module_name].append(int(spend))
html_doc.write("Module running time<ul>\n")
for m, spends in per_module.items():
html_doc.write("<li>%s : last %d [s], avg %d [s]</li>\n" % (m, spends[-1], sum(spends) / float(len(spends))))
html_doc.write("</ul>")
html_doc.write(
"Last running <pre>%s</pre>"
% (os.popen("tac /afs/cern.ch/user/c/cmst2/www/unified/logs/running | head -5").read())
)
html_doc.write("</div>\n")
lap("done with jobs")
text = ""
count = 0
for (c, info) in campaignInfo().campaigns.items():
# if 'go' in info and info['go']:
text += "<li>%s <br> <pre>%s</pre> </li>" % (c, json.dumps(info, indent=2))
count += 1
html_doc.write(
"""Campaign configuration
<a href="javascript:showhide('campaign')">[Click to show/hide]</a>
<br>
<div id="campaign" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""
% (text)
)
text = ""
count = 0
n_column = 4
SI = siteInfo()
for t in SI.types():
text += "<li>%s<table border=1>" % t
c = 0
for site in getattr(SI, t):
cpu = SI.cpu_pledges[site] if site in SI.cpu_pledges else "N/A"
disk = SI.disk[SI.CE_to_SE(site)] if SI.CE_to_SE(site) in SI.disk else "N/A"
if c == 0:
text += "<tr>"
text += (
'<td><a href=http://dashb-ssb.cern.ch/dashboard/templates/sitePendingRunningJobs.html?site=%s>%s</a><br><a href="http://cms-gwmsmon.cern.ch/prodview/%s" target="_blank"><img src="http://cms-gwmsmon.cern.ch/prodview/graphs/%s/daily" style="height:50px"></a><br>CPU pledge: %s<br>Disk available: %s</td>'
% (site, site, site, site, cpu, disk)
)
if c == n_column:
c = 0
else:
c += 1
text += "</table></li>"
lap("done with campaigns")
open("/afs/cern.ch/user/c/cmst2/www/unified/siteInfo.json", "w").write(
json.dumps(dict([(t, getattr(SI, t)) for t in SI.types()]), indent=2)
)
lap("done with sites json")
chart_data = defaultdict(list)
for site in SI.quota:
chart_data[site].append(
"""
var data_%s = google.visualization.arrayToDataTable([
['Overall', 'Space in TB'],
//['Quota' , %s],
['Locked' , %s],
['Free' , %s]
]);
"""
% (site, SI.quota[site], SI.locked[site], SI.disk[site])
)
chart_data[site].append(
"""
var chart_%s = new google.visualization.PieChart(document.getElementById('donutchart_%s'));
chart_%s.draw(data_%s, {title: '%s %s [TB]', pieHole:0.4, slices:{0:{color:'red'},1:{color:'green'}}});
"""
% (site, site, site, site, site, SI.quota[site])
)
chart_data[site].append(
"""
<div id="donutchart_%s" style="height: 200px;"></div>
"""
% (site)
)
## make the locked/available donut chart
donut_html = open("/afs/cern.ch/user/c/cmst2/www/unified/locked.html", "w")
tables = "\n".join([info[0] for site, info in chart_data.items()])
draws = "\n".join([info[1] for site, info in chart_data.items()])
divs = "\n".join([info[2] for site, info in chart_data.items()])
divs_table = "<table border=0>"
for c, site in enumerate(sorted(chart_data.keys())):
if c % 6 == 0:
divs_table += "<tr>"
divs_table += "<td>%s</td>" % (chart_data[site][2])
divs_table += "</table>"
donut_html.write(
"""
<html>
<head>
<script type="text/javascript" src="https://www.google.com/jsapi"></script>
<script type="text/javascript">
google.load("visualization", "1", {packages:["corechart"]});
google.setOnLoadCallback(drawChart);
function drawChart() {
%s
%s
}
</script>
</head>
<body>
%s
</body>
</html>
"""
% (tables, draws, divs_table)
)
donut_html.close()
html_doc.write(
"""Site configuration
<a href="javascript:showhide('site')">[Click to show/hide]</a>
<br>
<div id="site" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""
% (text)
)
lap("done with space")
UC = unifiedConfiguration()
text = ""
for param in UC.configs:
text += "<li>%s</li><ul>\n" % param
for sub in sorted(UC.configs[param].keys()):
text += "<li> %s : %s </li>\n" % (sub, UC.configs[param][sub])
text += "</ul>\n"
html_doc.write(
"""Unified configuration
<a href="javascript:showhide('config')">[Click to show/hide]</a>
<br>
<div id="config" style="display:none;">
<br>
<ul>
%s
</ul></div>
"""
% (text)
)
lap("done with configuration")
print "... done with status page."
html_doc.write(
"""
</body>
</html>
"""
)
html_doc.close()
html_doc = open("/afs/cern.ch/user/c/cmst2/www/unified/statuses.html", "w")
html_doc.write(
""" <html>
<table border=1>
<thead>
<tr>
<th> workflow </th><th> status </th><th> wm status</th>
</tr>
</thead>
"""
)
wfs = {}
for wfo in session.query(Workflow).all():
wfs[wfo.name] = (wfo.status, wfo.wm_status)
open("/afs/cern.ch/user/c/cmst2/www/unified/statuses.json", "w").write(json.dumps(wfs))
for wfn in sorted(wfs.keys()):
html_doc.write(
'<tr><td><a id="%s">%s</a></td><td>%s</td><td>%s</td></tr>\n' % (wfn, wfn, wfs[wfn][0], wfs[wfn][1])
)
html_doc.write("</table>")
html_doc.write("<br>" * 100)
html_doc.write("end of page</html>")
html_doc.close()
def completor(url, specific):
mlock = moduleLock(silent=True)
if mlock(): return
use_mcm = True
up = componentInfo(soft=['mcm', 'wtc', 'jira'])
if not up.check(): return
use_mcm = up.status['mcm']
if use_mcm:
mcm = McMClient(dev=False)
safe_mode = False
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
JC = JIRAClient() if up.status.get('jira', False) else None
wfs = []
wfs.extend(session.query(Workflow).filter(Workflow.status == 'away').all())
wfs.extend(
session.query(Workflow).filter(
Workflow.status.startswith('assistance')).all())
## just take it in random order so that not always the same is seen
random.shuffle(wfs)
max_per_round = UC.get('max_per_round').get('completor', None)
if max_per_round and not specific: wfs = wfs[:max_per_round]
all_stuck = set()
## take into account what stagor was saying
for itry in range(5):
try:
all_stuck.update(
json.loads(eosRead('%s/stuck_transfers.json' %
monitor_pub_dir)))
break
except:
time.sleep(2)
for itry in range(5):
try:
## take into account the block that needed to be repositioned recently
all_stuck.update([
b.split('#')[0] for b in json.loads(
eosRead('%s/missing_blocks.json' % monitor_dir))
])
break
except:
time.sleep(2)
## take into account all stuck block and dataset from transfer team
all_stuck.update(getAllStuckDataset())
good_fractions = {}
overdoing_fractions = {}
truncate_fractions = {}
timeout = {}
campaign_injection_delay = {}
for c in CI.campaigns:
if 'force-complete' in CI.campaigns[c]:
good_fractions[c] = CI.campaigns[c]['force-complete']
if 'truncate-complete' in CI.campaigns[c]:
truncate_fractions[c] = CI.campaigns[c]['truncate-complete']
if 'force-timeout' in CI.campaigns[c]:
timeout[c] = CI.campaigns[c]['force-timeout']
if 'injection-delay' in CI.campaigns[c]:
campaign_injection_delay[c] = CI.campaigns[c]['injection-delay']
if 'overdoing-complete' in CI.campaigns[c]:
overdoing_fractions[c] = CI.campaigns[c]['overdoing-complete']
long_lasting = {}
WI = wtcInfo()
overrides = WI.getForce()
if use_mcm:
## add all workflow that mcm wants to get force completed
mcm_force = mcm.get('/restapi/requests/forcecomplete')
## assuming this will be a list of actual prepids
overrides['mcm'] = mcm_force
print "can force complete on"
print json.dumps(good_fractions, indent=2)
print "can truncate complete on"
print json.dumps(truncate_fractions, indent=2)
print "can overide on"
print json.dumps(overrides, indent=2)
max_force = UC.get("max_force_complete")
max_priority = UC.get("max_tail_priority")
injection_delay_threshold = UC.get("injection_delay_threshold")
injection_delay_priority = UC.get("injection_delay_priority")
delay_priority_increase = UC.get("delay_priority_increase")
default_fraction_overdoing = UC.get('default_fraction_overdoing')
set_force_complete = set()
# priority and time above which to fire a JIRA
jira_priority_and_delays = {
110000: 21,
90000: 28,
# 80000 : 60,
#0 : 90
}
for wfo in wfs:
if specific and not specific in wfo.name: continue
print "looking at", wfo.name
## get all of the same
wfi = workflowInfo(url, wfo.name)
pids = wfi.getPrepIDs()
skip = False
campaigns = wfi.getCampaigns()
#if not any([c in good_fractions.keys() for c in campaigns]): skip=True
#if not any([c in truncate_fractions.keys() for c in campaigns]): skip=True
for user, spec in overrides.items():
if not spec: continue
spec = filter(None, spec)
if not wfi.request['RequestStatus'] in [
'force-complete', 'completed'
]:
if any(s in wfo.name
for s in spec) or (wfo.name in spec) or any(
pid in spec for pid in pids) or any(s in pids
for s in spec):
wfi = workflowInfo(url, wfo.name)
forceComplete(url, wfi)
skip = True
wfi.notifyRequestor(
"The workflow %s was force completed by request of %s"
% (wfo.name, user),
do_batch=False)
wfi.sendLog(
'completor',
'%s is asking for %s to be force complete' %
(user, wfo.name))
break
if wfo.status.startswith('assistance'): skip = True
if skip:
continue
priority = wfi.request['RequestPriority']
if not 'Campaign' in wfi.request: continue
if not wfi.request['RequestStatus'] in [
'acquired', 'running-open', 'running-closed'
]:
continue
## until we can map the output to task ...
output_per_task = wfi.getOutputPerTask(
) ## can use that one, and follow mapping
good_fraction_per_out = {}
good_fraction_nodelay_per_out = {}
truncate_fraction_per_out = {}
#allowed_delay_per_out = {}
for task, outs in output_per_task.items():
task_campaign = wfi.getCampaignPerTask(task)
for out in outs:
good_fraction_per_out[out] = good_fractions.get(
task_campaign, 1000.)
good_fraction_nodelay_per_out[out] = overdoing_fractions.get(
task_campaign, default_fraction_overdoing)
truncate_fraction_per_out[out] = truncate_fractions.get(
task_campaign, 1000.)
#allowed_delay_per_out[out] = timeout.get(task_campaign, 14)
#print "force at", json.dumps( good_fraction_per_out, indent=2)
#print "truncate at",json.dumps( truncate_fraction_per_out, indent=2)
now = time.mktime(time.gmtime()) / (60 * 60 * 24.)
priority_log = filter(lambda change: change['Priority'] == priority,
wfi.request.get('PriorityTransition', []))
if not priority_log:
print "\tHas no priority log"
priority_delay = 0
else:
then = max([change['UpdateTime']
for change in priority_log]) / (60. * 60. * 24.)
priority_delay = now - then ## in days
print "priority was set to", priority, priority_delay, "[days] ago"
running_log = filter(
lambda change: change["Status"
] in ["running-open", "running-closed"],
wfi.request['RequestTransition'])
if not running_log:
print "\tHas no running log"
delay = 0
else:
then = max([change['UpdateTime']
for change in running_log]) / (60. * 60. * 24.)
delay = now - then ## in days
#further check on delays
cpuh = wfi.getComputingTime(unit='d')
wfi.sendLog(
'completor',
"Requires %.1f [CPUd], running since %.1f [day], last priority change to %d done %.1f [day] ago"
% (cpuh, delay, priority, priority_delay))
if priority_delay != 0 and priority_delay < delay:
## regardless when it started running, set the delay to when priority was changed last
delay = priority_delay
## this is supposed to be the very initial request date, inherited from clones
injection_delay = None
original = wfi
if 'OriginalRequestName' in original.request:
## go up the clone chain
original = workflowInfo(url,
original.request['OriginalRequestName'])
injected_log = filter(
lambda change: change["Status"] in ["assignment-approved"],
original.request['RequestTransition'])
if injected_log:
injected_on = injected_log[-1]['UpdateTime'] / (60. * 60. * 24.)
injection_delay = now - injected_on
delay_for_priority_increase = injection_delay
#delay_for_priority_increase = delay
(w, d) = divmod(delay, 7)
print "\t" * int(
w) + "Running since", delay, "[days] priority=", priority
pop_a_jira = False
ping_on_jira = 7 * (24 * 60 * 60) # 7 days
for jp, jd in jira_priority_and_delays.items():
if priority >= jp and delay >= jd: pop_a_jira = True
if pop_a_jira and JC:
j, reopened, just_created = JC.create_or_last(
prepid=wfi.request['PrepID'],
priority=wfi.request['RequestPriority'],
label='Late',
reopen=True)
last_time = JC.last_time(j)
since_last_ping = time.mktime(time.gmtime()) - last_time
if since_last_ping > ping_on_jira or just_created:
j_comment = "Running since %.1f [days] at priority %d" % (
delay, priority)
JC.comment(j.key, j_comment)
if delay_for_priority_increase != None and delay_for_priority_increase > injection_delay_threshold and priority >= injection_delay_priority:
quantized = 5000 ## quantize priority
tail_cutting_priority = wfi.request['InitialPriority'] + int(
(delay_priority_increase *
(delay_for_priority_increase - injection_delay_threshold) / 7)
/ quantized) * quantized
tail_cutting_priority += 101 ## to signal it is from this mechanism
tail_cutting_priority = min(
400000, tail_cutting_priority) ## never go above 400k priority
tail_cutting_priority = max(
tail_cutting_priority,
priority) ## never go below the current value
if priority < tail_cutting_priority:
if max_priority:
sendLog(
'completor',
"%s Injected since %s [days] priority=%s, increasing to %s"
% (wfo.name, delay_for_priority_increase, priority,
tail_cutting_priority),
level='critical')
wfi.sendLog(
'completor',
'bumping priority to %d for being injected since %s' %
(tail_cutting_priority, delay_for_priority_increase))
reqMgrClient.changePriorityWorkflow(
url, wfo.name, tail_cutting_priority)
max_priority -= 1
else:
sendLog(
'completor',
"%s Injected since %s [days] priority=%s, would like to increase to %s"
% (wfo.name, delay_for_priority_increase, priority,
tail_cutting_priority),
level='critical')
wfi.sendLog(
'completor',
'would like to bump priority to %d for being injected since %s'
% (tail_cutting_priority, delay_for_priority_increase))
print "Could be changing the priority to higher value, but too many already were done"
_, prim, _, _ = wfi.getIO()
is_stuck = all_stuck & prim
if is_stuck:
wfi.sendLog('completor', '%s is stuck' % ','.join(is_stuck))
monitor_delay = 7
allowed_delay = max([timeout.get(c, 14) for c in campaigns])
monitor_delay = min(monitor_delay, allowed_delay)
### just skip if too early, just for the sake of not computing the completion fraction just now.
# maybe this is fast enough that we can do it for all
if delay <= monitor_delay:
print "not enough time has passed yet"
continue
long_lasting[wfo.name] = {
"delay": delay,
"injection_delay": injection_delay
}
percent_completions = wfi.getCompletionFraction(caller='completor')
if not percent_completions:
sendLog('completor',
'%s has no output at all' % wfo.name,
level='critical')
continue
is_over_allowed_delay = (all([
percent_completions[out] >= good_fraction_per_out.get(out, 1000.)
for out in percent_completions
]) and delay >= allowed_delay)
is_over_truncation_delay = (is_stuck and (all([
percent_completions[out] >= truncate_fraction_per_out.get(
out, 1000.) for out in percent_completions
])) and delay >= allowed_delay)
is_over_completion = (all([
percent_completions[out] >= good_fraction_nodelay_per_out.get(
out, 1000.) for out in percent_completions
]))
if is_over_completion:
wfi.sendLog(
'completor', "all is over completed %s\n %s" %
(json.dumps(good_fraction_nodelay_per_out, indent=2),
json.dumps(percent_completions, indent=2)))
elif is_over_allowed_delay:
wfi.sendLog(
'completor', "all is above %s \n%s" %
(json.dumps(good_fraction_per_out, indent=2),
json.dumps(percent_completions, indent=2)))
elif is_over_truncation_delay:
wfi.sendLog(
'completor',
"all is above %s truncation level, and the input is stuck\n%s"
% (json.dumps(truncate_fraction_per_out, indent=2),
json.dumps(percent_completions, indent=2)))
else:
long_lasting[wfo.name].update({
'completion':
sum(percent_completions.values()) / len(percent_completions),
'completions':
percent_completions
})
## do something about the agents this workflow is in
long_lasting[wfo.name]['agents'] = wfi.getAgents()
wfi.sendLog(
'completor',
"%s not over bound \ncomplete at %s \n truncate at %s \nRunning %s"
% (json.dumps(percent_completions, indent=2),
json.dumps(good_fraction_per_out, indent=2),
json.dumps(truncate_fraction_per_out, indent=2),
json.dumps(long_lasting[wfo.name]['agents'], indent=2)))
continue
#for output in percent_completions:
# completions[output]['injected'] = then
ran_at = wfi.request['SiteWhitelist']
wfi.sendLog('completor', "Required %s, time spend %s" % (cpuh, delay))
##### WILL FORCE COMPLETE BELOW
# only really force complete after n days
## find ACDCs that might be running
if max_force > 0:
print "going for force-complete of", wfo.name
if not safe_mode:
forceComplete(url, wfi)
set_force_complete.add(wfo.name)
wfi.sendLog('completor', 'going for force completing')
wfi.notifyRequestor(
"The workflow %s was force completed for running too long"
% wfo.name)
max_force -= 1
else:
sendEmail(
'completor',
'The workflow %s is ready for force complete, but completor is in safe mode'
% wfo.name)
else:
wfi.sendLog(
'completor',
"too many completion this round, cannot force complete")
if set_force_complete:
sendLog(
'completor', 'The followings were set force-complete \n%s' %
('\n'.join(set_force_complete)))
#open('%s/completions.json'%monitor_dir,'w').write( json.dumps( completions , indent=2))
text = "These have been running for long"
#open('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 ))
eosFile('%s/longlasting.json' % monitor_dir,
'w').write(json.dumps(long_lasting, indent=2)).close()
for wf, info in sorted(long_lasting.items(),
key=lambda tp: tp[1]['delay'],
reverse=True):
delay = info['delay']
text += "\n %s : %s days" % (wf, delay)
if 'completion' in info:
text += " %d%%" % (info['completion'] * 100)
print text
def stuckor(url = reqmgr_url):
if duplicateLock(): return
datasets_by_phid = json.loads(open('datasets_by_phid.json').read())
really_stuck_dataset = set(json.loads(open('really_stuck_dataset.json').read()))
UC = unifiedConfiguration()
print "make a report of stuck transfers"
bad_destinations = defaultdict(set)
bad_sources = defaultdict(set)
report = ""
transfer_timeout = UC.get("transfer_timeout")
transfer_lowrate = UC.get("transfer_lowrate")
for phid,datasets in datasets_by_phid.items():
issues = checkTransferLag( url, phid , datasets=list(datasets) )
for dataset in issues:
for block in issues[dataset]:
for destination in issues[dataset][block]:
(block_size,destination_size,delay,rate,dones) = issues[dataset][block][destination]
## count x_Buffer and x_MSS as one source
redones=[]
for d in dones:
if d.endswith('Buffer') or d.endswith('Export'):
if d.replace('Buffer','MSS').replace('Export','MSS') in dones:
continue
else:
redones.append( d )
else:
redones.append( d )
dones = list(set( redones ))
#dones = filter(lambda s : (s.endswith('Buffer') and not s.replace('Buffer','MSS') in dones) or (not s.endswith('Buffer')) , dones)
if delay>transfer_timeout and rate<transfer_lowrate:
if len(dones)>1:
## its the destination that sucks
bad_destinations[destination].add( block )
else:
dum=[bad_sources[d].add( block ) for d in dones]
really_stuck_dataset.add( dataset )
print "add",dataset,"to really stuck"
report += "%s is not getting to %s, out of %s faster than %f [GB/s] since %f [d]\n"%(block,destination,", ".join(dones), rate, delay)
print "\n"*2
## create tickets right away ?
report+="\nbad sources "+",".join(bad_sources.keys())+"\n"
for site,blocks in bad_sources.items():
report+="\n\n%s:"%site+"\n\t".join(['']+list(blocks))
report+="\nbad destinations "+",".join(bad_destinations.keys())+"\n"
for site,blocks in bad_destinations.items():
report+="\n\n%s:"%site+"\n\t".join(['']+list(blocks))
print '\n'*2,"Datasets really stuck"
print '\n'.join( really_stuck_dataset )
print '\n'*2,"report written at %s/logs/incomplete_transfers.log"%unified_url
print report
missing_in_action = json.loads(open('%s/incomplete_transfers.json'%monitor_dir).read())
stuck_transfers = dict([(k,v) for (k,v) in missing_in_action.items() if k in really_stuck_dataset])
print '\n'*2,'Stuck dataset transfers'
print json.dumps(stuck_transfers , indent=2)
open('%s/stuck_transfers.json'%monitor_pub_dir,'w').write( json.dumps(stuck_transfers , indent=2) )
open('%s/logs/incomplete_transfers.log'%monitor_dir,'w').write( report )
def new_recoveror(url, specific, options=None):
if userLock('recoveror'): return
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): return
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
wfs = session.query(Workflow).filter(
Workflow.status.contains('recovery')).all()
if specific:
wfs.extend(
session.query(Workflow).filter(
Workflow.status == 'assistance-manual').all())
try:
from_operator = json.loads(
os.popen(
'curl -s http://vocms0113.cern.ch/actions/test.json').read())
## now we have a list of things that we can take action on
except:
pass
for wfo in wfs:
if specific and not specific in wfo.name: continue
if not specific and 'manual' in wfo.status: continue
wfi = workflowInfo(url, wfo.name)
send_recovery = False ## will make all acdc
send_clone = False ## will make a clone
send_back = False ## should just reject. manual ?
send_manual = False ## will set in manual
where_to_run, missing_to_run = wfi.getRecoveryInfo()
task_to_recover = where_to_run.keys()
## if the site at which the recovery could run in drain or out ?
for task in task_to_recover:
not_ready = set(where_to_run[task]) - set(SI.sites_ready)
if not_ready:
print "the following sites are not ready for the ACDC", ",".join(
sorted(not_ready))
## do we have a way of telling if a site is going to be out for a long time ?
# check on priority: high prio, restart
if wfi.request['RequestPriority'] >= 85000:
send_clone = True
# check on age of the request
injection_time = time.mktime(
time.strptime(
'.'.join(map(str, wfi.request['RequestDate'])),
"%Y.%m.%d.%H.%M.%S")) / (60. * 60.)
now = time.mktime(time.gmtime()) / (60. * 60.)
if float(now - injection_time) < 14.:
## less than 14 days, start over
send_clone = True
else:
send_manual = True
if not send_recovery:
## check on whether the stats is very low
pass
if send_recovery:
## make acdc for all tasks
for task in task_to_recover:
actions = list(
set([
case['solution']
for code, case in task_to_recover[task]
]))
acdc = singleRecovery(url, task, wfi.request, actions, do=True)
elif send_clone:
## this will get it cloned
wfo.status = 'assistance-clone'
session.commit()
elif send_manual:
wfo.status = 'assistance-manual'
def equalizor(url, specific=None, options=None):
up = componentInfo(mcm=False, soft=['mcm'])
if not up.check(): return
if not specific:
workflows = getWorkflows(url, status='running-closed', details=True)
workflows.extend(getWorkflows(url, status='running-open',
details=True))
## start from scratch
modifications = defaultdict(dict)
## define regionality site => fallback allowed. feed on an ssb metric ??
mapping = defaultdict(list)
reversed_mapping = defaultdict(list)
regions = defaultdict(list)
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
for site in SI.sites_ready:
region = site.split('_')[1]
if not region in ['US', 'DE', 'IT']: continue
regions[region] = [region]
def site_in_depletion(s):
return True
if s in SI.sites_pressure:
(m, r, pressure) = SI.sites_pressure[s]
if float(m) < float(r):
print s, m, r, "lacking pressure"
return True
else:
print s, m, r, "pressure"
pass
return False
for site in SI.sites_ready:
region = site.split('_')[1]
## fallback to the region, to site with on-going low pressure
mapping[site] = [
fb for fb in SI.sites_ready
if any([('_%s_' %
(reg) in fb and fb != site and site_in_depletion(fb))
for reg in regions[region]])
]
use_T0 = ('T0_CH_CERN' in UC.get("site_for_overflow"))
if options.t0: use_T0 = True
#if options.augment : use_T0 = True
use_HLT = ('T2_CH_CERN_HLT' in UC.get("site_for_overflow"))
if options.hlt: use_HLT = True
#if options.augment : use_HLT=True
if use_HLT:
mapping['T2_CH_CERN'].append('T2_CH_CERN_HLT')
if use_T0:
mapping['T2_CH_CERN'].append('T0_CH_CERN')
#mapping['T1_FR_CCIN2P3'].append('T0_CH_CERN')
#mapping['T2_IT_Legnaro'].append('T1_IT_CNAF')
for reg in ['IT', 'DE', 'UK']:
mapping['T2_CH_CERN'].extend(
[fb for fb in SI.sites_ready if '_%s_' % reg in fb])
## make them appear as OK to use
force_sites = []
## overflow CERN to underutilized T1s
upcoming = json.loads(open('%s/GQ.json' % monitor_dir).read())
for possible in SI.sites_T1s:
if not possible in upcoming:
mapping['T2_CH_CERN'].append(possible)
## remove add-hoc sites from overflow mapping
prevent_sites = ['T2_US_Purdue']
for prevent in prevent_sites:
if prevent in mapping: mapping.pop(prevent)
for src in mapping:
for prevent in prevent_sites:
if prevent in mapping[src]:
mapping[src].remove(prevent)
## create the reverse mapping for the condor module
for site, fallbacks in mapping.items():
for fb in fallbacks:
reversed_mapping[fb].append(site)
## this is the fallback mapping
print "Direct mapping : site => overflow"
print json.dumps(mapping, indent=2)
print "Reverse mapping : dest <= from origin"
print json.dumps(reversed_mapping, indent=2)
altered_tasks = set()
def running_idle(wfi, task_name):
gmon = wfi.getGlideMon()
#print gmon
if not gmon: return (0, 0)
if not task_name in gmon: return (0, 0)
return (gmon[task_name]['Running'], gmon[task_name]['Idle'])
def needs_action(wfi, task, min_idled=100, pressure=0.2):
task_name = task.pathName.split('/')[-1]
running, idled = running_idle(wfi, task_name)
go = True
if not idled and not running:
go = False
if idled < 100:
go = False
if (not running and idled) or (running and
(idled / float(running) > pressure)):
go = True
else:
go = False
return go, task_name, running, idled
def getPerf(task):
task = task.split('/')[1] + '/' + task.split('/')[-1]
try:
u = 'http://cms-gwmsmon.cern.ch/prodview/json/history/memoryusage720/%s' % task
print u
perf_data = json.loads(os.popen('curl -s --retry 5 %s' % u).read())
except Exception as e:
print str(e)
return (None, None)
buckets = perf_data['aggregations']["2"]['buckets']
s_m = sum(bucket['key'] * bucket['doc_count'] for bucket in buckets)
w_m = sum(bucket['doc_count'] for bucket in buckets)
m_m = max(bucket['key'] for bucket in buckets) if buckets else None
b_m = None
if w_m > 100:
b_m = m_m
try:
perf_data = json.loads(
os.popen(
'curl -s --retry 5 http://cms-gwmsmon.cern.ch/prodview/json/history/runtime720/%s'
% task).read())
except Exception as e:
print str(e)
return (b_m, None)
buckets = perf_data['aggregations']["2"]['buckets']
s_t = sum(bucket['key'] * bucket['doc_count'] for bucket in buckets)
w_t = sum(bucket['doc_count'] for bucket in buckets)
m_t = max(bucket['key'] for bucket in buckets) if buckets else None
b_t = None
if w_t > 100:
b_t = m_t
return (b_m, b_t)
def getcampaign(task):
taskname = task.pathName.split('/')[-1]
if hasattr(task, 'prepID'):
return task.prepID.split('-')[1]
elif taskname.count('-') >= 1:
return taskname.split('-')[1]
else:
return None
def close(interface):
open('%s/equalizor.json.new' % monitor_dir,
'w').write(json.dumps(interface, indent=2))
os.system('mv %s/equalizor.json.new %s/equalizor.json' %
(monitor_dir, monitor_dir))
os.system('cp %s/equalizor.json %s/logs/equalizor/equalizor.%s.json' %
(monitor_dir, monitor_dir, time.mktime(time.gmtime())))
interface = {'reversed_mapping': reversed_mapping, 'modifications': {}}
if options.augment or options.remove:
interface['modifications'] = json.loads(
open('%s/equalizor.json' % monitor_dir).read())['modifications']
if options.remove:
if specific in interface['modifications']:
print "poping", specific
interface['modifications'].pop(specific)
close(interface)
return
PU_locations = {}
PU_overflow = {}
LHE_overflow = {}
tune_performance = []
pending_HLT = 0
max_HLT = 60000
pending_T0 = 0
max_T0 = 60000
try:
gmon = json.loads(
os.popen(
'curl -s http://cms-gwmsmon.cern.ch/prodview/json/T2_CH_CERN_HLT'
).read())
pending_HLT += gmon["Running"]
pending_HLT += gmon["MatchingIdle"]
except:
pass
stay_within_site_whitelist = False
specific_task = None
if specific and ":" in specific:
specific, specific_task = specific.split(':')
if specific:
wfs = session.query(Workflow).filter(
Workflow.name.contains(specific)).all()
else:
wfs = session.query(Workflow).filter(Workflow.status == 'away').all()
performance = {}
no_routing = []
random.shuffle(wfs)
for wfo in wfs:
if wfo.name in no_routing and not options.augment:
continue
if specific and not specific in wfo.name:
continue
if specific:
wfi = workflowInfo(url, wfo.name)
else:
cached = filter(lambda d: d['RequestName'] == wfo.name, workflows)
if not cached: continue
wfi = workflowInfo(url, wfo.name, request=cached[0])
## only running should get re-routed
if not wfi.request['RequestStatus'] in [
'running-open', 'running-closed'
] and not specific:
continue
tasks_and_campaigns = []
for task in wfi.getWorkTasks():
tasks_and_campaigns.append((task, getcampaign(task)))
_, _, _, sec = wfi.getIO()
## check needs override
needs_overide = False
if not needs_overide and options.augment: needs_overide = True
def overide_from_agent(wfi, needs_overide):
bad_agents = [] #'http://cmssrv219.fnal.gov:5984']
if not bad_agents: return needs_overide
if needs_overide: return True
agents = wfi.getAgents()
wqss = ['Running', 'Acquired']
if any([
agent in agents.get(wqs, {}).keys()
for wqs, agent in itertools.product(wqss, bad_agents)
]):
print "overriding the need for bad agent"
needs_overide = True
return needs_overide
## now parse this for action
for i_task, (task, campaign) in enumerate(tasks_and_campaigns):
if options.augment:
print task.pathName
print campaign
tune = CI.get(campaign, 'tune', options.tune)
if tune and not campaign in tune_performance:
tune_performance.append(campaign)
overflow = CI.get(campaign, 'overflow', {})
if overflow:
if "PU" in overflow and not campaign in PU_overflow:
PU_overflow[campaign] = copy.deepcopy(overflow['PU'])
print "adding", campaign, "to PU overflow rules"
if "LHE" in overflow and not campaign in LHE_overflow:
print "adding", campaign, "to light input overflow rules"
site_list = overflow['LHE']['site_list']
LHE_overflow[campaign] = copy.deepcopy(
getattr(SI, site_list))
### get the task performance, for further massaging.
if campaign in tune_performance or options.tune:
print "performance", task.taskType, task.pathName
if task.taskType in ['Processing', 'Production']:
set_memory, set_time = getPerf(task.pathName)
#print "Performance %s GB %s min"%( set_memory,set_time)
wfi.sendLog(
'equalizor', 'Performance tuning to %s GB %s min' %
(set_memory, set_time))
## get values from gmwsmon
# massage the values : 95% percentile
performance[task.pathName] = {}
if set_memory:
performance[task.pathName]['memory'] = set_memory
if set_time and False:
performance[task.pathName]['time'] = set_time
### rule to avoid the issue of taskchain secondary jobs being stuck at sites processing the initial step
if campaign in LHE_overflow:
if task.taskType in ['Processing']:
needs, task_name, running, idled = needs_action(wfi, task)
needs_overide = overide_from_agent(wfi, needs_overide)
extend_to = list(set(copy.deepcopy(
LHE_overflow[campaign])))
if stay_within_site_whitelist:
extend_to = list(
set(extend_to) & set(wfi.request['SiteWhitelist'])
) ## restrict to stupid-site-whitelist
extend_to = list(
set(extend_to) & set(SI.sites_ready + force_sites))
if extend_to and needs or needs_overide:
modifications[wfo.name][task.pathName] = {
"ReplaceSiteWhitelist": extend_to,
"Running": running,
"Pending": idled,
"Priority": wfi.request['RequestPriority']
}
wfi.sendLog(
'equalizor',
'%s of %s is running %d and pending %d, taking action : ReplaceSiteWhitelist \n %s'
%
(task_name, wfo.name, running, idled,
json.dumps(
sorted(modifications[wfo.name][task.pathName]
['ReplaceSiteWhitelist']))))
altered_tasks.add(task.pathName)
else:
wfi.sendLog(
'equalizor',
'%s of %s is running %d and pending %d' %
(task_name, wfo.name, running, idled))
### overflow the 76 digi-reco to the site holding the pileup
if campaign in PU_overflow:
force = PU_overflow[campaign][
'force'] if 'force' in PU_overflow[campaign] else False
secondary_locations = set(SI.sites_ready + force_sites)
for s in sec:
if not s in PU_locations:
presence = getDatasetPresence(url, s)
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
one_secondary_locations = [
site for (site, (there, frac)) in presence.items()
if frac > 98.
]
PU_locations[s] = one_secondary_locations
print "secondary is at", sorted(PU_locations[s])
secondary_locations = set(
[SI.SE_to_CE(site)
for site in PU_locations[s]]) & secondary_locations
## we should add all sites that hold the secondary input if any
### given that we have the secondary location available, it is not necessary to use the add-hoc list
##secondary_locations = list(set(PU_overflow[campaign]['sites']) & set( SI.sites_ready ))
if any([
task.pathName.endswith(finish)
for finish in ['_0', 'StepOneProc', 'Production']
]):
needs, task_name, running, idled = needs_action(wfi, task)
## removing the ones in the site whitelist already since they encode the primary input location
if stay_within_site_whitelist:
original_site_in_use = set(
wfi.request['SiteWhitelist'])
else:
original_site_in_use = set(secondary_locations)
## remove the sites that have already running jobs
gmon = wfi.getGlideMon()
if gmon and task_name in gmon and 'Sites' in gmon[
task_name]:
site_in_use = set(gmon[task_name]['Sites'])
print "removing", sorted(site_in_use)
## that determines where you want to run in addition
augment_by = list((set(secondary_locations) -
site_in_use) & original_site_in_use)
else:
print "no existing running site"
augment_by = list(original_site_in_use)
needs_overide = overide_from_agent(wfi, needs_overide)
if augment_by and (
needs or needs_overide
or force) and PU_overflow[campaign][
'pending'] < PU_overflow[campaign]['max']:
PU_overflow[campaign]['pending'] += idled
print "raising overflow to", PU_overflow[campaign][
'pending'], "for", PU_overflow[campaign]['max']
## the step with an input ought to be the digi part : make this one go anywhere
modifications[wfo.name][task.pathName] = {
"AddWhitelist": augment_by,
"Running": running,
"Pending": idled,
"Priority": wfi.request['RequestPriority']
}
altered_tasks.add(task.pathName)
wfi.sendLog(
'equalizor',
'%s of %s is running %d and pending %d, taking action : AddWhitelist \n %s'
% (task_name, wfo.name, running, idled,
json.dumps(sorted(augment_by), indent=2)))
else:
print task_name, "of", wfo.name, "running", running, "and pending", idled
### overflow the skims back to multi-core
if campaign in ['Run2015D', 'Run2015C_25ns'
] and task.taskType == 'Skim':
original_swl = wfi.request['SiteWhitelist']
needs, task_name, running, idled = needs_action(wfi, task)
if (needs or needs_overide):
modifications[wfo.name][task.pathName] = {
'AddWhitelist': original_swl,
"Running": running,
"Pending": idled,
"Priority": wfi.request['RequestPriority']
}
altered_tasks.add(task.pathName)
wfi.sendLog(
'equalizor',
'%s of %s is running %d and pending %d, taking action : AddWhitelist \n %s'
% (task_name, wfo.name, running, idled,
json.dumps(sorted(original_swl), indent=2)))
if options.augment:
print sorted(wfi.request['SiteWhitelist']), i_task, use_HLT
### add the HLT at partner of CERN
if 'T2_CH_CERN' in wfi.request['SiteWhitelist'] and i_task in [
0, 1
] and use_HLT:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs = True
needs = True
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide) and pending_HLT < max_HLT:
pending_HLT += idled
if task.pathName in modifications[
wfo.name] and 'AddWhitelist' in modifications[
wfo.name][task.pathName]:
modifications[wfo.name][task.pathName][
"AddWhitelist"].append("T2_CH_CERN_HLT")
print "\t", wfo.name, "adding addHLT up to", pending_HLT, "for", max_HLT
print task.pathName
## this Replace does not work at all for HLT
#elif task.pathName in modifications[wfo.name] and 'ReplaceSiteWhitelist' in modifications[wfo.name][task.pathName]:
#modifications[wfo.name][task.pathName]["ReplaceSiteWhitelist"].append( "T2_CH_CERN_HLT" )
#print "\t",wfo.name,"adding replace HLT up to",pending_HLT,"for",max_HLT
else:
modifications[wfo.name][task.pathName] = {
"AddWhitelist": ["T2_CH_CERN_HLT"],
"Priority": wfi.request['RequestPriority'],
"Running": running,
"Pending": idled
}
wfi.sendLog(
'equalizor',
'adding the HLT in whitelist of %s to %d for %d' %
(task.pathName, pending_HLT, max_HLT))
if i_task == 0 and not sec and use_T0:
needs, task_name, running, idled = needs_action(wfi, task)
if options.augment: needs = True
#needs = True
good_type = wfi.request['RequestType'] in [
'MonteCarlo', 'MonteCarloFromGEN'
]
read_lhe = ((not 'LheInputFiles' in wfi.request)
or bool(wfi.request['LheInputFiles']))
good_type &= not read_lhe
if not good_type and not options.augment: needs = False
##needs = random.random()<0.40 remove the random, just add up to a limit
if (needs or needs_overide):
pending_T0 += idled
if task.pathName in modifications[
wfo.name] and 'AddWhitelist' in modifications[
wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][
task.pathName]["AddWhitelist"]:
modifications[wfo.name][task.pathName][
"AddWhitelist"].append("T0_CH_CERN")
wfi, sendLog(
'equalizor',
'adding the T0 for %s to %d for %d' %
(task.pathName, pending_T0, max_T0))
elif task.pathName in modifications[
wfo.
name] and 'ReplaceSiteWhitelist' in modifications[
wfo.name][task.pathName]:
if not "T0_CH_CERN" in modifications[wfo.name][
task.pathName]["ReplaceSiteWhitelist"]:
modifications[wfo.name][task.pathName][
"ReplaceSiteWhitelist"].append("T0_CH_CERN")
wfi, sendLog(
'equalizor',
'adding the T0 to replacement for %s to %d for %d'
% (task.pathName, pending_T0, max_T0))
else:
modifications[wfo.name][task.pathName] = {
"AddWhitelist": ["T0_CH_CERN"],
"Priority": wfi.request['RequestPriority'],
"Running": running,
"Pending": idled
}
wfi, sendLog(
'equalizor', 'adding the T0 for %s to %d for %d' %
(task.pathName, pending_T0, max_T0))
interface['modifications'].update(modifications)
### manage the number of core and job resizing
interface['cores'] = {
'T2_CH_CERN_HLT': {
'min': 4,
'max': 16
},
'default': {
'min': 1,
'max': 4
}
}
interface['resizes'] = ['RunIISpring16DR80']
### manage the modification of the memory and target time
interface['time'] = defaultdict(list)
interface['memory'] = defaultdict(list)
max_N_mem = 10
max_N_time = 10
## discretize the memory to 10 at most values
mems = set([o['memory'] for t, o in performance.items() if 'memory' in o])
times = set([o['time'] for t, o in performance.items() if 'time' in o])
if len(mems) > max_N_mem:
mem_step = int((max(mems) - min(mems)) / float(max_N_mem))
for t in performance:
if not 'memory' in performance[t]: continue
(m, r) = divmod(performance[t]['memory'], mem_step)
performance[t]['memory'] = (m + 1) * mem_step
if len(times) > max_N_time:
time_step = int((max(times) - min(times)) / float(max_N_time))
for t in performance:
if not 'time' in performance[t]: continue
(m, r) = divmod(performance[t]['time'], time_step)
performance[t]['time'] = (m + 1) * time_step
for t, o in performance.items():
if 'time' in o:
interface['time'][str(o['time'])].append(t)
if 'memory' in o:
interface['memory'][str(o['memory'])].append(t)
## close and save
close(interface)
def assignor(url, specific=None, talk=True, options=None):
if userLock() and not options.manual: return
mlock = moduleLock()
if mlock() and not options.manual: return
if not componentInfo().check() and not options.manual: return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = siteInfo()
SI = global_SI()
###NLI = newLockInfo()
###if not NLI.free() and not options.go: return
LI = lockInfo()
#if not LI.free() and not options.go and not options.manual: return
n_assigned = 0
n_stalled = 0
wfos = []
fetch_from = []
if specific or options.early:
fetch_from.extend(['considered', 'staging'])
if specific:
fetch_from.extend(['considered-tried'])
if options.early:
print "Option Early is on"
fetch_from.extend(['staged'])
if options.from_status:
fetch_from = options.from_status.split(',')
print "Overriding to read from", fetch_from
for status in fetch_from:
print "getting wf in", status
wfos.extend(
session.query(Workflow).filter(Workflow.status == status).all())
print len(wfos)
## in case of partial, go for fetching a list from json ?
#if options.partial and not specific:
# pass
aaa_mapping = json.loads(eosRead('%s/equalizor.json' %
monitor_pub_dir))['mapping']
all_stuck = set()
all_stuck.update(
json.loads(eosRead('%s/stuck_transfers.json' % monitor_pub_dir)))
max_per_round = UC.get('max_per_round').get('assignor', None)
max_cpuh_block = UC.get('max_cpuh_block')
# Temporarily switch off prioritization
random.shuffle(wfos)
##order by priority instead of random
"""
if options.early:
cache = sorted(getWorkflows(url, 'assignment-approved', details=True), key = lambda r : r['RequestPriority'])
cache = [r['RequestName'] for r in cache]
def rank( wfn ):
return cache.index( wfn ) if wfn in cache else 0
wfos = sorted(wfos, key = lambda wfo : rank( wfo.name ),reverse=True)
print "10 first",[wfo.name for wfo in wfos[:10]]
print "10 last",[wfo.name for wfo in wfos[-10:]]
else:
random.shuffle( wfos )
"""
for wfo in wfos:
if options.limit and (n_stalled + n_assigned) > options.limit:
break
if max_per_round and (n_stalled + n_assigned) > max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name, specific.split(','))):
continue
#if not specific in wfo.name: continue
if not options.manual and 'rucio' in (wfo.name).lower(): continue
print "\n\n"
wfh = workflowInfo(url, wfo.name)
if wfh.request['RequestStatus'] in [
'rejected', 'aborted', 'aborted-completed', 'aborted-archived',
'rejected-archived'
] and wfh.isRelval():
wfo.status = 'forget'
session.commit()
n_stalled += 1
continue
if options.priority and int(
wfh.request['RequestPriority']) < options.priority:
continue
options_text = ""
if options.early: options_text += ", early option is ON"
wfh.sendLog('assignor',
"%s to be assigned %s" % (wfo.name, options_text))
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput, primary, parent, secondary, sites_allowed,
sites_not_allowed) = wfh.getSiteWhiteList()
output_tiers = list(
set([o.split('/')[-1] for o in wfh.request['OutputDatasets']]))
if not output_tiers:
n_stalled += 1
wfh.sendLog('assignor', 'There is no output at all')
sendLog('assignor',
'Workflow %s has no output at all' % (wfo.name),
level='critical')
continue
is_stuck = (all_stuck & primary)
if is_stuck:
wfh.sendLog('assignor',
"%s are stuck input" % (','.join(is_stuck)))
## check if by configuration we gave it a GO
no_go = False
if not wfh.go(log=True) and not options.go:
no_go = True
allowed_secondary = {}
assign_parameters = {}
check_secondary = (not wfh.isRelval())
for campaign in wfh.getCampaigns():
if campaign in CI.campaigns:
assign_parameters.update(CI.campaigns[campaign])
if campaign in CI.campaigns and 'secondaries' in CI.campaigns[
campaign]:
if CI.campaigns[campaign]['secondaries']:
allowed_secondary.update(
CI.campaigns[campaign]['secondaries'])
check_secondary = True
if campaign in CI.campaigns and 'banned_tier' in CI.campaigns[
campaign]:
banned_tier = list(
set(CI.campaigns[campaign]['banned_tier'])
& set(output_tiers))
if banned_tier:
no_go = True
wfh.sendLog(
'assignor', 'These data tiers %s are not allowed' %
(','.join(banned_tier)))
sendLog('assignor',
'These data tiers %s are not allowed' %
(','.join(banned_tier)),
level='critical')
if secondary and check_secondary:
if (set(secondary) & set(allowed_secondary.keys()) !=
set(secondary)):
msg = '%s is not an allowed secondary' % (
', '.join(set(secondary) - set(allowed_secondary.keys())))
wfh.sendLog('assignor', msg)
critical_msg = msg + '\nWorkflow URL: https://dmytro.web.cern.ch/dmytro/cmsprodmon/workflows.php?prep_id=task_{}'.format(
wfh.getPrepIDs()[0])
sendLog('assignor', critical_msg, level='critical')
if not options.go:
no_go = True
## then get whether there is something more to be done by secondary
for sec in secondary:
if sec in allowed_secondary: # and 'parameters' in allowed_secondary[sec]:
assign_parameters.update(allowed_secondary[sec])
if no_go:
n_stalled += 1
## make a very loud noise if >100k priority stalled
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] != 'assignment-approved':
if not options.test:
wfh.sendLog('assignor',
"setting %s away and skipping" % wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name, wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version = wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor', "cannot decide on version number")
n_stalled += 1
wfo.status = 'trouble'
session.commit()
continue
wfh.sendLog('assignor', "Site white list %s" % sorted(sites_allowed))
blocks = wfh.getBlocks()
if blocks:
wfh.sendLog(
'assignor',
"Needs {} blocks in input {}".format(len(blocks),
'\n'.join(blocks)))
wfh.sendLog('assignor', "Allowed %s" % sorted(sites_allowed))
primary_aaa = options.primary_aaa
secondary_aaa = options.secondary_aaa
if 'Campaign' in wfh.request and wfh.request[
'Campaign'] in CI.campaigns:
assign_parameters.update(CI.campaigns[wfh.request['Campaign']])
if 'primary_AAA' in assign_parameters and primary:
primary_aaa = primary_aaa or assign_parameters['primary_AAA']
if 'secondary_AAA' in assign_parameters:
secondary_aaa = secondary_aaa or assign_parameters['secondary_AAA']
wfh.sendLog(
'assignor',
"Initial values for primary_AAA=%s and secondary_AAA=%s" %
(primary_aaa, secondary_aaa))
if primary_aaa:
if "T2_CH_CERN_HLT" in sites_allowed:
sites_allowed.remove("T2_CH_CERN_HLT")
if "T2_CH_CERN_HLT" not in sites_not_allowed:
sites_not_allowed.append("T2_CH_CERN_HLT")
## keep track of this, after secondary input location restriction : that's how you want to operate it
initial_sites_allowed = copy.deepcopy(sites_allowed)
set_lfn = '/store/mc' ## by default
for prim in list(primary):
set_lfn = getLFNbase(prim)
## if they are requested for processing, they should bbe all closed already
# FIXME: remove this closeAllBlocks
#closeAllBlocks(url, prim, blocks)
## should be 2 but for the time-being let's lower it to get things going
_copies_wanted, cpuh = wfh.getNCopies()
wfh.sendLog('assignor', "we need %s CPUh" % cpuh)
if cpuh > max_cpuh_block and not options.go:
#sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog(
'assignor',
'%s requires a large numbr of CPUh %s , not assigning, please check with requester'
% (wfo.name, cpuh),
level='critical')
wfh.sendLog(
'assignor',
"Requiring a large number of CPUh %s, not assigning" % cpuh)
continue
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
## default back to white list to original white list with any data
wfh.sendLog('assignor', "Allowed sites :%s" % sorted(sites_allowed))
# TODO Alan on 1/april/2020: keep the AAA functionality
if primary_aaa:
## remove the sites not reachable localy if not in having the data
if not sites_allowed:
wfh.sendLog('assignor',
"Overiding the primary on AAA setting to Off")
primary_aaa = False
else:
aaa_grid = set(sites_allowed)
for site in list(aaa_grid):
aaa_grid.update(aaa_mapping.get(site, []))
sites_allowed = list(set(initial_sites_allowed) & aaa_grid)
wfh.sendLog(
'assignor', "Selected to read primary through xrootd %s" %
sorted(sites_allowed))
isStoreResults = ('StoreResults' == wfh.request.setdefault(
'RequestType', None))
if isStoreResults:
if 'MergedLFNBase' in wfh.request:
set_lfn = wfh.request['MergedLFNBase']
else:
n_stalled += 1
wfh.sendLog(
'assignor',
"Cannot assign StoreResults request because MergedLFN is missing"
)
sendLog(
'assignor',
'Cannot assign StoreResults request because MergedLFN is missing',
level='critical')
continue
if not primary_aaa:
if isStoreResults:
## if we are dealing with a StoreResults request, we don't need to check dataset availability and
## should use the SiteWhiteList set in the original request
if 'SiteWhitelist' in wfh.request:
sites_allowed = wfh.request['SiteWhitelist']
else:
wfh.sendLog(
'assignor',
"Cannot assign StoreResults request because SiteWhitelist is missing"
)
sendLog(
'assignor',
'Cannot assign StoreResults request because SiteWhitelist is missing',
level='critical')
n_stalled += 1
continue
wfh.sendLog('assignor',
"Selected for any data %s" % sorted(sites_allowed))
#if not len(sites_allowed):
# if not options.early:
# wfh.sendLog('assignor',"cannot be assign with no matched sites")
# sendLog('assignor','%s has no whitelist'% wfo.name, level='critical')
# n_stalled+=1
# continue
if not len(sites_allowed) and not options.SiteWhitelist:
if not options.early:
wfh.sendLog('assignor',
"cannot be assign with no matched sites")
sendLog('assignor',
'%s has no whitelist' % wfo.name,
level='critical')
n_stalled += 1
continue
t1t2_only = [
ce for ce in sites_allowed
if [ce.startswith('T1') or ce.startswith('T2')]
]
if t1t2_only:
# try to pick from T1T2 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1t2_only])]
# then pick any otherwise
else:
sites_out = [
SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])
]
print "available=", SI.disk[sites_out[0]]
wfh.sendLog('assignor', "Placing the output on %s" % sites_out)
parameters = {
'SiteWhitelist': sites_allowed,
'SiteBlacklist': sites_not_allowed,
'NonCustodialSites': sites_out,
'AutoApproveSubscriptionSites': list(set(sites_out)),
'AcquisitionEra': wfh.acquisitionEra(),
'ProcessingString': wfh.processingString(),
'MergedLFNBase': set_lfn,
'ProcessingVersion': version,
}
if primary_aaa:
parameters['TrustSitelists'] = True
wfh.sendLog(
'assignor',
"Reading primary through xrootd at %s" % sorted(sites_allowed))
if secondary_aaa:
# Do not set TrustPUSitelist to True if there is no secondary
if secondary:
parameters['TrustPUSitelists'] = True
wfh.sendLog(
'assignor', "Reading secondary through xrootd at %s" %
sorted(sites_allowed))
## plain assignment here
team = 'production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
parameters['Team'] = team
if lheinput:
## throttle reading LHE article
wfh.sendLog('assignor',
'Setting the number of events per job to 500k max')
parameters['EventsPerJob'] = 500000
def pick_options(options, parameters):
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v = getattr(options, key)
if v != None:
if type(v) == str and ',' in v:
parameters[key] = filter(None, v.split(','))
else:
parameters[key] = v
def pick_campaign(assign_parameters, parameters):
## pick up campaign specific assignment parameters
parameters.update(assign_parameters.get('parameters', {}))
if options.force_options:
pick_campaign(assign_parameters, parameters)
pick_options(options, parameters)
else:
## campaign parameters update last
pick_options(options, parameters)
pick_campaign(assign_parameters, parameters)
if not options.test:
parameters['execute'] = True
hold_split, split_check = wfh.checkSplitting()
if hold_split and not options.go:
if split_check:
wfh.sendLog(
'assignor', 'Holding on to the change in splitting %s' %
('\n\n'.join([str(i) for i in split_check])))
else:
wfh.sendLog('assignor', 'Change of splitting is on hold')
n_stalled += 1
continue
if split_check == None or split_check == False:
n_stalled += 1
continue
elif split_check:
## operate all recommended changes
reqMgrClient.setWorkflowSplitting(url, wfo.name, split_check)
wfh.sendLog(
'assignor', 'Applying the change in splitting %s' %
('\n\n'.join([str(i) for i in split_check])))
split_check = True ## bypass completely and use the above
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi) / float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents / (reqJobs * 1.4))
lumisPerJob = int(eventsPerJob / eventsPerLumi)
if lumisPerJob == 0:
#sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog('assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob),
level='critical')
wfh.sendLog(
'assignor',
"%s needs to be split by event with %s per job" %
(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
# FIXME: decide which of the lines below needs to remain...
eventsPerJobEstimated = spl[
'events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl[
'avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
#sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog('assignor',
"%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob),
level='critical')
wfh.sendLog(
'assignor', "%s was assigned with %s lumis/job" %
(wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
#sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog(
'assignor',
"leaving splitting untouched for %s, please check on %s"
% (pstring, wfo.name),
level='critical')
wfh.sendLog(
'assignor',
"leaving splitting untouched for PU_RD*, please check."
)
## make sure to autoapprove all NonCustodialSites
parameters['AutoApproveSubscriptionSites'] = list(
set(parameters['NonCustodialSites'] +
parameters.get('AutoApproveSubscriptionSites', [])))
result = reqMgrClient.assignWorkflow(
url, wfo.name, None,
parameters) ## team is not relevant anymore here
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned += 1
wfh.sendLog(
'assignor', "Properly assigned\n%s" %
(json.dumps(parameters, indent=2)))
if wfh.producePremix() and (not wfh.isRelval()):
title = "Heavy workflow assigned to {}".format(
parameters['SiteWhitelist'])
body = "Workflow name: {}".format(
wfh.request['RequestName'])
body += "\nOutput dataset(s): {}".format(
wfh.request['OutputDatasets'])
body += "\nAssigned to: {}".format(
parameters['SiteWhitelist'])
sendEmail(
title,
body,
destination=[
'*****@*****.**'
])
try:
## refetch information and lock output
new_wfi = workflowInfo(url, wfo.name)
(_, prim, _, sec) = new_wfi.getIO()
for secure in list(prim) + list(
sec) + new_wfi.request['OutputDatasets']:
## lock all outputs
LI.lock(secure, reason='assigning')
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output", str(e))
else:
wfh.sendLog(
'assignor',
"Failed to assign %s.\n%s \n Please check the logs" %
(wfo.name, reqMgrClient.assignWorkflow.errorMessage))
sendLog('assignor',
"Failed to assign %s.\n%s \n Please check the logs" %
(wfo.name, reqMgrClient.assignWorkflow.errorMessage),
level='critical')
print "ERROR could not assign", wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor', "Assigned %d Stalled %s" % (n_assigned, n_stalled))
if n_stalled and not options.go and not options.early:
sendLog('assignor',
"%s workflows cannot be assigned. Please take a look" %
(n_stalled),
level='critical')
def stuckor(url=reqmgr_url):
if duplicateLock(): return
datasets_by_phid = json.loads(
open('%s/datasets_by_phid.json' % base_eos_dir).read())
really_stuck_dataset = set(
json.loads(open('%s/really_stuck_dataset.json' % base_eos_dir).read()))
UC = unifiedConfiguration()
print "make a report of stuck transfers"
bad_destinations = defaultdict(set)
bad_sources = defaultdict(set)
report = ""
transfer_timeout = UC.get("transfer_timeout")
transfer_lowrate = UC.get("transfer_lowrate")
for phid, datasets in datasets_by_phid.items():
issues = checkTransferLag(url, phid, datasets=list(datasets))
for dataset in issues:
for block in issues[dataset]:
for destination in issues[dataset][block]:
(block_size, destination_size, delay, rate,
dones) = issues[dataset][block][destination]
## count x_Buffer and x_MSS as one source
redones = []
for d in dones:
if d.endswith('Buffer') or d.endswith('Export'):
if d.replace('Buffer',
'MSS').replace('Export',
'MSS') in dones:
continue
else:
redones.append(d)
else:
redones.append(d)
dones = list(set(redones))
#dones = filter(lambda s : (s.endswith('Buffer') and not s.replace('Buffer','MSS') in dones) or (not s.endswith('Buffer')) , dones)
if delay > transfer_timeout and rate < transfer_lowrate:
if len(dones) > 1:
## its the destination that sucks
bad_destinations[destination].add(block)
else:
dum = [bad_sources[d].add(block) for d in dones]
really_stuck_dataset.add(dataset)
print "add", dataset, "to really stuck"
report += "%s is not getting to %s, out of %s faster than %f [GB/s] since %f [d]\n" % (
block, destination, ", ".join(dones), rate, delay)
print "\n" * 2
## create tickets right away ?
report += "\nbad sources " + ",".join(bad_sources.keys()) + "\n"
for site, blocks in bad_sources.items():
report += "\n\n%s:" % site + "\n\t".join([''] + list(blocks))
report += "\nbad destinations " + ",".join(bad_destinations.keys()) + "\n"
for site, blocks in bad_destinations.items():
report += "\n\n%s:" % site + "\n\t".join([''] + list(blocks))
print '\n' * 2, "Datasets really stuck"
print '\n'.join(really_stuck_dataset)
print '\n' * 2, "report written at %s/logs/incomplete_transfers.log" % unified_url
print report
missing_in_action = json.loads(
open('%s/incomplete_transfers.json' % monitor_dir).read())
stuck_transfers = dict([(k, v) for (k, v) in missing_in_action.items()
if k in really_stuck_dataset])
print '\n' * 2, 'Stuck dataset transfers'
print json.dumps(stuck_transfers, indent=2)
open('%s/stuck_transfers.json' % monitor_pub_dir,
'w').write(json.dumps(stuck_transfers, indent=2))
open('%s/logs/incomplete_transfers.log' % monitor_dir, 'w').write(report)
def completor(url, specific):
use_mcm = True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check(): return
use_mcm = up.status['mcm']
if use_mcm:
mcm = McMClient(dev=False)
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
wfs = []
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance')).all() )
## just take it in random order so that not always the same is seen
random.shuffle( wfs )
max_per_round = UC.get('max_per_round').get('completor',None)
if max_per_round and not specific: wfs = wfs[:max_per_round]
## by workflow a list of fraction / timestamps
completions = json.loads( open('%s/completions.json'%monitor_dir).read())
good_fractions = {}
timeout = {}
for c in CI.campaigns:
if 'force-complete' in CI.campaigns[c]:
good_fractions[c] = CI.campaigns[c]['force-complete']
if 'force-timeout' in CI.campaigns[c]:
timeout[c] = CI.campaigns[c]['force-timeout']
long_lasting = {}
overrides = getForceCompletes()
if use_mcm:
## add all workflow that mcm wants to get force completed
mcm_force = mcm.get('/restapi/requests/forcecomplete')
## assuming this will be a list of actual prepids
overrides['mcm'] = mcm_force
print "can force complete on"
print json.dumps( good_fractions ,indent=2)
print json.dumps( overrides, indent=2)
max_force = UC.get("max_force_complete")
#wfs_no_location_in_GQ = set()
#block_locations = defaultdict(lambda : defaultdict(list))
#wfs_no_location_in_GQ = defaultdict(list)
set_force_complete = set()
for wfo in wfs:
if specific and not specific in wfo.name: continue
print "looking at",wfo.name
## get all of the same
wfi = workflowInfo(url, wfo.name)
pids = wfi.getPrepIDs()
skip=False
if not any([c in wfo.name for c in good_fractions]): skip=True
for user,spec in overrides.items():
if wfi.request['RequestStatus']!='force-complete':
if any(s in wfo.name for s in spec) or (wfo.name in spec) or any(pid in spec for pid in pids) or any(s in pids for s in spec):
sendEmail('force-complete requested','%s is asking for %s to be force complete'%(user,wfo.name))
wfi = workflowInfo(url, wfo.name)
forceComplete(url , wfi )
skip=True
wfi.notifyRequestor("The workflow %s was force completed by request of %s"%(wfo.name,user), do_batch=False)
wfi.sendLog('completor','%s is asking for %s to be force complete'%(user,wfo.name))
break
if wfo.status.startswith('assistance'): skip = True
if skip:
continue
priority = wfi.request['RequestPriority']
if not 'Campaign' in wfi.request: continue
if not wfi.request['RequestStatus'] in ['acquired','running-open','running-closed']: continue
c = wfi.request['Campaign']
if not c in good_fractions: continue
good_fraction = good_fractions[c]
ignore_fraction = 2.
lumi_expected = None
event_expected = None
if not 'TotalInputEvents' in wfi.request:
if 'RequestNumEvents' in wfi.request:
event_expected = wfi.request['RequestNumEvents']
else:
print "truncated, cannot do anything"
continue
else:
lumi_expected = wfi.request['TotalInputLumis']
event_expected = wfi.request['TotalInputEvents']
now = time.mktime(time.gmtime()) / (60*60*24.)
running_log = filter(lambda change : change["Status"] in ["running-open","running-closed"],wfi.request['RequestTransition'])
if not running_log:
print "\tHas no running log"
# cannot figure out when the thing started running
continue
then = running_log[-1]['UpdateTime'] / (60.*60.*24.)
delay = now - then ## in days
(w,d) = divmod(delay, 7 )
print "\t"*int(w)+"Running since",delay,"[days] priority=",priority
monitor_delay = 7
allowed_delay = 14
if c in timeout:
allowed_delay = timeout[c]
monitor_delay = min(monitor_delay, allowed_delay)
### just skip if too early
if delay <= monitor_delay: continue
long_lasting[wfo.name] = { "delay" : delay }
percent_completions = {}
for output in wfi.request['OutputDatasets']:
if "/DQM" in output: continue ## that does not count
if not output in completions: completions[output] = { 'injected' : None, 'checkpoints' : [], 'workflow' : wfo.name}
## get completion fraction
event_count,lumi_count = getDatasetEventsAndLumis(dataset=output)
lumi_completion=0.
event_completion=0.
if lumi_expected:
lumi_completion = lumi_count / float( lumi_expected )
if event_expected:
event_completion = event_count / float( event_expected )
#take the less optimistic
percent_completions[output] = min( lumi_completion, event_completion )
completions[output]['checkpoints'].append( (now, event_completion ) )
if all([percent_completions[out] >= good_fraction for out in percent_completions]):
wfi.sendLog('completor', "all is above %s \n%s"%( good_fraction,
json.dumps( percent_completions, indent=2 )
))
else:
long_lasting[wfo.name].update({
'completion': sum(percent_completions.values()) / len(percent_completions),
'completions' : percent_completions
})
## do something about the agents this workflow is in
long_lasting[wfo.name]['agents'] = wfi.getAgents()
wfi.sendLog('completor', "%s not over bound %s\n%s"%(percent_completions.values(), good_fraction,
json.dumps( long_lasting[wfo.name]['agents'], indent=2) ))
continue
if all([percent_completions[out] >= ignore_fraction for out in percent_completions]):
print "all is done, just wait a bit"
continue
for output in percent_completions:
completions[output]['injected'] = then
#further check on delays
cpuh = wfi.getComputingTime(unit='d')
ran_at = wfi.request['SiteWhitelist']
wfi.sendLog('completor',"Required %s, time spend %s"%( cpuh, delay))
##### WILL FORCE COMPLETE BELOW
# only really force complete after n days
if delay <= allowed_delay: continue
## find ACDCs that might be running
if max_force>0:
forceComplete(url, wfi )
set_force_complete.add( wfo.name )
print "going for force-complete of",wfo.name
wfi.sendLog('completor','going for force completing')
wfi.notifyRequestor("The workflow %s was force completed for running too long"% wfo.name)
max_force -=1
else:
wfi.sendLog('completor',"too many completion this round, cannot force complete")
## do it once only for testing
#break
if set_force_complete:
sendLog('completor','The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
#sendEmail('set force-complete', 'The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
open('%s/completions.json'%monitor_dir,'w').write( json.dumps( completions , indent=2))
text="These have been running for long"
open('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 ))
for wf,info in sorted(long_lasting.items(), key=lambda tp:tp[1]['delay'], reverse=True):
delay = info['delay']
text += "\n %s : %s days"% (wf, delay)
if 'completion' in info:
text += " %d%%"%( info['completion']*100 )
#if wfs_no_location_in_GQ:
# sendEmail('workflow with no location in GQ',"there won't be able to run anytime soon\n%s"%( '\n'.join(wfs_no_location_in_GQ)))
#sendEmail("long lasting workflow",text)
## you can check the log
print text
def completor(url, specific):
mlock = moduleLock(silent=True)
if mlock(): return
use_mcm = True
up = componentInfo(soft=['mcm','wtc','jira'])
if not up.check(): return
use_mcm = up.status['mcm']
if use_mcm:
mcm = McMClient(dev=False)
safe_mode = False
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
JC = JIRAClient() if up.status.get('jira',False) else None
wfs = []
wfs.extend( session.query(Workflow).filter(Workflow.status == 'away').all() )
wfs.extend( session.query(Workflow).filter(Workflow.status.startswith('assistance')).all() )
## just take it in random order so that not always the same is seen
random.shuffle( wfs )
max_per_round = UC.get('max_per_round').get('completor',None)
if max_per_round and not specific: wfs = wfs[:max_per_round]
all_stuck = set()
## take into account what stagor was saying
for itry in range(5):
try:
all_stuck.update( json.loads( eosRead('%s/stuck_transfers.json'%monitor_pub_dir)))
break
except:
time.sleep(2)
for itry in range(5):
try:
## take into account the block that needed to be repositioned recently
all_stuck.update( [b.split('#')[0] for b in json.loads( eosRead('%s/missing_blocks.json'%monitor_dir)) ] )
break
except:
time.sleep(2)
## take into account all stuck block and dataset from transfer team
all_stuck.update( getAllStuckDataset())
good_fractions = {}
overdoing_fractions = {}
truncate_fractions = {}
timeout = {}
campaign_injection_delay = {}
for c in CI.campaigns:
if 'force-complete' in CI.campaigns[c]:
good_fractions[c] = CI.campaigns[c]['force-complete']
if 'truncate-complete' in CI.campaigns[c]:
truncate_fractions[c] = CI.campaigns[c]['truncate-complete']
if 'force-timeout' in CI.campaigns[c]:
timeout[c] = CI.campaigns[c]['force-timeout']
if 'injection-delay' in CI.campaigns[c]:
campaign_injection_delay[c] = CI.campaigns[c]['injection-delay']
if 'overdoing-complete' in CI.campaigns[c]:
overdoing_fractions[c] = CI.campaigns[c]['overdoing-complete']
long_lasting = {}
WI = wtcInfo()
overrides = WI.getForce()
if use_mcm:
## add all workflow that mcm wants to get force completed
mcm_force = mcm.get('/restapi/requests/forcecomplete')
## assuming this will be a list of actual prepids
overrides['mcm'] = mcm_force
print "can force complete on"
print json.dumps( good_fractions ,indent=2)
print "can truncate complete on"
print json.dumps( truncate_fractions ,indent=2)
print "can overide on"
print json.dumps( overrides, indent=2)
max_force = UC.get("max_force_complete")
max_priority = UC.get("max_tail_priority")
injection_delay_threshold = UC.get("injection_delay_threshold")
injection_delay_priority = UC.get("injection_delay_priority")
delay_priority_increase = UC.get("delay_priority_increase")
default_fraction_overdoing = UC.get('default_fraction_overdoing')
set_force_complete = set()
# priority and time above which to fire a JIRA
jira_priority_and_delays = { 110000 : 21,
90000 : 28,
# 80000 : 60,
#0 : 90
}
for wfo in wfs:
if specific and not specific in wfo.name: continue
print "looking at",wfo.name
## get all of the same
wfi = workflowInfo(url, wfo.name)
pids = wfi.getPrepIDs()
skip=False
campaigns = wfi.getCampaigns()
#if not any([c in good_fractions.keys() for c in campaigns]): skip=True
#if not any([c in truncate_fractions.keys() for c in campaigns]): skip=True
for user,spec in overrides.items():
if not spec: continue
spec = filter(None, spec)
if not wfi.request['RequestStatus'] in ['force-complete', 'completed']:
if any(s in wfo.name for s in spec) or (wfo.name in spec) or any(pid in spec for pid in pids) or any(s in pids for s in spec):
wfi = workflowInfo(url, wfo.name)
forceComplete(url , wfi )
skip=True
wfi.notifyRequestor("The workflow %s was force completed by request of %s"%(wfo.name,user), do_batch=False)
wfi.sendLog('completor','%s is asking for %s to be force complete'%(user,wfo.name))
break
if wfo.status.startswith('assistance'): skip = True
if skip:
continue
priority = wfi.request['RequestPriority']
if not 'Campaign' in wfi.request: continue
if not wfi.request['RequestStatus'] in ['acquired','running-open','running-closed']: continue
## until we can map the output to task ...
output_per_task = wfi.getOutputPerTask() ## can use that one, and follow mapping
good_fraction_per_out = {}
good_fraction_nodelay_per_out = {}
truncate_fraction_per_out = {}
#allowed_delay_per_out = {}
for task,outs in output_per_task.items():
task_campaign = wfi.getCampaignPerTask( task )
for out in outs:
good_fraction_per_out[out] = good_fractions.get(task_campaign,1000.)
good_fraction_nodelay_per_out[out] = overdoing_fractions.get(task_campaign,default_fraction_overdoing)
truncate_fraction_per_out[out] = truncate_fractions.get(task_campaign,1000.)
#allowed_delay_per_out[out] = timeout.get(task_campaign, 14)
#print "force at", json.dumps( good_fraction_per_out, indent=2)
#print "truncate at",json.dumps( truncate_fraction_per_out, indent=2)
now = time.mktime(time.gmtime()) / (60*60*24.)
priority_log = filter(lambda change: change['Priority'] == priority,wfi.request.get('PriorityTransition',[]))
if not priority_log:
print "\tHas no priority log"
priority_delay = 0
else:
then = max([change['UpdateTime'] for change in priority_log]) / (60.*60.*24.)
priority_delay = now - then ## in days
print "priority was set to",priority,priority_delay,"[days] ago"
running_log = filter(lambda change : change["Status"] in ["running-open","running-closed"],wfi.request['RequestTransition'])
if not running_log:
print "\tHas no running log"
delay = 0
else:
then = max([change['UpdateTime'] for change in running_log]) / (60.*60.*24.)
delay = now - then ## in days
#further check on delays
cpuh = wfi.getComputingTime(unit='d')
wfi.sendLog('completor',"Requires %.1f [CPUd], running since %.1f [day], last priority change to %d done %.1f [day] ago"%( cpuh, delay, priority, priority_delay))
if priority_delay!=0 and priority_delay < delay:
## regardless when it started running, set the delay to when priority was changed last
delay = priority_delay
## this is supposed to be the very initial request date, inherited from clones
injection_delay = None
original = wfi
if 'OriginalRequestName' in original.request:
## go up the clone chain
original = workflowInfo(url, original.request['OriginalRequestName'])
injected_log = filter(lambda change : change["Status"] in ["assignment-approved"],original.request['RequestTransition'])
if injected_log:
injected_on = injected_log[-1]['UpdateTime'] / (60.*60.*24.)
injection_delay = now - injected_on
delay_for_priority_increase = injection_delay
#delay_for_priority_increase = delay
(w,d) = divmod(delay, 7 )
print "\t"*int(w)+"Running since",delay,"[days] priority=",priority
pop_a_jira = False
ping_on_jira = 7 *(24*60*60) # 7 days
for jp,jd in jira_priority_and_delays.items():
if priority >= jp and delay >= jd: pop_a_jira = True
if pop_a_jira and JC:
j,reopened,just_created = JC.create_or_last( prepid = wfi.request['PrepID'],
priority = wfi.request['RequestPriority'],
label = 'Late',
reopen = True)
last_time = JC.last_time( j )
since_last_ping = time.mktime(time.gmtime()) - last_time
if since_last_ping > ping_on_jira or just_created:
j_comment = "Running since %.1f [days] at priority %d"%( delay, priority)
JC.comment(j.key, j_comment)
if delay_for_priority_increase!=None and delay_for_priority_increase > injection_delay_threshold and priority >= injection_delay_priority:
quantized = 5000 ## quantize priority
tail_cutting_priority = wfi.request['InitialPriority']+ int((delay_priority_increase * (delay_for_priority_increase - injection_delay_threshold) / 7) / quantized) * quantized
tail_cutting_priority += 101 ## to signal it is from this mechanism
tail_cutting_priority = min(400000, tail_cutting_priority) ## never go above 400k priority
tail_cutting_priority = max(tail_cutting_priority, priority) ## never go below the current value
if priority < tail_cutting_priority:
if max_priority:
sendLog('completor',"%s Injected since %s [days] priority=%s, increasing to %s"%(wfo.name,delay_for_priority_increase,priority, tail_cutting_priority), level='critical')
wfi.sendLog('completor','bumping priority to %d for being injected since %s'%( tail_cutting_priority, delay_for_priority_increase))
reqMgrClient.changePriorityWorkflow(url, wfo.name, tail_cutting_priority)
max_priority-=1
else:
sendLog('completor',"%s Injected since %s [days] priority=%s, would like to increase to %s"%(wfo.name,delay_for_priority_increase,priority, tail_cutting_priority), level='critical')
wfi.sendLog('completor','would like to bump priority to %d for being injected since %s'%( tail_cutting_priority, delay_for_priority_increase))
print "Could be changing the priority to higher value, but too many already were done"
_,prim,_,_ = wfi.getIO()
is_stuck = all_stuck & prim
if is_stuck: wfi.sendLog('completor','%s is stuck'%','.join(is_stuck))
monitor_delay = 7
allowed_delay = max([timeout.get(c,14) for c in campaigns])
monitor_delay = min(monitor_delay, allowed_delay)
### just skip if too early, just for the sake of not computing the completion fraction just now.
# maybe this is fast enough that we can do it for all
if delay <= monitor_delay:
print "not enough time has passed yet"
continue
long_lasting[wfo.name] = { "delay" : delay,
"injection_delay" : injection_delay }
percent_completions = wfi.getCompletionFraction(caller='completor')
if not percent_completions:
sendLog('completor','%s has no output at all'% wfo.name, level='critical')
continue
is_over_allowed_delay = (all([percent_completions[out] >= good_fraction_per_out.get(out,1000.) for out in percent_completions]) and delay >= allowed_delay)
is_over_truncation_delay = (is_stuck and (all([percent_completions[out] >= truncate_fraction_per_out.get(out,1000.) for out in percent_completions])) and delay >= allowed_delay)
is_over_completion = (all([percent_completions[out] >= good_fraction_nodelay_per_out.get(out,1000.) for out in percent_completions]))
if is_over_completion:
wfi.sendLog('completor', "all is over completed %s\n %s"%( json.dumps( good_fraction_nodelay_per_out, indent=2 ),
json.dumps( percent_completions, indent=2 )
))
elif is_over_allowed_delay:
wfi.sendLog('completor', "all is above %s \n%s"%( json.dumps(good_fraction_per_out, indent=2 ),
json.dumps( percent_completions, indent=2 )
))
elif is_over_truncation_delay:
wfi.sendLog('completor', "all is above %s truncation level, and the input is stuck\n%s"%( json.dumps(truncate_fraction_per_out, indent=2 ),
json.dumps( percent_completions, indent=2 ) ) )
else:
long_lasting[wfo.name].update({
'completion': sum(percent_completions.values()) / len(percent_completions),
'completions' : percent_completions
})
## do something about the agents this workflow is in
long_lasting[wfo.name]['agents'] = wfi.getAgents()
wfi.sendLog('completor', "%s not over bound \ncomplete at %s \n truncate at %s \nRunning %s"%(json.dumps( percent_completions, indent=2),
json.dumps(good_fraction_per_out, indent=2),
json.dumps( truncate_fraction_per_out, indent=2),
json.dumps( long_lasting[wfo.name]['agents'], indent=2) ))
continue
#for output in percent_completions:
# completions[output]['injected'] = then
ran_at = wfi.request['SiteWhitelist']
wfi.sendLog('completor',"Required %s, time spend %s"%( cpuh, delay))
##### WILL FORCE COMPLETE BELOW
# only really force complete after n days
## find ACDCs that might be running
if max_force>0:
print "going for force-complete of",wfo.name
if not safe_mode:
forceComplete(url, wfi )
set_force_complete.add( wfo.name )
wfi.sendLog('completor','going for force completing')
wfi.notifyRequestor("The workflow %s was force completed for running too long"% wfo.name)
max_force -=1
else:
sendEmail('completor', 'The workflow %s is ready for force complete, but completor is in safe mode'%wfo.name)
else:
wfi.sendLog('completor',"too many completion this round, cannot force complete")
if set_force_complete:
sendLog('completor','The followings were set force-complete \n%s'%('\n'.join(set_force_complete)))
#open('%s/completions.json'%monitor_dir,'w').write( json.dumps( completions , indent=2))
text="These have been running for long"
#open('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 ))
eosFile('%s/longlasting.json'%monitor_dir,'w').write( json.dumps( long_lasting, indent=2 )).close()
for wf,info in sorted(long_lasting.items(), key=lambda tp:tp[1]['delay'], reverse=True):
delay = info['delay']
text += "\n %s : %s days"% (wf, delay)
if 'completion' in info:
text += " %d%%"%( info['completion']*100 )
print text
def stagor(url, specific=None, options=None):
if not componentInfo().check(): return
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
TS = transferStatuses()
cached_transfer_statuses = TS.content()
transfer_statuses = {}
done_by_wf_id = {}
done_by_input = {}
completion_by_input = {}
good_enough = 100.0
lost_blocks = json.loads(
eosRead('%s/lost_blocks_datasets.json' % monitor_dir))
lost_files = json.loads(
eosRead('%s/lost_files_datasets.json' % monitor_dir))
known_lost_blocks = {}
known_lost_files = {}
for dataset in set(lost_blocks.keys() + lost_files.keys()):
b, f = findLostBlocksFiles(url, dataset)
if dataset in lost_blocks and not b:
print dataset, "has no really lost blocks"
else:
known_lost_blocks[dataset] = [i['name'] for i in b]
if dataset in lost_files and not f:
print dataset, "has no really lost files"
else:
known_lost_files[dataset] = [i['name'] for i in f]
def time_point(label="", sub_lap=False):
now = time.mktime(time.gmtime())
nows = time.asctime(time.gmtime())
print "Time check (%s) point at : %s" % (label, nows)
print "Since start: %s [s]" % (now - time_point.start)
if sub_lap:
print "Sub Lap : %s [s]" % (now - time_point.sub_lap)
time_point.sub_lap = now
else:
print "Lap : %s [s]" % (now - time_point.lap)
time_point.lap = now
time_point.sub_lap = now
time_point.sub_lap = time_point.lap = time_point.start = time.mktime(
time.gmtime())
time_point("Check cached transfer")
## collect all datasets that are needed for wf in staging, correcting the status of those that are not really in staging
wfois = []
needs = defaultdict(list)
needs_by_priority = defaultdict(list)
for wfo in session.query(Workflow).filter(
Workflow.status == 'staging').all():
wfi = workflowInfo(url, wfo.name)
if wfi.request['RequestStatus'] in [
'running-open', 'running-closed', 'completed', 'assigned',
'acquired'
]:
wfi.sendLog('stagor',
"is in status %s" % wfi.request['RequestStatus'])
wfo.status = 'away'
session.commit()
continue
if not wfi.request['RequestStatus'] in ['assignment-approved']:
## should be setting 'away' too
## that usually happens for relvals
if wfi.request['RequestStatus'] in [
'rejected', 'aborted', 'aborted-completed',
'aborted-archived', 'rejected-archived'
] and wfi.isRelval():
wfo.status = 'forget'
session.commit()
continue
else:
print wfo.name, "is", wfi.request['RequestStatus']
#sendEmail("wrong status in staging. debug","%s is in %s, should set away."%(wfo.name,wfi.request['RequestStatus']))
sendLog("stagor",
"%s is in %s, set away" %
(wfo.name, wfi.request['RequestStatus']),
level='critical')
wfo.status = 'away'
session.commit()
continue
wfois.append((wfo, wfi))
_, primaries, _, secondaries = wfi.getIO()
for dataset in list(primaries) + list(secondaries):
needs[wfo.name].append(dataset)
done_by_input[dataset] = {}
completion_by_input[dataset] = {}
needs_by_priority[wfi.request['RequestPriority']].append(dataset)
wfi.sendLog('stagor', '%s needs %s' % (wfo.name, dataset))
time_point("Check staging workflows")
open('%s/dataset_requirements.json' % monitor_dir,
'w').write(json.dumps(needs, indent=2))
for prio in needs_by_priority:
needs_by_priority[prio] = list(set(needs_by_priority[prio]))
open('%s/dataset_priorities.json' % monitor_dir,
'w').write(json.dumps(needs_by_priority, indent=2))
dataset_endpoints = defaultdict(set)
endpoint_in_downtime = defaultdict(set)
#endpoint_completed = defaultdict(set)
endpoint_incompleted = defaultdict(set)
#endpoint = defaultdict(set)
send_back_to_considered = set()
## first check if anything is inactive
all_actives = set([
transfer.phedexid for transfer in session.query(TransferImp).filter(
TransferImp.active).all()
])
for active_phedexid in all_actives:
skip = True
transfers_phedexid = session.query(TransferImp).filter(
TransferImp.phedexid == active_phedexid).all()
for imp in transfers_phedexid:
if imp.workflow.status == 'staging':
skip = False
sendLog(
'stagor', "\t%s is staging for %s" %
(imp.phedexid, imp.workflow.name))
if skip:
sendLog('stagor', "setting %s inactive" % active_phedexid)
for imp in transfers_phedexid:
imp.active = False
session.commit()
all_actives = sorted(
set([
transfer.phedexid for transfer in session.query(
TransferImp).filter(TransferImp.active).all()
]))
for phedexid in all_actives:
if specific: continue
## check on transfer completion
not_cached = False
if phedexid in cached_transfer_statuses:
### use a cache for transfer that already looked done
sendLog('stagor', "read %s from cache" % phedexid)
checks = cached_transfer_statuses[phedexid]
else:
## I actually would like to avoid that all I can
sendLog('stagor',
'Performing spurious transfer check on %s' % phedexid,
level='critical')
checks = checkTransferStatus(url, phedexid, nocollapse=True)
try:
print json.dumps(checks, indent=2)
except:
print checks
if not checks:
## this is going to bias quite heavily the rest of the code. we should abort here
#sendLog('stagor','Ending stagor because of skewed input from checkTransferStatus', level='critical')
#return False
sendLog(
'stagor',
'Stagor has got a skewed input from checkTransferStatus',
level='critical')
checks = {}
pass
else:
TS.add(phedexid, checks)
time_point("Check transfer status %s" % phedexid, sub_lap=True)
if not specific:
for dsname in checks:
if not dsname in done_by_input: done_by_input[dsname] = {}
if not dsname in completion_by_input:
completion_by_input[dsname] = {}
done_by_input[dsname][phedexid] = all(
map(lambda i: i >= good_enough, checks[dsname].values()))
completion_by_input[dsname][phedexid] = checks[dsname].values()
if checks:
sendLog(
'stagor', "Checks for %s are %s" %
(phedexid, [node.values() for node in checks.values()]))
done = all(
map(
lambda i: i >= good_enough,
list(
itertools.chain.from_iterable(
[node.values() for node in checks.values()]))))
else:
## it is empty, is that a sign that all is done and away ?
if not_cached:
print "Transfer status was not cached"
else:
print "ERROR with the scubscriptions API of ", phedexid
print "Most likely something else is overiding the transfer request. Need to work on finding the replacement automatically, if the replacement exists"
done = False
transfers_phedexid = session.query(TransferImp).filter(
TransferImp.phedexid == phedexid).all()
for imp in transfers_phedexid:
tr_wf = imp.workflow
if tr_wf: # and tr_wf.status == 'staging':
if not tr_wf.id in done_by_wf_id: done_by_wf_id[tr_wf.id] = {}
done_by_wf_id[tr_wf.id][phedexid] = done
if done:
imp.active = False
session.commit()
for ds in checks:
for s, v in checks[ds].items():
dataset_endpoints[ds].add(s)
if done:
sendLog('stagor', "%s is done" % phedexid)
TS.add(phedexid, checks)
else:
sendLog(
'stagor',
"%s is not finished %s" % (phedexid, pprint.pformat(checks)))
##pprint.pprint( checks )
## check if the destination is in down-time
for ds in checks:
sites_incomplete = [
SI.SE_to_CE(s) for s, v in checks[ds].items()
if v < good_enough
]
sites_incomplete_down = [
s for s in sites_incomplete if not s in SI.sites_ready
]
## no space means no transfer should go there : NO, it does not work in the long run
#sites_incomplete_down = [SI.SE_to_CE(s) for s,v in checks[ds].items() if (v<good_enough and (SI.disk[s]==0 or (not SI.SE_to_CE(s) in SI.sites_ready)))]
if sites_incomplete_down:
sendLog(
'stagor',
"%s are in downtime, while waiting for %s to get there"
% (",".join(sites_incomplete_down), ds))
endpoint_in_downtime[ds].update(sites_incomplete_down)
if sites_incomplete:
endpoint_incompleted[ds].update(sites_incomplete)
time_point("Check on-going transfers")
print "End points"
for k in dataset_endpoints:
dataset_endpoints[k] = list(dataset_endpoints[k])
print json.dumps(dataset_endpoints, indent=2)
print "End point in down time"
for k in endpoint_in_downtime:
endpoint_in_downtime[k] = list(endpoint_in_downtime[k])
print json.dumps(endpoint_in_downtime, indent=2)
print "End point incomplete in down time"
for k in endpoint_incompleted:
endpoint_incompleted[k] = list(endpoint_incompleted[k])
print json.dumps(endpoint_incompleted, indent=2)
#open('%s/transfer_statuses.json'%monitor_dir,'w').write( json.dumps( transfer_statuses, indent=2))
eosFile('%s/transfer_statuses.json' % monitor_dir,
'w').write(json.dumps(TS.content(), indent=2)).close()
eosFile('%s/dataset_endpoints.json' % monitor_dir,
'w').write(json.dumps(dataset_endpoints, indent=2)).close()
already_stuck = json.loads(
eosRead('%s/stuck_transfers.json' % monitor_pub_dir)).keys()
already_stuck.extend(getAllStuckDataset())
missing_in_action = defaultdict(list)
print "-" * 10, "Checking on workflows in staging", "-" * 10
#forget_about = ['/MinBias_TuneCUETP8M1_13TeV-pythia8/RunIISummer15GS-MCRUN2_71_V1-v2/GEN-SIM']
#for what in forget_about:
# if not done_by_input[what]:
# done_by_input[what] = {'fake':True}
## come back to workflows and check if they can go
available_cache = defaultdict(lambda: defaultdict(float))
presence_cache = defaultdict(dict)
time_point("Preparing for more")
for wfo, wfi in wfois:
print "#" * 30
time_point("Forward checking %s" % wfo.name, sub_lap=True)
## the site white list takes site, campaign, memory and core information
(_, primaries, _, secondaries,
sites_allowed) = wfi.getSiteWhiteList(verbose=False)
se_allowed = [SI.CE_to_SE(site) for site in sites_allowed]
se_allowed.sort()
se_allowed_key = ','.join(se_allowed)
readys = {}
for need in list(primaries) + list(secondaries):
if not need in done_by_input:
wfi.sendLog('stagor', "missing transfer report for %s" % need)
readys[need] = False
## should warn someone about this !!!
## it cannot happen, by construction
sendEmail('missing transfer report',
'%s does not have a transfer report' % (need))
continue
if not done_by_input[need] and need in list(secondaries):
wfi.sendLog(
'stagor',
"assuming it is OK for secondary %s to have no attached transfers"
% need)
readys[need] = True
done_by_input[need] = {"fake": True}
continue
if len(done_by_input[need]) and all(done_by_input[need].values()):
wfi.sendLog('stagor', "%s is ready" % need)
print json.dumps(done_by_input[need], indent=2)
readys[need] = True
else:
wfi.sendLog(
'stagor', "%s is not ready \n%s" %
(need, json.dumps(done_by_input[need], indent=2)))
readys[need] = False
if readys and all(readys.values()):
if wfo.status == 'staging':
wfi.sendLog('stagor',
"all needs are fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
wfi.sendLog('stagor', "all needs are fullfilled, already")
print json.dumps(readys, indent=2)
else:
wfi.sendLog('stagor', "missing requirements")
copies_needed, _ = wfi.getNCopies()
jump_ahead = False
re_transfer = False
## there is missing input let's do something more elaborated
for need in list(primaries): #+list(secondaries):
if endpoint_in_downtime[need] and endpoint_in_downtime[
need] == endpoint_incompleted[need]:
#print need,"is going to an end point in downtime"
wfi.sendLog(
'stagor',
"%s has only incomplete endpoint in downtime\n%s" %
(need, endpoint_in_downtime[need]))
re_transfer = True
if not se_allowed_key in available_cache[need]:
available_cache[need][
se_allowed_key] = getDatasetBlocksFraction(
url, need, sites=se_allowed)
if available_cache[need][se_allowed_key] >= copies_needed:
wfi.sendLog(
'stagor',
"assuming it is OK to move on like this already for %s"
% need)
jump_ahead = True
else:
wfi.sendLog(
'stagor', "Available %s times" %
available_cache[need][se_allowed_key])
missing_and_downtime = list(
set(endpoint_in_downtime[need])
& set(endpoint_incompleted[need]))
if missing_and_downtime:
wfi.sendLog(
'stagor',
"%s is incomplete at %s which is in downtime, trying to move along"
% (need, ','.join(missing_and_downtime)))
jump_ahead = True
else:
wfi.sendLog(
'stagor',
"continue waiting for transfers for optimum production performance."
)
## compute a time since staging to filter jump starting ?
# check whether the inputs is already in the stuck list ...
for need in list(primaries) + list(secondaries):
if need in already_stuck:
wfi.sendLog('stagor',
"%s is stuck, so try to jump ahead" % need)
jump_ahead = True
if jump_ahead or re_transfer:
details_text = "checking on availability for %s to jump ahead" % wfo.name
details_text += '\n%s wants %s copies' % (wfo.name,
copies_needed)
copies_needed = max(1, copies_needed - 1)
details_text += '\nlowering by one unit to %s' % copies_needed
wfi.sendLog('stagor', details_text)
all_check = True
prim_where = set()
for need in list(primaries):
if not se_allowed_key in presence_cache[need]:
presence_cache[need][
se_allowed_key] = getDatasetPresence(
url, need, within_sites=se_allowed)
presence = presence_cache[need][se_allowed_key]
prim_where.update(presence.keys())
available = available_cache[need][se_allowed_key]
this_check = (available >= copies_needed)
wfi.sendLog(
'stagor', "%s is available %s times (%s), at %s" %
(need, available, this_check, se_allowed_key))
all_check &= this_check
if not all_check: break
for need in list(secondaries):
## I do not want to check on the secon
## this below does not function because the primary could be all available, and the secondary not complete at a certain site that does not matter at that point
this_check = all(done_by_input[need].values())
wfi.sendLog(
'stagor', "%s is this much transfered %s" %
(need, json.dumps(done_by_input[need], indent=2)))
all_check &= this_check
#if not se_allowed_key in presence_cache[need]:
# presence_cache[need][se_allowed_key] = getDatasetPresence( url, need , within_sites=se_allowed)
## restrict to where the primary is
#presence = dict([(k,v) for (k,v) in presence_cache[need][se_allowed_key].items() if k in prim_where])
#this_check = all([there for (there,frac) in presence.values()])
#print need,"is present at all sites:",this_check
#all_check&= this_check
if all_check and not re_transfer:
wfi.sendLog(
'stagor',
"needs are sufficiently fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
print wfo.name, "has to wait a bit more"
wfi.sendLog('stagor', "needs to wait a bit more")
else:
wfi.sendLog('stagor', "not checking availability")
if re_transfer:
wfi.sendLog(
'stagor',
"Sending back to considered because of endpoint in downtime"
)
if wfo.status == 'staging':
wfo.status = 'considered'
session.commit()
send_back_to_considered.add(wfo.name)
time_point("Checked affected workflows")
if send_back_to_considered:
#sendEmail("transfer to endpoint in downtime","sending back to considered the following workflows \n%s"%('\n'.join( send_back_to_considered)))
sendLog('stagor',
"sending back to considered the following workflows \n%s" %
('\n'.join(send_back_to_considered)),
level='critical')
print "-" * 10, "Checking on non-available datasets", "-" * 10
## now check on those that are not fully available
for dsname in available_cache.keys():
## squash the se_allowed_key key
available_cache[dsname] = min(available_cache[dsname].values())
really_stuck_dataset = set()
for dsname, available in available_cache.items():
using_its = getWorkflowByInput(url, dsname)
#print using_its
using_wfos = []
for using_it in using_its:
wf = session.query(Workflow).filter(
Workflow.name == using_it).first()
if wf:
using_wfos.append(wf)
if not len(done_by_input[dsname]):
print "For dataset", dsname, "there are no transfer report. That's an issue."
for wf in using_wfos:
if wf.status == 'staging':
if UC.get("stagor_sends_back"):
print "sending", wf.name, "back to considered"
wf.status = 'considered'
session.commit()
#sendEmail( "send back to considered","%s was send back and might be trouble"% wf.name)
sendLog('stagor',
"%s was send back and might be trouble" %
wf.name,
level='critical')
else:
print "would send", wf.name, "back to considered"
#sendEmail( "subscription lagging behind","susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name)
sendLog(
'stagor',
"susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."
% wf.name,
level='critical')
continue
## not compatible with checking on secondary availability
#if all([wf.status != 'staging' for wf in using_wfos]):
# ## means despite all checks that input is not needed
# continue
if available < 1.:
print "incomplete", dsname
## there is a problem in the method below that it does not account for files stuck in T1*Buffer only
lost_blocks, lost_files = findLostBlocksFiles(
url, dsname) if (not dsname.endswith('/RAW')) else ([], [])
lost_block_names = [item['name'] for item in lost_blocks]
lost_file_names = [item['name'] for item in lost_files]
if lost_blocks:
#print json.dumps( lost , indent=2 )
## estimate for how much !
fraction_loss, _, n_missing = getDatasetBlockFraction(
dsname, lost_block_names)
print "We have lost", len(
lost_block_names
), "blocks", lost_block_names, "for %f%%" % (100. *
fraction_loss)
if fraction_loss > 0.05: ## 95% completion mark
#sendEmail('we have lost too many blocks','%s is missing %d blocks, for %d events, %f %% loss'%(dsname, len(lost_block_names), n_missing, fraction_loss))
sendLog(
'stagor',
'%s is missing %d blocks, for %d events, %3.2f %% loss'
% (dsname, len(lost_block_names), n_missing,
100 * fraction_loss),
level='critical')
## the workflow should be rejected !
for wf in using_wfos:
if wf.status == 'staging':
print wf.name, "is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
sendLog(
'stagor',
'%s has too much loss on the input dataset %s. Missing %d blocks, for %d events, %3.2f %% loss'
% (wf.name, dsname, len(lost_block_names),
n_missing, 100 * fraction_loss),
level='critical')
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_blocks:
#sendEmail('we have lost a few blocks', '%s is missing %d blocks, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_block_names), n_missing, fraction_loss, '\n'.join( lost_block_names ) ))
sendLog(
'stagor',
'%s is missing %d blocks, for %d events, %f %% loss\n\n%s'
% (dsname, len(lost_block_names), n_missing,
fraction_loss, '\n'.join(lost_block_names)),
level='critical')
known_lost_blocks[dsname] = [
i['name'] for i in lost_blocks
]
really_stuck_dataset.add(dsname)
if lost_files:
fraction_loss, _, n_missing = getDatasetFileFraction(
dsname, lost_file_names)
print "We have lost", len(
lost_file_names
), "files", lost_file_names, "for %f%%" % fraction_loss
if fraction_loss > 0.05:
#sendEmail('we have lost too many files','%s is missing %d files, for %d events, %f %% loss'%(dsname, len(lost_file_names),n_missing, fraction_loss))
sendLog(
'stagor',
'%s is missing %d files, for %d events, %f %% loss' %
(dsname, len(lost_file_names), n_missing,
fraction_loss),
level='critical')
for wf in using_wfos:
if wf.status == 'staging':
print wf.name, "is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_files:
#sendEmail('we have lost a few files','%s is missing %d files, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_file_names),n_missing, fraction_loss, '\n'.join(lost_file_names)))
sendLog(
'stagor',
'%s is missing %d files, for %d events, %f %% loss\n\n%s'
% (dsname, len(lost_file_names), n_missing,
fraction_loss, '\n'.join(lost_file_names)),
level='critical')
known_lost_files[dsname] = [
i['name'] for i in lost_files
]
## should the status be change to held-staging and pending on a ticket
missings = [
pid for (pid, d) in done_by_input[dsname].items() if d == False
]
print "\t", done_by_input[dsname]
print "\tneeds", len(done_by_input[dsname])
print "\tgot", done_by_input[dsname].values().count(True)
print "\tmissing", missings
missing_in_action[dsname].extend(missings)
rr = eosFile('%s/lost_blocks_datasets.json' % monitor_dir, 'w')
rr.write(json.dumps(known_lost_blocks, indent=2))
rr.close()
rr = eosFile('%s/lost_files_datasets.json' % monitor_dir, 'w')
rr.write(json.dumps(known_lost_files, indent=2))
rr.close()
eosFile('%s/incomplete_transfers.json' % monitor_dir,
'w').write(json.dumps(missing_in_action, indent=2)).close()
print "Stuck transfers and datasets"
print json.dumps(missing_in_action, indent=2)
TD = transferDataset()
datasets_by_phid = defaultdict(set)
for dataset in missing_in_action:
for phid in missing_in_action[dataset]:
#print dataset,"stuck through",phid
datasets_by_phid[phid].add(dataset)
for k in datasets_by_phid:
#datasets_by_phid[k] = list(datasets_by_phid[k])
TD.add(k, list(datasets_by_phid[k]))
#eosFile('%s/datasets_by_phid.json'%base_eos_dir,'w').write( json.dumps(datasets_by_phid, indent=2 )).close()
eosFile('%s/really_stuck_dataset.json' % base_eos_dir,
'w').write(json.dumps(list(really_stuck_dataset),
indent=2)).close()
print '\n' * 2, "Datasets really stuck"
print '\n'.join(really_stuck_dataset)
#############
## not going further for what matters
#############
return
def new_recoveror(url, specific, options=None):
if userLock('recoveror'): return
up = componentInfo(soft=['mcm','wtc','jira'])
if not up.check(): return
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
wfs = session.query(Workflow).filter(Workflow.status.contains('recovery')).all()
if specific:
wfs.extend( session.query(Workflow).filter(Workflow.status == 'assistance-manual').all() )
try:
from_operator = json.loads(os.popen('curl -s http://vocms0113.cern.ch/actions/test.json').read())
## now we have a list of things that we can take action on
except:
pass
for wfo in wfs:
if specific and not specific in wfo.name:continue
if not specific and 'manual' in wfo.status: continue
wfi = workflowInfo(url, wfo.name)
send_recovery = False ## will make all acdc
send_clone = False ## will make a clone
send_back = False ## should just reject. manual ?
send_manual = False ## will set in manual
where_to_run, missing_to_run = wfi.getRecoveryInfo()
task_to_recover = where_to_run.keys()
## if the site at which the recovery could run in drain or out ?
for task in task_to_recover:
not_ready = set(where_to_run[task]) - set(SI.sites_ready)
if not_ready:
print "the following sites are not ready for the ACDC",",".join( sorted(not_ready) )
## do we have a way of telling if a site is going to be out for a long time ?
# check on priority: high prio, restart
if wfi.request['RequestPriority'] >= 85000:
send_clone = True
# check on age of the request
injection_time = time.mktime(time.strptime('.'.join(map(str,wfi.request['RequestDate'])),"%Y.%m.%d.%H.%M.%S")) / (60.*60.)
now = time.mktime(time.gmtime()) / (60.*60.)
if float(now - injection_time) <14.:
## less than 14 days, start over
send_clone = True
else:
send_manual = True
if not send_recovery:
## check on whether the stats is very low
pass
if send_recovery:
## make acdc for all tasks
for task in task_to_recover:
actions = list(set([case['solution'] for code,case in task_to_recover[task] ]))
acdc = singleRecovery(url, task, wfi.request , actions, do = True)
elif send_clone:
## this will get it cloned
wfo.status = 'assistance-clone'
session.commit()
elif send_manual:
wfo.status = 'assistance-manual'
url = reqmgr_url
use_mcm=True
up = componentInfo(mcm=use_mcm, soft=['mcm'])
if not up.check():
sys.exit(1)
use_mcm = up.status['mcm']
mcm=None
if use_mcm:
print "mcm interface is up"
mcm = McMClient(dev=False)
statuses = ['assignment-approved','assigned','failed','acquired','running-open','running-closed','force-complete','completed','closed-out']
UC = unifiedConfiguration()
CI = campaignInfo()
tier_no_custodial = UC.get('tiers_with_no_custodial')
tiers_keep_on_disk = UC.get("tiers_keep_on_disk")
now = time.mktime( time.gmtime())
## can we catch the datasets that actually should go to tape ?
custodial_override = {}
for c in CI.campaigns:
if 'custodial_override' in CI.campaigns[c]:
custodial_override[c] = CI.campaigns[c]['custodial_override']
## those that are already in lock
already_locked = set(json.loads(open('%s/globallocks.json'%monitor_dir).read()))
if not already_locked:
def recoveror(url,specific,options=None):
if userLock('recoveror'): return
up = componentInfo(soft=['mcm','wtc','jira'])
if not up.check(): return
CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
use_recoveror = UC.get('use_recoveror')
if not use_recoveror and not options.go:
print "We are told not to run recoveror"
return
def make_int_keys( d ):
for code in d:
d[int(code)] = d.pop(code)
error_codes_to_recover = UC.get('error_codes_to_recover')
error_codes_to_block = UC.get('error_codes_to_block')
error_codes_to_notify = UC.get('error_codes_to_notify')
make_int_keys( error_codes_to_recover )
make_int_keys( error_codes_to_block )
make_int_keys( error_codes_to_notify )
#wfs = session.query(Workflow).filter(Workflow.status == 'assistance-recovery').all()
wfs = session.query(Workflow).filter(Workflow.status.contains('recovery')).all()
if specific:
wfs.extend( session.query(Workflow).filter(Workflow.status == 'assistance-manual').all() )
for wfo in wfs:
if specific and not specific in wfo.name:continue
if not specific and 'manual' in wfo.status: continue
wfi = workflowInfo(url, wfo.name)
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
all_errors = {}
try:
## this is clearly very truncated and should be changed completely
wfi.getSummary()
all_errors = wfi.summary['errors']
except:
pass
print '-'*100
print "Looking at",wfo.name,"for recovery options"
recover = True
if not 'MergedLFNBase' in wfi.request:
print "f****d up"
sendEmail('missing lfn','%s wl cache is screwed up'%wfo.name)
recover = False
if not len(all_errors):
print "\tno error for",wfo.name
recover = False
task_to_recover = defaultdict(list)
message_to_ops = ""
message_to_user = ""
if 'LheInputFilese' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
recover = False
if wfi.request['RequestType'] in ['MonteCarlo','ReReco']:
recover = False
if 'Campaign' in wfi.request:
c = wfi.request['Campaign']
if c in CI.campaigns and 'recover' in CI.campaigns[c]:
recover=CI.campaigns[c]['recover']
for task,errors in all_errors.items():
print "\tTask",task
## collect all error codes and #jobs regardless of step at which it occured
all_codes = []
for name, codes in errors.items():
if type(codes)==int: continue
all_codes.extend( [(int(code),info['jobs'],name,list(set([e['type'] for e in info['errors']])),list(set([e['details'] for e in info['errors']])) ) for code,info in codes.items()] )
all_codes.sort(key=lambda i:i[1], reverse=True)
sum_failed = sum([l[1] for l in all_codes])
for errorCode,njobs,name,types,details in all_codes:
rate = 100*njobs/float(sum_failed)
#print ("\t\t %10d (%6s%%) failures with error code %10d (%"+str(max_legend)+"s) at stage %s")%(njobs, "%4.2f"%rate, errorCode, legend, name)
print ("\t\t %10d (%6s%%) failures with error code %10d (%30s) at stage %s")%(njobs, "%4.2f"%rate, errorCode, ','.join(types), name)
added_in_recover=False
#if options.go:
# force the recovery of any task with error ?
if errorCode in error_codes_to_recover:
## the error code is registered
for case in error_codes_to_recover[errorCode]:
match = case['details']
matched= (match==None)
if not matched:
matched=False
for detail in details:
if match in detail:
print "[recover] Could find keyword",match,"in"
print 50*"#"
print detail
print 50*"#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => we should be able to recover that", case['legend']
task_to_recover[task].append( (code,case) )
added_in_recover=True
message_to_user = ""
else:
print "\t\t recoverable but not frequent enough, needs",case['rate']
if errorCode in error_codes_to_block:
for case in error_codes_to_block[errorCode]:
match = case['details']
matched= (match==None)
if not matched:
matched=False
for detail in details:
if match in detail:
print "[block] Could find keyword",match,"in"
print 50*"#"
print detail
print 50*"#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => that error means no ACDC on that workflow", case['legend']
if not options.go:
message_to_ops += "%s has an error %s blocking an ACDC.\n%s\n "%( wfo.name, errorCode, '#'*50 )
recover = False
added_in_recover=False
if errorCode in error_codes_to_notify and not added_in_recover:
print "\t\t => we should notify people on this"
message_to_user += "%s has an error %s in processing.\n%s\n" %( wfo.name, errorCode, '#'*50 )
if message_to_user:
print wfo.name,"to be notified to user(DUMMY)",message_to_user
if message_to_ops:
#sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**'])
sendLog('recoveror',message_to_ops,level='warning')
if len(task_to_recover) != len(all_errors):
print "Should not be doing partial ACDC. skipping"
#sendEmail('recoveror','do not want to make partial acdc on %s'%wfo.name)
sendLog('recoveror','do not want to make partial acdc on %s'%wfo.name, level='warning')
recover = False
if task_to_recover and recover:
print "Initiating recovery"
print ', '.join(task_to_recover.keys()),"to be recovered"
recovering=set()
for task in task_to_recover:
print "Will be making a recovery workflow for",task
## from here you can fetch known solutions, to known error codes
actions = list(set([case['solution'] for code,case in task_to_recover[task] ]))
acdc = singleRecovery(url, task, wfi.request , actions, do = options.do)
if not acdc:
if options.do:
if recovering:
print wfo.name,"has been partially ACDCed. Needs manual attention"
#sendEmail( "failed ACDC partial recovery","%s has had %s/%s recoveries %s only"%( wfo.name, len(recovering), len(task_to_recover), list(recovering)), destination=['*****@*****.**'])
sendLog('recoveror', "%s has had %s/%s recoveries %s only"%( wfo.name, len(recovering), len(task_to_recover), list(recovering)), level='critical')
continue
else:
print wfo.name,"failed recovery once"
#break
continue
else:
print "no action to take further"
sendLog('recoveror', "ACDC for %s can be done automatically"% wfo.name, level='critical')
continue
## and assign it ?
team = wfi.request['Team']
#assign_to_sites = set(SI.sites_ready) ## that needs to be massaged to prevent assigning to something out.
assign_to_sites = set(SI.all_sites)
parameters={
#'SiteWhitelist' : wfi.request['SiteWhitelist'],
'SiteWhitelist' : sorted(assign_to_sites),
'AcquisitionEra' : wfi.acquisitionEra(),
'ProcessingString' : wfi.processingString(),
'MergedLFNBase' : wfi.request['MergedLFNBase'],
'ProcessingVersion' : wfi.request['ProcessingVersion'],
}
## hackery for ACDC merge assignment
if wfi.request['RequestType'] == 'TaskChain' and 'Merge' in task.split('/')[-1]:
parameters['AcquisitionEra'] = None
parameters['ProcessingString'] = None
## xrootd setttings on primary and secondary
if 'TrustSitelists' in wfi.request and wfi.request['TrustSitelists']:
parameters['TrustSitelists'] = True
if 'TrustPUSitelists' in wfi.request and wfi.request['TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
if options.ass:
print "really doing the assignment of the ACDC",acdc
parameters['execute']=True
wfi.sendLog('recoveror',"%s was assigned for recovery"% acdc)
else:
print "no assignment done with this ACDC",acdc
sendLog('recoveror',"%s needs to be assigned"%(acdc), level='critical')
result = reqMgrClient.assignWorkflow(url, acdc, team, parameters)
if not result:
print acdc,"was not asigned"
sendLog('recoveror',"%s needs to be assigned"%(acdc), level='critical')
else:
recovering.add( acdc )
current = None
if recovering:
#if all went well, set the status to -recovering
current = wfo.status
if options.ass:
current = current.replace('recovery','recovering')
else:
current = 'assistance-manual'
print 'created ACDC: '+', '.join( recovering )
else:
## was set to be recovered, and no acdc was made
current = 'assistance-manual'
if current:
print wfo.name,"setting the status to",current
wfo.status = current
session.commit()
else:
## this workflow should be handled manually at that point
print wfo.name,"needs manual intervention"
wfo.status = 'assistance-manual'
session.commit()
def recoveror(url,specific,options=None):
if userLock('recoveror'): return
up = componentInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
def make_int_keys( d ):
for code in d:
d[int(code)] = d.pop(code)
error_codes_to_recover = UC.get('error_codes_to_recover')
error_codes_to_block = UC.get('error_codes_to_block')
error_codes_to_notify = UC.get('error_codes_to_notify')
make_int_keys( error_codes_to_recover )
make_int_keys( error_codes_to_block )
make_int_keys( error_codes_to_notify )
wfs = session.query(Workflow).filter(Workflow.status == 'assistance-recovery').all()
if specific:
wfs.extend( session.query(Workflow).filter(Workflow.status == 'assistance-manual').all() )
for wfo in wfs:
if specific and not specific in wfo.name:continue
if not specific and 'manual' in wfo.status: continue
wfi = workflowInfo(url, wfo.name, deprecated=True) ## need deprecated info for mergedlfnbase
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
all_errors = None
try:
wfi.getSummary()
all_errors = wfi.summary['errors']
except:
pass
print '-'*100
print "Looking at",wfo.name,"for recovery options"
if not len(all_errors):
print "\tno error for",wfo.name
task_to_recover = defaultdict(list)
message_to_ops = ""
message_to_user = ""
recover=True
if 'LheInputFilese' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
recover = False
if 'Campaign' in wfi.request:
c = wfi.request['Campaign']
if c in CI.campaigns and 'recover' in CI.campaigns[c]:
recover=CI.campaigns[c]['recover']
for task,errors in all_errors.items():
print "\tTask",task
## collect all error codes and #jobs regardless of step at which it occured
all_codes = []
for name, codes in errors.items():
if type(codes)==int: continue
all_codes.extend( [(int(code),info['jobs'],name,list(set([e['type'] for e in info['errors']])),list(set([e['details'] for e in info['errors']])) ) for code,info in codes.items()] )
all_codes.sort(key=lambda i:i[1], reverse=True)
sum_failed = sum([l[1] for l in all_codes])
for errorCode,njobs,name,types,details in all_codes:
rate = 100*njobs/float(sum_failed)
#print ("\t\t %10d (%6s%%) failures with error code %10d (%"+str(max_legend)+"s) at stage %s")%(njobs, "%4.2f"%rate, errorCode, legend, name)
print ("\t\t %10d (%6s%%) failures with error code %10d (%30s) at stage %s")%(njobs, "%4.2f"%rate, errorCode, ','.join(types), name)
added_in_recover=False
#if options.go:
# force the recovery of any task with error ?
if errorCode in error_codes_to_recover:
## the error code is registered
for case in error_codes_to_recover[errorCode]:
match = case['details']
matched= (match==None)
if not matched:
matched=False
for detail in details:
if match in detail:
print "[recover] Could find keyword",match,"in"
print 50*"#"
print detail
print 50*"#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => we should be able to recover that", case['legend']
task_to_recover[task].append( (code,case) )
added_in_recover=True
message_to_user = ""
else:
print "\t\t recoverable but not frequent enough, needs",case['rate']
if errorCode in error_codes_to_block:
for case in error_codes_to_block[errorCode]:
match = case['details']
matched= (match==None)
if not matched:
matched=False
for detail in details:
if match in detail:
print "[block] Could find keyword",match,"in"
print 50*"#"
print detail
print 50*"#"
matched = True
break
if matched and rate > case['rate']:
print "\t\t => that error means no ACDC on that workflow", case['legend']
if not options.go:
message_to_ops += "%s has an error %s blocking an ACDC.\n%s\n "%( wfo.name, errorCode, '#'*50 )
recover = False
added_in_recover=False
if errorCode in error_codes_to_notify and not added_in_recover:
print "\t\t => we should notify people on this"
message_to_user += "%s has an error %s in processing.\n%s\n" %( wfo.name, errorCode, '#'*50 )
if message_to_user:
print wfo.name,"to be notified to user(DUMMY)",message_to_user
if message_to_ops:
sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**','*****@*****.**'])
if task_to_recover and recover:
print "Initiating recovery"
print ', '.join(task_to_recover.keys()),"to be recovered"
recovering=set()
for task in task_to_recover:
print "Will be making a recovery workflow for",task
## from here you can fetch known solutions, to known error codes
actions = list(set([case['solution'] for code,case in task_to_recover[task] ]))
acdc = singleRecovery(url, task, wfi.request , actions, do = options.do)
if not acdc:
if options.do:
if recovering:
print wfo.name,"has been partially ACDCed. Needs manual attention"
sendEmail( "failed ACDC partial recovery","%s has had %s/%s recoveries %s only"%( wfo.name, len(recovering), len(task_to_recover), list(recovering)), destination=['*****@*****.**','*****@*****.**'])
continue
else:
print wfo.name,"failed recovery once"
break
else:
print "no action to take further"
sendEmail("an ACDC that can be done automatically","please check https://cmst2.web.cern.ch/cmst2/unified/logs/recoveror/last.log for details", destination=['*****@*****.**','*****@*****.**'])
continue
## and assign it ?
team = wfi.request['Teams'][0]
parameters={
'SiteWhitelist' : wfi.request['SiteWhitelist'],
'AcquisitionEra' : wfi.acquisitionEra(),
'ProcessingString' : wfi.processingString(),
'MergedLFNBase' : wfi.deprecated_request['MergedLFNBase'],
'ProcessingVersion' : wfi.request['ProcessingVersion'],
}
if options.ass:
print "really doing the assignment of the ACDC",acdc
parameters['execute']=True
sendEmail("an ACDC was done and WAS assigned", "%s was assigned, please check https://cmst2.web.cern.ch/cmst2/unified/logs/recoveror/last.log for details"%( acdc ), destination=['*****@*****.**','*****@*****.**'])
else:
print "no assignment done with this ACDC",acdc
sendEmail("an ACDC was done and need to be assigned", "%s needs to be assigned, please check https://cmst2.web.cern.ch/cmst2/unified/logs/recoveror/last.log for details"%( acdc ), destination=['*****@*****.**','*****@*****.**'])
result = reqMgrClient.assignWorkflow(url, acdc, team, parameters)
recovering.add( acdc )
if recovering:
#if all went well, set the status to -recovering
current = wfo.status
if options.ass:
current = current.replace('recovery','recovering')
else:
current = 'assistance-manual'
print wfo.name,"setting the status to",current
print ', '.join( recovering )
wfo.status = current
session.commit()
else:
## this workflow should be handled manually at that point
print wfo.name,"needs manual intervention"
wfo.status = 'assistance-manual'
session.commit()
def assignor(url, specific=None, talk=True, options=None):
if userLock():
return
if duplicateLock():
return
if not componentInfo().check():
return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = siteInfo()
NLI = newLockInfo()
n_assigned = 0
n_stalled = 0
wfos = []
if specific or options.early:
wfos.extend(session.query(Workflow).filter(Workflow.status == "considered").all())
wfos.extend(session.query(Workflow).filter(Workflow.status == "staging").all())
if specific:
wfos.extend(session.query(Workflow).filter(Workflow.status == "considered-tried").all())
wfos.extend(session.query(Workflow).filter(Workflow.status == "staged").all())
dataset_endpoints = json.loads(open("%s/dataset_endpoints.json" % monitor_dir).read())
max_per_round = UC.get("max_per_round").get("assignor", None)
max_cpuh_block = UC.get("max_cpuh_block")
random.shuffle(wfos)
for wfo in wfos:
if options.limit and (n_stalled + n_assigned) > options.limit:
break
if max_per_round and (n_stalled + n_assigned) > max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name, specific.split(","))):
continue
# if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo(url, wfo.name)
wfh.sendLog("assignor", "%s to be assigned" % wfo.name)
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput, primary, parent, secondary, sites_allowed) = wfh.getSiteWhiteList()
## check if by configuration we gave it a GO
no_go = False
if not wfh.go(log=True) and not options.go:
n_stalled += 1
no_go = True
allowed_secondary = set()
for campaign in wfh.getCampaigns():
if campaign in CI.campaigns and "secondaries" in CI.campaigns[campaign]:
allowed_secondary.update(CI.campaigns[campaign]["secondaries"])
if (secondary and allowed_secondary) and (set(secondary) & allowed_secondary != set(secondary)):
wfh.sendLog("assignor", "%s is not an allowed secondary" % (", ".join(set(secondary) - allowed_secondary)))
# sendEmail('secondary not allowed','%s is not an allowed secondary'%( ', '.join(set(secondary)-allowed_secondary)))
sendLog(
"assignor",
"%s is not an allowed secondary" % (", ".join(set(secondary) - allowed_secondary)),
level="critical",
)
if not options.go:
n_stalled += 1
no_go = True
if no_go:
continue
## check on current status for by-passed assignment
if wfh.request["RequestStatus"] != "assignment-approved":
if not options.test:
wfh.sendLog("assignor", "setting %s away and skipping" % wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request["RequestStatus"]
wfo.status = "away"
session.commit()
continue
else:
print wfo.name, wfh.request["RequestStatus"]
## retrieve from the schema, dbs and reqMgr what should be the next version
version = wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog("assignor", "cannot decide on version number")
n_stalled += 1
wfo.status = "trouble"
session.commit()
continue
original_sites_allowed = copy.deepcopy(sites_allowed)
wfh.sendLog("assignor", "Site white list %s" % sorted(sites_allowed))
override_sec_location = CI.get(wfh.request["Campaign"], "SecondaryLocation", [])
blocks = []
if "BlockWhitelist" in wfh.request:
blocks = wfh.request["BlockWhitelist"]
if "RunWhitelist" in wfh.request and wfh.request["RunWhitelist"]:
## augment with run white list
for dataset in primary:
blocks = list(set(blocks + getDatasetBlocks(dataset, runs=wfh.request["RunWhitelist"])))
wfh.sendLog("assignor", "Allowed %s" % sorted(sites_allowed))
secondary_locations = None
primary_aaa = options.primary_aaa
if (
"Campaign" in wfh.request
and wfh.request["Campaign"] in CI.campaigns
and "primary_AAA" in CI.campaigns[wfh.request["Campaign"]]
):
primary_aaa = primary_aaa or CI.campaigns[wfh.request["Campaign"]]["primary_AAA"]
secondary_aaa = options.secondary_aaa
if (
"Campaign" in wfh.request
and wfh.request["Campaign"] in CI.campaigns
and "secondary_AAA" in CI.campaigns[wfh.request["Campaign"]]
):
secondary_aaa = secondary_aaa or CI.campaigns[wfh.request["Campaign"]]["secondary_AAA"]
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
sendEmail("tempting to pass sec location check", "but we cannot yet IMO")
# pass
if secondary_aaa:
# just continue without checking
continue
presence = getDatasetPresence(url, sec)
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [site for (site, (there, frac)) in presence.items() if frac > 98.0]
# one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations == None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
# sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [site for site in sites_allowed if SI.CE_to_SE(site) in one_secondary_locations]
wfh.sendLog("assignor", "From secondary requirement, now Allowed%s" % sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy(
sites_allowed
) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy(sites_allowed)
sites_with_data = copy.deepcopy(sites_allowed)
sites_with_any_data = copy.deepcopy(sites_allowed)
primary_locations = None
available_fractions = {}
set_lfn = "/store/mc" ## by default
endpoints = set()
for prim in list(primary):
if prim in dataset_endpoints:
print "endpoints from stagor", dataset_endpoints[prim]
endpoints.update(dataset_endpoints[prim])
set_lfn = getLFNbase(prim)
presence = getDatasetPresence(url, prim, only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(
url, prim, sites=[SI.CE_to_SE(site) for site in sites_allowed], only_blocks=blocks
)
# sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
# sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [
site
for site in sites_with_data
if SI.CE_to_SE(site) in [psite for (psite, (there, frac)) in presence.items() if there]
]
sites_with_data = [
site
for site in sites_with_data
if SI.CE_to_SE(site) in [psite for (psite, frac) in presence.items() if frac[1] > 90.0]
]
sites_with_any_data = [site for site in sites_with_any_data if SI.CE_to_SE(site) in presence.keys()]
wfh.sendLog(
"assignor",
"Holding the data but not allowed %s"
% sorted(
list(set([se_site for se_site in presence.keys() if not SI.SE_to_CE(se_site) in sites_allowed]))
),
)
if primary_locations == None:
primary_locations = presence.keys()
else:
primary_locations = list(set(primary_locations) & set(presence.keys()))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites = []
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
# opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site)
for site in list(
(set(secondary_locations) & set(primary_locations))
- set([SI.CE_to_SE(site) for site in sites_allowed])
)
]
elif primary_locations:
opportunistic_sites = [
SI.SE_to_CE(site)
for site in list(set(primary_locations) - set([SI.CE_to_SE(site) for site in sites_allowed]))
]
else:
opportunistic_sites = []
wfh.sendLog("assignor", "We could be running in addition at %s" % sorted(opportunistic_sites))
if any([osite in SI.sites_not_ready for osite in opportunistic_sites]):
wfh.sendLog(
"assignor",
"One of the usable site is in downtime %s"
% ([osite in SI.sites_not_ready for osite in opportunistic_sites]),
)
down_time = True
## should this be send back to considered ?
## should be 2 but for the time-being let's lower it to get things going
copies_wanted, cpuh = wfh.getNCopies()
wfh.sendLog("assignor", "we need %s CPUh" % cpuh)
if cpuh > max_cpuh_block and not options.go:
# sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog(
"assignor",
"%s requires a large numbr of CPUh %s , not assigning, please check with requester" % (wfo.name, cpuh),
level="critical",
)
wfh.sendLog("assignor", "Requiring a large number of CPUh %s, not assigning" % cpuh)
continue
if (
"Campaign" in wfh.request
and wfh.request["Campaign"] in CI.campaigns
and "maxcopies" in CI.campaigns[wfh.request["Campaign"]]
):
copies_needed_from_campaign = CI.campaigns[wfh.request["Campaign"]]["maxcopies"]
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(1, copies_wanted - less_copies_than_requested) # take one out for the efficiency
else:
## find out whether there is a site in the whitelist, that is lacking jobs and reduce to 1 copy needed to get things going
pass
wfh.sendLog("assignor", "needed availability fraction %s" % copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
if available_fractions and not all([available >= copies_wanted for available in available_fractions.values()]):
not_even_once = not all([available >= 1.0 for available in available_fractions.values()])
wfh.sendLog(
"assignor",
"The input dataset is not available %s times, only %s" % (copies_wanted, available_fractions.values()),
)
if down_time and not options.go and not options.early:
wfo.status = "considered"
session.commit()
wfh.sendLog("assignor", "sending back to considered because of site downtime, instead of waiting")
# sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
sendLog(
"assignor",
"%s is not sufficiently available, due to down time of a site in the whitelist. sending back to considered."
% (wfo.name),
level="delay",
)
continue
# pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open("cannot_assign.json").read())
except:
pass
if (
not wfo.name in known
and not options.limit
and not options.go
and not options.early
and not options.partial
):
wfh.sendLog(
"assignor",
"cannot be assigned, %s is not sufficiently available.\n %s"
% (wfo.name, json.dumps(available_fractions)),
)
sendEmail(
"cannot be assigned",
"%s is not sufficiently available.\n %s" % (wfo.name, json.dumps(available_fractions)),
)
known.append(wfo.name)
open("cannot_assign.json", "w").write(json.dumps(known, indent=2))
n_stalled += 1
if options.early:
if wfo.status == "considered":
wfh.sendLog("assignor", "setting considered-tried")
wfo.status = "considered-tried"
session.commit()
else:
print "tried but status is", wfo.status
if options.partial:
print "Will move on with partial locations"
else:
continue
## default back to white list to original white list with any data
print "Allowed", sorted(sites_allowed)
if primary_aaa:
sites_allowed = initial_sites_allowed
options.TrustSitelists = True
wfh.sendLog("assignor", "Selected to read primary through xrootd %s" % sorted(sites_allowed))
else:
sites_allowed = sites_with_any_data
wfh.sendLog("assignor", "Selected for any data %s" % sorted(sites_allowed))
if secondary_aaa:
options.TrustPUSitelists = True
wfh.sendLog("assignor", "Reading secondary through xrootd from %s" % sorted(sites_allowed))
### check on endpoints for on-going transfers
if endpoints and options.partial:
sites_allowed = list(set(sites_allowed + [SI.SE_to_CE(s) for s in endpoints]))
print "with added endpoints", sorted(sites_allowed)
if not len(sites_allowed):
wfh.sendLog("assignor", "cannot be assign with no matched sites")
sendLog("assignor", "%s has no whitelist" % wfo.name, level="critical")
n_stalled += 1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith("T1")]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])]
wfh.sendLog("assignor", "Placing the output on %s" % sites_out)
parameters = {
"SiteWhitelist": sites_allowed,
"NonCustodialSites": sites_out,
"AutoApproveSubscriptionSites": list(set(sites_out)),
"AcquisitionEra": wfh.acquisitionEra(),
"ProcessingString": wfh.processingString(),
"MergedLFNBase": set_lfn,
"ProcessingVersion": version,
}
## plain assignment here
team = "production"
if os.getenv("UNIFIED_TEAM"):
team = os.getenv("UNIFIED_TEAM")
if options and options.team:
team = options.team
if False and "T2_CH_CERN" in parameters["SiteWhitelist"]:
## add some check on
### the amount pending to HLT
### the size of the request
### the priority of the request (maybe not if we decide to overflow during runs)
parameters["SiteWhitelist"] = ["T2_CH_CERN_HLT"]
team = "hlt"
## reduce the splitting by factor of 4, regardless of type of splitting
sendEmail("sending work to HLT", "%s was assigned to HLT" % wfo.name)
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v = getattr(options, key)
if v != None:
if type(v) == str and "," in v:
parameters[key] = filter(None, v.split(","))
else:
parameters[key] = v
if lheinput:
## throttle reading LHE article
wfh.sendLog("assignor", "Setting the number of events per job to 500k max")
parameters["EventsPerJob"] = 500000
## pick up campaign specific assignment parameters
parameters.update(CI.parameters(wfh.request["Campaign"]))
if not options.test:
parameters["execute"] = True
split_check = wfh.checkWorkflowSplitting()
if split_check != True:
parameters.update(split_check)
if "EventBased" in split_check.values():
wfh.sendLog("assignor", "Falling back to event splitting.")
# sendEmail("Fallback to EventBased","the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"%wfo.name)
sendLog(
"assignor",
"the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"
% wfo.name,
level="critical",
)
elif "EventsPerJob" in split_check.values():
wfh.sendLog("assignor", "Modifying the number of job per event")
# sendEmail("Modifying the job per events","the workflow %s is too heavy in number of jobs explosion"%wfo.name)
sendLog(
"assignor", "the workflow %s is too heavy in number of jobs explosion" % wfo.name, level="critical"
)
# Handle run-dependent MC
pstring = wfh.processingString()
if "PU_RD" in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi) / float(len(eventsPerLumi))
reqJobs = 500
if "PU_RD2" in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents / (reqJobs * 1.4))
lumisPerJob = int(eventsPerJob / eventsPerLumi)
if lumisPerJob == 0:
# sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog(
"assignor",
"%s needs to be split by event with %s per job" % (wfo.name, eventsPerJob),
level="critical",
)
wfh.sendLog("assignor", "%s needs to be split by event with %s per job" % (wfo.name, eventsPerJob))
parameters["EventsPerJob"] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl["events_per_job"] if "events_per_job" in spl else None
eventsPerJobEstimated = spl["avg_events_per_job"] if "avg_events_per_job" in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
# sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog(
"assignor", "%s was assigned with %s lumis/job" % (wfo.name, lumisPerJob), level="critical"
)
wfh.sendLog("assignor", "%s was assigned with %s lumis/job" % (wfo.name, lumisPerJob))
parameters["LumisPerJob"] = lumisPerJob
else:
# sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog(
"assignor",
"leaving splitting untouched for %s, please check on %s" % (pstring, wfo.name),
level="critical",
)
wfh.sendLog("assignor", "leaving splitting untouched for PU_RD*, please check.")
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = "away"
session.commit()
n_assigned += 1
wfh.sendLog("assignor", "Properly assigned\n%s" % (json.dumps(parameters, indent=2)))
try:
## refetch information and lock output
new_wfi = workflowInfo(url, wfo.name)
(_, prim, _, sec) = new_wfi.getIO()
for secure in list(prim) + list(sec) + new_wfi.request["OutputDatasets"]:
## lock all outputs flat
NLI.lock(secure)
# for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output", str(e))
else:
print "ERROR could not assign", wfo.name
else:
pass
print "Assignment summary:"
sendLog("assignor", "Assigned %d Stalled %s" % (n_assigned, n_stalled))
def assignor(url ,specific = None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
#if notRunningBefore( 'stagor' ): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = global_SI
#LI = lockInfo()
NLI = newLockInfo()
n_assigned = 0
n_stalled = 0
wfos=[]
if specific:
wfos = session.query(Workflow).filter(Workflow.name==specific).all()
if not wfos:
if specific:
wfos = session.query(Workflow).filter(Workflow.status=='considered').all()
wfos.extend( session.query(Workflow).filter(Workflow.status=='staging').all())
wfos.extend(session.query(Workflow).filter(Workflow.status=='staged').all())
for wfo in wfos:
if specific:
if not any(map(lambda sp: sp in wfo.name,specific.split(','))): continue
#if not specific in wfo.name: continue
print "\n\n",wfo.name,"\n\tto be assigned"
wfh = workflowInfo( url, wfo.name)
## check if by configuration we gave it a GO
if not CI.go( wfh.request['Campaign'] ) and not options.go:
print "No go for",wfh.request['Campaign']
n_stalled+=1
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] !='assignment-approved':
if not options.test:
print wfo.name,wfh.request['RequestStatus'],"setting away and skipping"
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name,wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version=wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
print "cannot decide on version number"
n_stalled+=1
continue
#(lheinput,primary,parent,secondary) = wfh.getIO()
#sites_allowed = getSiteWhiteList( (lheinput,primary,parent,secondary) )
(lheinput,primary,parent,secondary, sites_allowed) = wfh.getSiteWhiteList()
print "Site white list",sorted(sites_allowed)
override_sec_location = CI.get(wfh.request['Campaign'], 'SecondaryLocation', [])
c_sites_allowed = CI.get(wfh.request['Campaign'], 'SiteWhitelist' , [])
if c_sites_allowed:
print "Would like to use the new whitelist, but will not until things went through a bit"
sendEmail("using a restricted site white list","for %s"%(c_sites_allowed))
sites_allowed = list(set(sites_allowed) & set(c_sites_allowed))
c_black_list = CI.get(wfh.request['Campaign'], 'SiteBlacklist', [])
if c_black_list:
print "Reducing the whitelist due to black list in campaign configuration"
print "Removing",c_black_list
sites_allowed = list(set(sites_allowed) - set(c_black_list))
blocks = []
if 'BlockWhitelist' in wfh.request:
blocks = wfh.request['BlockWhitelist']
ncores = wfh.request.get('Multicore',1)
memory_allowed = SI.sitesByMemory( wfh.request['Memory'] , maxCore=ncores)
if memory_allowed!=None:
print "sites allowing", wfh.request['Memory'],"MB and",ncores,"core are",memory_allowed
sites_allowed = list(set(sites_allowed) & set(memory_allowed))
print "Allowed",sorted(sites_allowed)
secondary_locations=None
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
sendEmail("tempting to pass sec location check","but we cannot yet IMO")
#pass
presence = getDatasetPresence( url, sec )
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations==None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [site for site in sites_allowed if SI.CE_to_SE(site) in one_secondary_locations]
print "From secondary requirement, now Allowed",sorted(sites_allowed)
initial_sites_allowed = copy.deepcopy( sites_allowed ) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy( sites_allowed )
sites_with_data = copy.deepcopy( sites_allowed )
sites_with_any_data = copy.deepcopy( sites_allowed )
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
for prim in list(primary):
set_lfn = getLFNbase( prim )
presence = getDatasetPresence( url, prim , only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(url, prim, sites = [SI.CE_to_SE(site) for site in sites_allowed] , only_blocks = blocks)
#sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
#sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,(there,frac)) in presence.items() if there]]
sites_with_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,frac) in presence.items() if frac[1]>90.]]
sites_with_any_data = [site for site in sites_with_any_data if SI.CE_to_SE(site) in presence.keys()]
print "Holding the data but not allowed",list(set([se_site for se_site in presence.keys() if not SI.SE_to_CE(se_site) in sites_allowed]))
if primary_locations==None:
primary_locations = presence.keys()
else:
primary_locations = list(set(primary_locations) & set(presence.keys() ))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites=[]
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations)) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
elif primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list(set(primary_locations) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
else:
opportunistic_sites = []
print "We could be running at",sorted(opportunistic_sites),"in addition"
if any([osite in SI.sites_not_ready for osite in opportunistic_sites]):
print "One of the destination site is in downtime"
down_time = True
## should this be send back to considered ?
"""
if available_fractions and not all([available>=1. for available in available_fractions.values()]):
print "The input dataset is not located in full over sites"
print json.dumps(available_fractions)
if not options.test and not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known:
sendEmail( "cannot be assigned","%s is not full over sites \n %s"%(wfo.name,json.dumps(available_fractions)))
known.append( wfo.name )
open('cannot_assign.json','w').write(json.dumps( known, indent=2))
n_stalled+=1
continue ## skip skip skip
"""
## should be 2 but for the time-being let's lower it to get things going
copies_wanted,cpuh = wfh.getNCopies()
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(1,copies_wanted-less_copies_than_requested) # take one out for the efficiency
if available_fractions and not all([available>=copies_wanted for available in available_fractions.values()]):
not_even_once = not all([available>=1. for available in available_fractions.values()])
print "The input dataset is not available",copies_wanted,"times, only",available_fractions.values()
if down_time and not options.go:
wfo.status = 'considered'
session.commit()
print "sending back to considered because of site downtime, instead of waiting"
sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known:
sendEmail( "cannot be assigned","%s is not sufficiently available. Probably phedex information lagging behind. \n %s"%(wfo.name,json.dumps(available_fractions)))
known.append( wfo.name )
open('cannot_assign.json','w').write(json.dumps( known, indent=2))
n_stalled+=1
continue
## default back to white list to original white list with any data
print "Allowed",sites_allowed
if options.primary_aaa:
sites_allowed = initial_sites_allowed
options.useSiteListAsLocation = True
else:
sites_allowed = sites_with_any_data
print "Selected for any data",sites_allowed
if options.restrict:
print "Allowed",sites_allowed
sites_allowed = sites_with_any_data
print "Selected",sites_allowed
else:
if set(sites_with_data) != set(sites_allowed):
## the data is not everywhere we wanted to run at : enable aaa
print "Sites with 90% data not matching site white list (block choping!)"
print "Resorting to AAA reading for",list(set(sites_allowed) - set(sites_with_data)),"?"
print "Whitelist site with any data",list(set(sites_allowed) - set(sites_with_any_data))
#options.useSiteListAsLocation = True
#print "Not commissioned yet"
#continue
#print "We could be running at",opportunistic_sites,"in addition"
##sites_allowed = list(set(sites_allowed+ opportunistic_sites))
if not len(sites_allowed):
print wfo.name,"cannot be assign with no matched sites"
sendEmail( "cannot be assigned","%s has no whitelist"%(wfo.name))
n_stalled+=1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])]
print "Placing the output on", sites_out
parameters={
'SiteWhitelist' : sites_allowed,
#'CustodialSites' : sites_custodial,
'NonCustodialSites' : sites_out,
'AutoApproveSubscriptionSites' : list(set(sites_out)),
'AcquisitionEra' : wfh.acquisitionEra(),
'ProcessingString' : wfh.processingString(),
'MergedLFNBase' : set_lfn,
'ProcessingVersion' : version,
}
## plain assignment here
team='production'
if options and options.team:
team = options.team
#if wfh.request['RequestPriority'] >= 100000 and (wfh.request['TimePerEvent']*int(wfh.getRequestNumEvents()))/(8*3600.) < 10000:
# team = 'highprio'
# sendEmail("sending work with highprio team","%s"% wfo.name, destination=['*****@*****.**'])
if "T2_US_UCSD" in sites_with_data and random.random() < -0.5 and wfh.request['Campaign']=='RunIISpring15DR74' and int(wfh.getRequestNumEvents()) < 600000 and not any([out.endswith('RAW') for out in wfh.request['OutputDatasets']]):
## consider SDSC
parameters['SiteWhitelist'] = ['T2_US_UCSD','T3_US_SDSC']
parameters['useSiteListAsLocation'] = True
team = 'allocation-based'
sendEmail("sending work to SDSC","%s was assigned to SDSC/UCSD"% wfo.name, destination=['*****@*****.**'])
if wfh.request['Campaign']=='RunIIWinter15GS' and random.random() < -1.0:
parameters['SiteWhitelist'] = ['T3_US_SDSC']
team = 'allocation-based'
sendEmail("sending work to SDSC","%s was assigned to SDSC"% wfo.name, destination=['*****@*****.**'])
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v=getattr(options,key)
if v!=None:
if type(v)==str and ',' in v:
parameters[key] = filter(None,v.split(','))
else:
parameters[key] = v
## pick up campaign specific assignment parameters
parameters.update( CI.parameters(wfh.request['Campaign']) )
if not options.test:
parameters['execute'] = True
split_check = wfh.checkWorkflowSplitting()
if split_check!=True:
parameters.update( split_check )
if 'EventBased' in split_check.values():
print "Falling back to event splitting."
sendEmail("Fallback to EventBased","the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"%wfo.name)
elif 'EventsPerJob' in split_check.values():
print "Modifying the number of job per event"
sendEmail("Modifying the job per events","the workflow %s is too heavy in number of jobs explosion"%wfo.name)
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi)/float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents/(reqJobs*1.4))
lumisPerJob = int(eventsPerJob/eventsPerLumi)
if lumisPerJob==0:
print "There is no go for assigning that request without event splitting"
sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
print "need to go down to",eventsPerJob,"events per job"
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl['events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl['avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
print "need to go down to",lumisPerJob,"in assignment"
sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
print "the regular splitting should work for",pstring
sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned+=1
try:
## refetch information and lock output
new_wfi = workflowInfo( url, wfo.name)
(_,prim,_,sec) = new_wfi.getIO()
for secure in list(prim)+list(sec)+new_wfi.request['OutputDatasets']:
## lock all outputs flat
NLI.lock( secure )
#for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output",str(e))
else:
print "ERROR could not assign",wfo.name
else:
pass
print "Assignment summary:"
print "Assigned",n_assigned
print "Stalled",n_stalled
def assignor(url ,specific = None, talk=True, options=None):
if userLock(): return
if duplicateLock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
SI = global_SI
#LI = lockInfo()
NLI = newLockInfo()
n_assigned = 0
n_stalled = 0
wfos=[]
if specific or options.early:
wfos.extend( session.query(Workflow).filter(Workflow.status=='considered').all())
wfos.extend( session.query(Workflow).filter(Workflow.status=='staging').all())
if specific:
wfos.extend( session.query(Workflow).filter(Workflow.status=='considered-tried').all())
wfos.extend(session.query(Workflow).filter(Workflow.status=='staged').all())
dataset_endpoints = json.loads(open('%s/dataset_endpoints.json'%monitor_dir).read())
max_per_round = UC.get('max_per_round').get('assignor',None)
max_cpuh_block = UC.get('max_cpuh_block')
random.shuffle( wfos )
for wfo in wfos:
if options.limit and (n_stalled+n_assigned)>options.limit:
break
if max_per_round and (n_stalled+n_assigned)>max_per_round:
break
if specific:
if not any(map(lambda sp: sp in wfo.name,specific.split(','))): continue
#if not specific in wfo.name: continue
print "\n\n"
wfh = workflowInfo( url, wfo.name)
wfh.sendLog('assignor',"%s to be assigned"%wfo.name)
## the site whitelist takes into account siteInfo, campaignInfo, memory and cores
(lheinput,primary,parent,secondary, sites_allowed) = wfh.getSiteWhiteList()
## check if by configuration we gave it a GO
no_go = False
allowed_secondary = set()
for campaign in wfh.getCampaigns():
if not CI.go( campaign ):
wfh.sendLog('assignor',"No go for %s"%campaign)
if not options.go:
n_stalled+=1
no_go = True
break
if campaign in CI.campaigns and 'secondaries' in CI.campaigns[campaign]:
allowed_secondary.update( CI.campaigns[campaign]['secondaries'] )
if (secondary and allowed_secondary) and (set(secondary)&allowed_secondary!=set(secondary)):
wfh.sendLog('assignor','%s is not an allowed secondary'%(', '.join(set(secondary)-allowed_secondary)))
#sendEmail('secondary not allowed','%s is not an allowed secondary'%( ', '.join(set(secondary)-allowed_secondary)))
sendLog('assignor','%s is not an allowed secondary'%(', '.join(set(secondary)-allowed_secondary)), level='critical')
if not options.go:
n_stalled+=1
no_go = True
if no_go:
continue
## check on current status for by-passed assignment
if wfh.request['RequestStatus'] !='assignment-approved':
if not options.test:
wfh.sendLog('assignor',"setting %s away and skipping"%wfo.name)
## the module picking up from away will do what is necessary of it
wfo.wm_status = wfh.request['RequestStatus']
wfo.status = 'away'
session.commit()
continue
else:
print wfo.name,wfh.request['RequestStatus']
## retrieve from the schema, dbs and reqMgr what should be the next version
version=wfh.getNextVersion()
if not version:
if options and options.ProcessingVersion:
version = options.ProcessingVersion
else:
wfh.sendLog('assignor',"cannot decide on version number")
n_stalled+=1
wfo.status = 'trouble'
session.commit()
continue
original_sites_allowed = copy.deepcopy( sites_allowed )
wfh.sendLog('assignor',"Site white list %s"%sorted(sites_allowed))
override_sec_location = CI.get(wfh.request['Campaign'], 'SecondaryLocation', [])
blocks = []
if 'BlockWhitelist' in wfh.request:
blocks = wfh.request['BlockWhitelist']
if 'RunWhitelist' in wfh.request and wfh.request['RunWhitelist']:
## augment with run white list
for dataset in primary:
blocks = list(set( blocks + getDatasetBlocks( dataset, runs=wfh.request['RunWhitelist'] ) ))
wfh.sendLog('assignor',"Allowed %s"%sorted(sites_allowed))
secondary_locations=None
for sec in list(secondary):
if override_sec_location:
print "We don't care where the secondary is"
print "Cannot pass for now"
sendEmail("tempting to pass sec location check","but we cannot yet IMO")
#pass
presence = getDatasetPresence( url, sec )
print sec
print json.dumps(presence, indent=2)
one_secondary_locations = [site for (site,(there,frac)) in presence.items() if frac>98.]
#one_secondary_locations = [site for (site,(there,frac)) in presence.items() if there]
if secondary_locations==None:
secondary_locations = one_secondary_locations
else:
secondary_locations = list(set(secondary_locations) & set(one_secondary_locations))
## reduce the site white list to site with secondary only
#sites_allowed = [site for site in sites_allowed if any([osite.startswith(site) for osite in one_secondary_locations])]
sites_allowed = [site for site in sites_allowed if SI.CE_to_SE(site) in one_secondary_locations]
wfh.sendLog('assignor',"From secondary requirement, now Allowed%s"%sorted(sites_allowed))
initial_sites_allowed = copy.deepcopy( sites_allowed ) ## keep track of this, after secondary input location restriction : that's how you want to operate it
sites_all_data = copy.deepcopy( sites_allowed )
sites_with_data = copy.deepcopy( sites_allowed )
sites_with_any_data = copy.deepcopy( sites_allowed )
primary_locations = None
available_fractions = {}
set_lfn = '/store/mc' ## by default
endpoints = set()
for prim in list(primary):
if prim in dataset_endpoints:
print "endpoints from stagor",dataset_endpoints[prim]
endpoints.update( dataset_endpoints[prim] )
set_lfn = getLFNbase( prim )
presence = getDatasetPresence( url, prim , only_blocks=blocks)
if talk:
print prim
print json.dumps(presence, indent=2)
available_fractions[prim] = getDatasetBlocksFraction(url, prim, sites = [SI.CE_to_SE(site) for site in sites_allowed] , only_blocks = blocks)
#sites_all_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,(there,frac)) in presence.items() if there]])]
#sites_with_data = [site for site in sites_with_data if any([osite.startswith(site) for osite in [psite for (psite,frac) in presence.items() if frac[1]>90.]])]
sites_all_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,(there,frac)) in presence.items() if there]]
sites_with_data = [site for site in sites_with_data if SI.CE_to_SE(site) in [psite for (psite,frac) in presence.items() if frac[1]>90.]]
sites_with_any_data = [site for site in sites_with_any_data if SI.CE_to_SE(site) in presence.keys()]
wfh.sendLog('assignor',"Holding the data but not allowed %s"%sorted(list(set([se_site for se_site in presence.keys() if not SI.SE_to_CE(se_site) in sites_allowed]))))
if primary_locations==None:
primary_locations = presence.keys()
else:
primary_locations = list(set(primary_locations) & set(presence.keys() ))
sites_with_data = list(set(sites_with_data))
sites_with_any_data = list(set(sites_with_any_data))
opportunistic_sites=[]
down_time = False
## opportunistic running where any piece of data is available
if secondary_locations or primary_locations:
## intersection of both any pieces of the primary and good IO
#opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations) & set(SI.sites_with_goodIO)) - set(sites_allowed))]
if secondary_locations and primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list((set(secondary_locations) & set(primary_locations)) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
elif primary_locations:
opportunistic_sites = [SI.SE_to_CE(site) for site in list(set(primary_locations) - set([SI.CE_to_SE(site) for site in sites_allowed]))]
else:
opportunistic_sites = []
wfh.sendLog('assignor',"We could be running in addition at %s"% sorted(opportunistic_sites))
if any([osite in SI.sites_not_ready for osite in opportunistic_sites]):
wfh.sendLog('assignor',"One of the usable site is in downtime %s"%([osite in SI.sites_not_ready for osite in opportunistic_sites]))
down_time = True
## should this be send back to considered ?
## should be 2 but for the time-being let's lower it to get things going
copies_wanted,cpuh = wfh.getNCopies()
wfh.sendLog('assignor',"we need %s CPUh"%cpuh)
if cpuh>max_cpuh_block and not options.go:
#sendEmail('large workflow','that wf %s has a large number of CPUh %s, not assigning, please check the logs'%(wfo.name, cpuh))#,destination=['*****@*****.**'])
sendLog('assignor','%s requires a large numbr of CPUh %s , not assigning, please check with requester'%( wfo.name, cpuh), level='critical')
wfh.sendLog('assignor',"Requiring a large number of CPUh %s, not assigning"%cpuh)
continue
if 'Campaign' in wfh.request and wfh.request['Campaign'] in CI.campaigns and 'maxcopies' in CI.campaigns[wfh.request['Campaign']]:
copies_needed_from_campaign = CI.campaigns[wfh.request['Campaign']]['maxcopies']
copies_wanted = min(copies_needed_from_campaign, copies_wanted)
if not options.early:
less_copies_than_requested = UC.get("less_copies_than_requested")
copies_wanted = max(1,copies_wanted-less_copies_than_requested) # take one out for the efficiency
else:
## find out whether there is a site in the whitelist, that is lacking jobs and reduce to 1 copy needed to get things going
pass
wfh.sendLog('assignor',"needed availability fraction %s"% copies_wanted)
## should also check on number of sources, if large enough, we should be able to overflow most, efficiently
if available_fractions and not all([available>=copies_wanted for available in available_fractions.values()]):
not_even_once = not all([available>=1. for available in available_fractions.values()])
wfh.sendLog('assignor',"The input dataset is not available %s times, only %s"%( copies_wanted, available_fractions.values()))
if down_time and not options.go and not options.early:
wfo.status = 'considered'
session.commit()
wfh.sendLog('assignor',"sending back to considered because of site downtime, instead of waiting")
#sendEmail( "cannot be assigned due to downtime","%s is not sufficiently available, due to down time of a site in the whitelist. check the assignor logs. sending back to considered."% wfo.name)
sendLog('assignor','%s is not sufficiently available, due to down time of a site in the whitelist. sending back to considered.'%( wfo.name ), level='delay')
continue
#pass
print json.dumps(available_fractions)
if (options.go and not_even_once) or not options.go:
known = []
try:
known = json.loads(open('cannot_assign.json').read())
except:
pass
if not wfo.name in known and not options.limit and not options.go and not options.early and not options.partial:
wfh.sendLog('assignor',"cannot be assigned, %s is not sufficiently available.\n %s"%(wfo.name,json.dumps(available_fractions)))
sendEmail( "cannot be assigned","%s is not sufficiently available.\n %s"%(wfo.name,json.dumps(available_fractions)))
known.append( wfo.name )
open('cannot_assign.json','w').write(json.dumps( known, indent=2))
n_stalled+=1
if options.early:
if wfo.status == 'considered':
wfh.sendLog('assignor',"setting considered-tried")
wfo.status = 'considered-tried'
session.commit()
else:
print "tried but status is",wfo.status
if options.partial:
print "Will move on with partial locations"
else:
continue
## default back to white list to original white list with any data
print "Allowed",sorted(sites_allowed)
if options.primary_aaa:
sites_allowed = initial_sites_allowed
#options.useSiteListAsLocation = True
options.TrustSitelists = True
else:
sites_allowed = sites_with_any_data
wfh.sendLog('assignor',"Selected for any data %s"%sorted(sites_allowed))
if options.restrict:
print "Allowed",sites_allowed
sites_allowed = sites_with_any_data
print "Selected",sites_allowed
else:
if set(sites_with_data) != set(sites_allowed):
## the data is not everywhere we wanted to run at : enable aaa
print "Sites with 90% data not matching site white list (block choping!)"
print "Resorting to AAA reading for",list(set(sites_allowed) - set(sites_with_data)),"?"
print "Whitelist site with any data",list(set(sites_allowed) - set(sites_with_any_data))
#options.useSiteListAsLocation = True
#print "Not commissioned yet"
#continue
#print "We could be running at",opportunistic_sites,"in addition"
##sites_allowed = list(set(sites_allowed+ opportunistic_sites))
### check on endpoints for on-going transfers
if endpoints and options.partial:
sites_allowed = list(set(sites_allowed + [SI.SE_to_CE(s) for s in endpoints]))
print "with added endpoints",sorted(sites_allowed)
#if options.partial:
# continue
if not len(sites_allowed):
wfh.sendLog('assignor',"cannot be assign with no matched sites")
#sendEmail( "cannot be assigned","%s has no whitelist"%(wfo.name))
sendLog('assignor','%s has no whitelist'% wfo.name, level='critical')
n_stalled+=1
continue
t1_only = [ce for ce in sites_allowed if ce.startswith('T1')]
if t1_only:
# try to pick from T1 only first
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in t1_only])]
else:
# then pick any otherwise
sites_out = [SI.pick_dSE([SI.CE_to_SE(ce) for ce in sites_allowed])]
## one last modification now that we know we can assign, and to make sure all ressource can be used by the request : set all ON sites to whitelist
###sites_allowed = original_sites_allowed ## not needed, afterall as secondary jobs go their own ways
wfh.sendLog('assignor',"Placing the output on %s"%sites_out)
parameters={
'SiteWhitelist' : sites_allowed,
#'CustodialSites' : sites_custodial,
'NonCustodialSites' : sites_out,
'AutoApproveSubscriptionSites' : list(set(sites_out)),
'AcquisitionEra' : wfh.acquisitionEra(),
'ProcessingString' : wfh.processingString(),
'MergedLFNBase' : set_lfn,
'ProcessingVersion' : version,
}
## plain assignment here
team='production'
if os.getenv('UNIFIED_TEAM'): team = os.getenv('UNIFIED_TEAM')
if options and options.team:
team = options.team
## high priority team agent
#if wfh.request['RequestPriority'] >= 100000 and (wfh.request['TimePerEvent']*int(wfh.getRequestNumEvents()))/(8*3600.) < 10000:
# team = 'highprio'
# sendEmail("sending work with highprio team","%s"% wfo.name, destination=['*****@*****.**'])
## SDSC redirection
#if "T2_US_UCSD" in sites_with_data and random.random() < -0.5 and wfh.request['Campaign']=='RunIISpring15DR74' and int(wfh.getRequestNumEvents()) < 600000 and not any([out.endswith('RAW') for out in wfh.request['OutputDatasets']]):
# ## consider SDSC
# parameters['SiteWhitelist'] = ['T2_US_UCSD','T3_US_SDSC']
# parameters['useSiteListAsLocation'] = True
# team = 'allocation-based'
# sendEmail("sending work to SDSC","%s was assigned to SDSC/UCSD"% wfo.name, destination=['*****@*****.**'])
## SDSC redirection
#if wfh.request['Campaign']==R'unIIWinter15GS' and random.random() < -1.0:
# parameters['SiteWhitelist'] = ['T3_US_SDSC']
# team = 'allocation-based'
# sendEmail("sending work to SDSC","%s was assigned to SDSC"% wfo.name, destination=['*****@*****.**'])
if False and 'T2_CH_CERN' in parameters['SiteWhitelist']:
## add some check on
### the amount pending to HLT
### the size of the request
### the priority of the request (maybe not if we decide to overflow during runs)
parameters['SiteWhitelist'] = ['T2_CH_CERN_HLT']
team = 'hlt'
## reduce the splitting by factor of 4, regardless of type of splitting
sendEmail("sending work to HLT","%s was assigned to HLT"%wfo.name)
##parse options entered in command line if any
if options:
for key in reqMgrClient.assignWorkflow.keys:
v=getattr(options,key)
if v!=None:
if type(v)==str and ',' in v:
parameters[key] = filter(None,v.split(','))
else:
parameters[key] = v
## pick up campaign specific assignment parameters
parameters.update( CI.parameters(wfh.request['Campaign']) )
if not options.test:
parameters['execute'] = True
split_check = wfh.checkWorkflowSplitting()
if split_check!=True:
parameters.update( split_check )
if 'EventBased' in split_check.values():
wfh.sendLog('assignor', "Falling back to event splitting.")
#sendEmail("Fallback to EventBased","the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting"%wfo.name)
sendLog('assignor','the workflow %s is too heavy to be processed as it is. Fallback to EventBased splitting'%wfo.name, level='critical')
elif 'EventsPerJob' in split_check.values():
wfh.sendLog('assignor', "Modifying the number of job per event")
#sendEmail("Modifying the job per events","the workflow %s is too heavy in number of jobs explosion"%wfo.name)
sendLog('assignor',"the workflow %s is too heavy in number of jobs explosion"%wfo.name, level='critical')
# Handle run-dependent MC
pstring = wfh.processingString()
if 'PU_RD' in pstring:
numEvents = wfh.getRequestNumEvents()
eventsPerLumi = [getDatasetEventsPerLumi(prim) for prim in primary]
eventsPerLumi = sum(eventsPerLumi)/float(len(eventsPerLumi))
reqJobs = 500
if 'PU_RD2' in pstring:
reqJobs = 2000
eventsPerJob = int(numEvents/(reqJobs*1.4))
lumisPerJob = int(eventsPerJob/eventsPerLumi)
if lumisPerJob==0:
#sendEmail("issue with event splitting for run-dependent MC","%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
sendLog('assignor', "%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob), level='critical')
wfh.sendLog('assignor', "%s needs to be split by event with %s per job"%(wfo.name, eventsPerJob))
parameters['EventsPerJob'] = eventsPerJob
else:
spl = wfh.getSplittings()[0]
eventsPerJobEstimated = spl['events_per_job'] if 'events_per_job' in spl else None
eventsPerJobEstimated = spl['avg_events_per_job'] if 'avg_events_per_job' in spl else None
if eventsPerJobEstimated and eventsPerJobEstimated > eventsPerJob:
#sendEmail("setting lumi splitting for run-dependent MC","%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
sendLog('assignor',"%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob), level='critical')
wfh.sendLog('assignor',"%s was assigned with %s lumis/job"%( wfo.name, lumisPerJob))
parameters['LumisPerJob'] = lumisPerJob
else:
#sendEmail("leaving splitting untouched for PU_RD*","please check on "+wfo.name)
sendLog('assignor',"leaving splitting untouched for %s, please check on %s"%( pstring, wfo.name), level='critical')
wfh.sendLog('assignor',"leaving splitting untouched for PU_RD*, please check.")
result = reqMgrClient.assignWorkflow(url, wfo.name, team, parameters)
# set status
if not options.test:
if result:
wfo.status = 'away'
session.commit()
n_assigned+=1
try:
## refetch information and lock output
new_wfi = workflowInfo( url, wfo.name)
(_,prim,_,sec) = new_wfi.getIO()
for secure in list(prim)+list(sec)+new_wfi.request['OutputDatasets']:
## lock all outputs flat
NLI.lock( secure )
#for site in [SI.CE_to_SE(site) for site in sites_allowed]:
# for output in new_wfi.request['OutputDatasets']:
# LI.lock( output, site, 'dataset in production')
# for primary in prim:
# LI.lock( primary, site, 'dataset used in input')
# for secondary in sec:
# LI.lock( secondary, site, 'required for mixing' )
except Exception as e:
print "fail in locking output"
print str(e)
sendEmail("failed locking of output",str(e))
else:
print "ERROR could not assign",wfo.name
else:
pass
print "Assignment summary:"
sendLog('assignor',"Assigned %d Stalled %s"%(n_assigned, n_stalled))
def actor(url, options=None):
mlock = moduleLock(wait=False, silent=True)
if mlock(): return
if userLock('actor'): return
up = componentInfo(soft=['mcm'])
if not up.check(): return
# CI = campaignInfo()
SI = siteInfo()
UC = unifiedConfiguration()
WC = wtcClient()
WI = wtcInfo()
JC = JIRAClient()
action_list = WC.get_actions()
if action_list is None:
print "Not able to load action list"
sendLog('actor', 'Not able to load action list', level='critical')
return
if options.actions:
action_list = json.loads(open(options.actions).read())
print json.dumps(action_list, indent=2)
if not action_list:
print "EMPTY!"
return
wf_list = action_list.keys()
print json.dumps(sorted(wf_list), indent=2)
if options.spec:
wf_list = [wf for wf in wf_list if options.spec in wf]
max_per_round = UC.get('max_per_round').get('actor', None)
if max_per_round:
random.shuffle(wf_list)
wf_list = wf_list[:max_per_round]
for wfname in wf_list:
print '-' * 100
print "Looking at", wfname, "for recovery options"
to_clone = False
to_acdc = False
to_force = False
to_hold = False
something_to_do = False
tasks = action_list[wfname].get('Parameters', None)
to_acdc = action_list[wfname].get('Action', None) == 'acdc'
to_clone = action_list[wfname].get('Action', None) == 'clone'
to_force = action_list[wfname].get(
'Action', None) == 'special' and action_list[wfname].get(
'Parameters', {}).get('action', None) in ['by-pass', 'bypass']
to_hold = action_list[wfname].get(
'Action', None) == 'special' and action_list[wfname].get(
'Parameters', {}).get('action', None) in ['onhold', 'on-hold']
if not to_acdc and not to_clone and not to_force and not to_hold:
sendLog(
'actor',
'Action submitted for something other than acdc, clone, bypass or hold for workflow %s'
% wfname,
level='critical')
print json.dumps(action_list[wfname], indent=2)
continue
if not tasks and to_acdc:
sendLog('actor',
'Empty action submitted for workflow %s' % wfname,
level='critical')
print "Moving on. Parameters is blank for " + wfname
continue
wfi = workflowInfo(url, wfname)
recover = True
message_to_ops = ""
message_to_user = ""
#===========================================================
if to_clone and options.do:
print "Let's try kill and clone: "
wfi.sendLog('actor', 'Going to clone %s' % wfname)
comment = ""
if 'comment' in tasks: comment = ", reason: " + tasks['comment']
wfi.sendLog(
'actor',
"invalidating the workflow by traffic controller %s" % comment)
#Reject all workflows in the family
inv_results = invalidate(url,
wfi,
only_resub=False,
with_output=True)
all_good = all(inv_results)
if all_good:
wfi.sendLog('actor', "%s and children are rejected" % wfname)
else:
wfi.sendLog('actor',
"Failed to reject the request and dependents")
sendLog('actor',
'Failed to reject the familly of %s' % wfname,
level='critical')
continue
cloned = None
try:
cloned = singleClone(url, wfname, tasks, comment, options.do)
except Exception as e:
sendLog(
'actor',
'Failed to create clone for %s! Check logs for more information. Action will need to be resubmitted.'
% wfname,
level='critical')
wfi.sendLog('actor', 'Failed to create clone for %s!' % wfname)
print str(e)
##let's not remove the action other the workflow goes to "trouble" and the WTC cannot set the action again
if not cloned:
recover = False
wfi.sendLog('actor', 'Failed to create clone for %s!' % wfname)
sendLog('actor',
'Failed to create clone for %s!' % wfname,
level='critical')
else:
wfi.sendLog('actor',
"Workflow %s cloned into %s" % (wfname, cloned))
## set to trouble for swift replacement
for wfo in session.query(Workflow).filter(
Workflow.name == wfname).all():
wfo.status = 'trouble'
session.commit()
#===========================================================
elif to_force:
wfi.sendLog(
'actor',
'Force-completing from workflow traffic controler request')
WI.add(action='force',
keyword=wfname,
user=action_list[wfname].get('user', 'unified'))
elif to_hold:
wfi.sendLog('actor',
'Holding on workflow traffic controler request')
WI.add(action='hold',
keyword=wfname,
user=action_list[wfname].get('user', 'unified'))
#===========================================================
elif to_acdc:
if 'AllSteps' in tasks:
allTasksDefaults = tasks['AllSteps']
tasks.pop('AllSteps')
for setting in allTasksDefaults:
for task in tasks:
if setting in tasks[task]:
tasks[task][setting] = allTasksDefaults[setting]
else:
tasks[task].append(
{setting: allTasksDefaults[setting]})
print "Tasks is "
print json.dumps(tasks, indent=2)
all_tasks = wfi.getAllTasks()
## need a way to verify that this is the first round of ACDC, since the second round will have to be on the ACDC themselves
try:
WMErr = wfi.getWMErrors()
# print WMErr
except:
sendLog(
'actor',
'Cannot create ACDCS for %s because WMErr cannot be reached.'
% wfname,
level='critical')
continue
if not WMErr:
wfi.sendLog('actor', 'WMErrors is blank for %s.' % wfname)
print "FYI getWMErrors is blank. Presumably there are only unreported errors"
# continue
try:
where_to_run, missing_to_run, missing_to_run_at = wfi.getRecoveryInfo(
)
print "Where to run = "
print where_to_run
if not where_to_run:
sendLog(
'actor',
'Cannot create ACDCS for %s because recovery info cannot be found.'
% wfname,
level='critical')
continue
except:
sendLog(
'actor',
'Cannot create ACDCS for %s because recovery info cannot be found.'
% wfname,
level='critical')
print "Moving on. Cannot access recovery info for " + wfname
continue
if not where_to_run:
sendLog(
'actor',
'Cannot create ACDCS for %s because site list cannot be found.'
% wfname,
level='critical')
print "Moving on. where to run is blank"
continue
message_to_ops = ""
message_to_user = ""
num_tasks_to_recover = 0
if WMErr:
for task in WMErr:
if 'LogCollect' in task: continue
if 'Cleanup' in task: continue
if not 'jobfailed' in WMErr[task]:
continue
else:
num_tasks_to_recover += 1
# print "Task to recover: " + task
if not num_tasks_to_recover:
print "\tno error for", wfname
# recover = False
if 'LheInputFiles' in wfi.request and wfi.request['LheInputFiles']:
## we do not try to recover pLHE
sendLog(
'actor',
'Cannot create ACDCS for %s because it is a pLHE workflow.'
% wfname,
level='critical')
print "We don't try to recover pLHE. Moving on."
recover = False
# sendEmail('cannot submit action', '%s is a pLHE workflow. We do not try to recover pLHE'%wfname)
# if wfi.request['RequestType'] in ['ReReco']:
# recover= False
# print 'cannot submit action. ReReco'
# sendEmail('cannot submit action', '%s is request type ReReco'%wfname)
recovering = set()
for task in tasks:
assign_to_sites = set()
print "Task names is " + task
fulltaskname = '/' + wfname + '/' + task
print "Full task name is " + fulltaskname
print where_to_run.keys()
wrong_task = False
for task_info in all_tasks:
if fulltaskname == task_info.pathName:
if task_info.taskType not in [
'Processing', 'Production', 'Merge'
]:
wrong_task = True
wfi.sendLog(
'actor',
"Skipping task %s because the taskType is %s. Can only ACDC Processing, Production, or Merge tasks"
% (fulltaskname, task_info.taskType))
if not fulltaskname in where_to_run.keys():
wrong_task = True
wfi.sendLog(
'actor',
"Skipping task %s because there is no acdc doc for it anyways."
% (fulltaskname))
if wrong_task:
continue
print tasks[task]
actions = tasks[task]
for action in actions:
if action.startswith('sites'):
if type(actions[action]) != list:
assign_to_sites = [SI.SE_to_CE(actions[action])]
else:
assign_to_sites = list(
set([
SI.SE_to_CE(site)
for site in actions[action]
]))
# if action.startswith('mem') and actions[action] != "" and actions[action] != 'Same' and wfi.request['RequestType'] in ['TaskChain']:
# recover = False;
# print "Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname
# wfi.sendLog('actor',"Skipping %s for now until Allie fixes memory parameter for TaskChain ACDCs."%wfname)
if not 'sites' in actions:
assign_to_sites = list(
set([
SI.SE_to_CE(site)
for site in where_to_run[fulltaskname]
]))
print "Found", sorted(
assign_to_sites
), "as sites where to run the ACDC at, from the acdc doc of ", wfname
print "Going to run at", sorted(assign_to_sites)
if recover:
print "Initiating recovery"
acdc = singleRecovery(url,
fulltaskname,
wfi.request,
actions,
do=options.do)
if not acdc:
if options.do:
if recovering:
print wfname + " has been partially ACDC'ed. Needs manual attention."
sendLog(
'actor',
"%s has had %s/%s recoveries %s only" %
(wfname, len(recovering),
num_tasks_to_recover, list(recovering)),
level='critical')
wfi.sendLog(
'actor',
"%s has had %s/%s recoveries %s only" %
(wfname, len(recovering),
num_tasks_to_recover, list(recovering)))
break
else:
print wfname + " failed recovery once"
recover = False
break
else:
print "no action to take further"
# sendLog('recoveror', "ACDC for %s can be done automatically"% wfname, level='critical')
continue
else: #ACDC was made correctly. Now we have to assign it.
wfi.sendLog(
'actor',
'ACDC created for task %s. Actions taken \n%s' %
(fulltaskname, json.dumps(actions)))
jira_comment = "%s created ACDC for task %s with action %s" % (
action_list[wfname].get('user', 'unified'),
task.split('/')[-1],
json.dumps(actions),
)
reason = action_list[wfname].get('Reason', None)
if reason:
jira_comment += '\ndue to: %s' % (reason)
#team = wfi.request['Teams'][0]
team = 'production'
parameters = {
'SiteWhitelist': sorted(assign_to_sites),
'AcquisitionEra': wfi.acquisitionEra(),
'ProcessingString': wfi.processingString(),
'MergedLFNBase': wfi.request['MergedLFNBase'],
'ProcessingVersion':
wfi.request['ProcessingVersion'],
}
## hackery for ACDC merge assignment
if wfi.request[
'RequestType'] == 'TaskChain' and 'Merge' in task.split(
'/')[-1]:
parameters['AcquisitionEra'] = None
parameters['ProcessingString'] = None
## xrootd setttings on primary and secondary
if 'xrootd' in actions:
if actions['xrootd'] == 'enabled':
print "Going to assign via xrootd"
parameters['TrustSitelists'] = True
elif actions['xrootd'] == 'disabled':
parameters['TrustSitelists'] = False
elif ('TrustSitelists' in wfi.request
and wfi.request['TrustSitelists'] == 'true'):
parameters['TrustSitelists'] = True
else:
parameters['TrustSitelists'] = False
if 'secondary' in actions:
if actions['secondary'] == 'enabled':
print 'Enabling reading the secondary input via xrootd'
parameters['TrustPUSitelists'] = True
elif actions['secondary'] == 'disabled':
parameters['TrustPUSitelists'] = False
#in case secondary is blank or not set to enabled or disabled
elif 'TrustPUSitelists' in wfi.request and wfi.request[
'TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
elif 'TrustPUSitelists' in wfi.request and wfi.request[
'TrustPUSitelists']:
parameters['TrustPUSitelists'] = True
if options.ass:
print "really doing the assignment of the ACDC", acdc
parameters['execute'] = True
#wfi.sendLog('actor',"%s was assigned for recovery"% acdc)
else:
print "no assignment done with this ACDC", acdc
sendLog('actor',
"%s needs to be assigned" % (acdc),
level='critical')
wfi.sendLog(
'actor',
"%s needs to be assigned by hand" % (acdc))
continue
# print parameters
result = reqMgrClient.assignWorkflow(
url, acdc, team, parameters)
if not result:
print acdc, "was not assigned"
sendLog('actor',
"%s failed to be assigned" % (acdc),
level='critical')
wfi.sendLog(
'actor',
"%s failed to get assigned for recovery" %
acdc)
else:
wfi.sendLog('actor',
"%s was assigned for recovery" % acdc)
recovering.add(acdc)
#wfi.sendLog('actor',"ACDCs created for %s"%wfname)
try:
if jira_comment:
jiras = JC.find(
{'prepid': wfi.request['PrepID']})
if len(jiras) == 1:
## put a comment on the single corresponding ticket
JC.comment(jiras[0].key, jira_comment)
JC.progress(jiras[0].key)
except Exception as e:
print "failed with JIRA"
print str(e)
#===========================================================
if recover and options.do:
r = WC.remove_action(wfname)
if not r:
sendLog(
'actor',
'not able to remove the action, interlocking the module',
level='critical')
os.system('touch %s/actor.failed-%s.lock' %
(base_eos_dir, os.getpid()))
sys.exit(-1)
## update the status with recovering removing manual
for wfo in session.query(Workflow).filter(
Workflow.name == wfname).all():
wfo.status = wfo.status.replace('manual', 'recovering')
session.commit()
if message_to_user:
print wfname, "to be notified to user(DUMMY)", message_to_user
if message_to_ops:
print 'message'
#sendEmail( "notification in recoveror" , message_to_ops, destination=['*****@*****.**'])
# sendLog('recoveror',message_to_ops,level='warning')
return
def stagor(url,specific =None, options=None):
if not componentInfo().check(): return
SI = siteInfo()
CI = campaignInfo()
UC = unifiedConfiguration()
done_by_wf_id = {}
done_by_input = {}
completion_by_input = {}
good_enough = 100.0
lost_blocks = json.loads(open('%s/lost_blocks_datasets.json'%monitor_dir).read())
lost_files = json.loads(open('%s/lost_files_datasets.json'%monitor_dir).read())
known_lost_blocks = {}
known_lost_files = {}
for dataset in set(lost_blocks.keys()+lost_files.keys()):
b,f = findLostBlocksFiles(url, dataset)
if dataset in lost_blocks and not b:
print dataset,"has no really lost blocks"
else:
known_lost_blocks[dataset] = [i['name'] for i in b]
if dataset in lost_files and not f:
print dataset,"has no really lost files"
else:
known_lost_files[dataset] = [i['name'] for i in f]
try:
cached_transfer_statuses = json.loads(open('cached_transfer_statuses.json').read())
except:
print "inexisting transfer statuses. starting fresh"
cached_transfer_statuses = {}
return False
transfer_statuses = {}
def time_point(label="",sub_lap=False):
now = time.mktime(time.gmtime())
nows = time.asctime(time.gmtime())
print "Time check (%s) point at : %s"%(label, nows)
print "Since start: %s [s]"% ( now - time_point.start)
if sub_lap:
print "Sub Lap : %s [s]"% ( now - time_point.sub_lap )
time_point.sub_lap = now
else:
print "Lap : %s [s]"% ( now - time_point.lap )
time_point.lap = now
time_point.sub_lap = now
time_point.sub_lap = time_point.lap = time_point.start = time.mktime(time.gmtime())
time_point("Check cached transfer")
## collect all datasets that are needed for wf in staging, correcting the status of those that are not really in staging
wfois = []
needs = defaultdict(list)
needs_by_priority = defaultdict(list)
for wfo in session.query(Workflow).filter(Workflow.status == 'staging').all():
wfi = workflowInfo(url, wfo.name)
if wfi.request['RequestStatus'] in ['running-open','running-closed','completed','assigned','acquired']:
wfi.sendLog('stagor', "is in status %s"%wfi.request['RequestStatus'])
wfi.status='away'
session.commit()
continue
if not wfi.request['RequestStatus'] in ['assignment-approved']:
## should be setting 'away' too
print wfo.name,"is",wfi.request['RequestStatus']
sendEmail("wrong status in staging. debug","%s is in %s, should set away."%(wfo.name,wfi.request['RequestStatus']))
wfois.append( (wfo,wfi) )
_,primaries,_,secondaries = wfi.getIO()
for dataset in list(primaries)+list(secondaries):
needs[wfo.name].append( dataset)
done_by_input[dataset] = {}
completion_by_input[dataset] = {}
needs_by_priority[wfi.request['RequestPriority']].append( dataset )
wfi.sendLog('stagor', '%s needs %s'%( wfo.name, dataset))
time_point("Check staging workflows")
open('%s/dataset_requirements.json'%monitor_dir,'w').write( json.dumps( needs, indent=2))
for prio in needs_by_priority: needs_by_priority[prio] = list(set(needs_by_priority[prio]))
open('%s/dataset_priorities.json'%monitor_dir,'w').write( json.dumps( needs_by_priority , indent=2))
dataset_endpoints = defaultdict(set)
endpoint_in_downtime = defaultdict(set)
#endpoint_completed = defaultdict(set)
endpoint_incompleted = defaultdict(set)
#endpoint = defaultdict(set)
send_back_to_considered = set()
## first check if anything is inactive
all_actives = set([transfer.phedexid for transfer in session.query(TransferImp).filter(TransferImp.active).all()])
for active_phedexid in all_actives:
skip = True
transfers_phedexid = session.query(TransferImp).filter(TransferImp.phedexid == active_phedexid).all()
for imp in transfers_phedexid:
if imp.workflow.status == 'staging':
skip =False
sendLog('stagor',"\t%s is staging for %s"%(imp.phedexid, imp.workflow.name))
if skip:
sendLog('stagor',"setting %s inactive" % active_phedexid)
for imp in transfers_phedexid:
imp.active = False
session.commit()
all_actives = sorted(set([transfer.phedexid for transfer in session.query(TransferImp).filter(TransferImp.active).all()]))
for phedexid in all_actives:
if specific: continue
## check on transfer completion
not_cached = False
if str(phedexid) in cached_transfer_statuses:
### use a cache for transfer that already looked done
sendLog('stagor',"read %s from cache"%phedexid)
checks = cached_transfer_statuses[str(phedexid)]
else:
## I actually would like to avoid that all I can
sendLog('stagor','Performing spurious transfer check on %s'% phedexid, level='critical')
checks = checkTransferStatus(url, phedexid, nocollapse=True)
if not checks:
## this is going to bias quite heavily the rest of the code. we should abort here
#sendLog('stagor','Ending stagor because of skewed input from checkTransferStatus', level='critical')
#return False
sendLog('stagor','Stagor has got a skewed input from checkTransferStatus', level='critical')
checks = {}
pass
#checks = {}
#not_cached = True
time_point("Check transfer status %s"% phedexid, sub_lap=True)
## just write this out
transfer_statuses[str(phedexid)] = copy.deepcopy(checks)
if not specific:
for dsname in checks:
if not dsname in done_by_input: done_by_input[dsname]={}
if not dsname in completion_by_input: completion_by_input[dsname] = {}
done_by_input[dsname][phedexid]=all(map(lambda i:i>=good_enough, checks[dsname].values()))
completion_by_input[dsname][phedexid]=checks[dsname].values()
if checks:
sendLog('stagor',"Checks for %s are %s"%( phedexid, [node.values() for node in checks.values()]))
done = all(map(lambda i:i>=good_enough,list(itertools.chain.from_iterable([node.values() for node in checks.values()]))))
else:
## it is empty, is that a sign that all is done and away ?
if not_cached:
print "Transfer status was not cached"
else:
print "ERROR with the scubscriptions API of ",phedexid
print "Most likely something else is overiding the transfer request. Need to work on finding the replacement automatically, if the replacement exists"
done = False
transfers_phedexid = session.query(TransferImp).filter(TransferImp.phedexid == phedexid).all()
for imp in transfers_phedexid:
tr_wf = imp.workflow
if tr_wf:# and tr_wf.status == 'staging':
if not tr_wf.id in done_by_wf_id: done_by_wf_id[tr_wf.id]={}
done_by_wf_id[tr_wf.id][phedexid]=done
if done:
imp.active = False
session.commit()
for ds in checks:
for s,v in checks[ds].items():
dataset_endpoints[ds].add( s )
if done:
sendLog('stagor',"%s is done"%phedexid)
cached_transfer_statuses[str(phedexid)] = copy.deepcopy(checks)
else:
sendLog('stagor',"%s is not finished %s"%(phedexid, pprint.pformat( checks )))
##pprint.pprint( checks )
## check if the destination is in down-time
for ds in checks:
sites_incomplete = [SI.SE_to_CE(s) for s,v in checks[ds].items() if v<good_enough]
sites_incomplete_down = [s for s in sites_incomplete if not s in SI.sites_ready]
## no space means no transfer should go there : NO, it does not work in the long run
#sites_incomplete_down = [SI.SE_to_CE(s) for s,v in checks[ds].items() if (v<good_enough and (SI.disk[s]==0 or (not SI.SE_to_CE(s) in SI.sites_ready)))]
if sites_incomplete_down:
sendLog('stagor',"%s are in downtime, while waiting for %s to get there"%( ",".join(sites_incomplete_down), ds))
#sites_complete = [SI.SE_to_CE(s) for s,v in checks[ds].items() if v>=good_enough]
#endpoint[ds].update( sites_complete )
#endpoint[ds].update( sites_incomplete )
#endpoint_completed[ds].update( sites_complete )
endpoint_incompleted[ds].update( sites_incomplete )
endpoint_in_downtime[ds].update( sites_incomplete_down )
time_point("Check on-going transfers")
print "End points"
for k in dataset_endpoints: dataset_endpoints[k] = list(dataset_endpoints[k])
print json.dumps( dataset_endpoints , indent=2)
print "End point in down time"
for k in endpoint_in_downtime: endpoint_in_downtime[k] = list(endpoint_in_downtime[k])
print json.dumps( endpoint_in_downtime , indent=2)
print "End point incomplete in down time"
for k in endpoint_incompleted: endpoint_incompleted[k] = list(endpoint_incompleted[k])
print json.dumps( endpoint_incompleted , indent=2)
#open('cached_transfer_statuses.json','w').write( json.dumps( cached_transfer_statuses, indent=2))
open('%s/transfer_statuses.json'%monitor_dir,'w').write( json.dumps( transfer_statuses, indent=2))
open('%s/dataset_endpoints.json'%monitor_dir,'w').write( json.dumps(dataset_endpoints, indent=2))
already_stuck = json.loads( open('%s/stuck_transfers.json'%monitor_pub_dir).read() ).keys()
already_stuck.extend( getAllStuckDataset() )
missing_in_action = defaultdict(list)
print "-"*10,"Checking on workflows in staging","-"*10
#forget_about = ['/MinBias_TuneCUETP8M1_13TeV-pythia8/RunIISummer15GS-MCRUN2_71_V1-v2/GEN-SIM']
#for what in forget_about:
# if not done_by_input[what]:
# done_by_input[what] = {'fake':True}
## come back to workflows and check if they can go
available_cache = defaultdict(lambda : defaultdict(float))
presence_cache = defaultdict(dict)
time_point("Preparing for more")
for wfo,wfi in wfois:
print "#"*30
time_point("Forward checking %s"% wfo.name,sub_lap=True)
## the site white list takes site, campaign, memory and core information
(_,primaries,_,secondaries,sites_allowed) = wfi.getSiteWhiteList(verbose=False)
se_allowed = [SI.CE_to_SE(site) for site in sites_allowed]
se_allowed.sort()
se_allowed_key = ','.join(se_allowed)
readys={}
for need in list(primaries)+list(secondaries):
if not need in done_by_input:
wfi.sendLog('stagor',"missing transfer report for %s"%need)
readys[need] = False
## should warn someone about this !!!
## it cannot happen, by construction
sendEmail('missing transfer report','%s does not have a transfer report'%(need))
continue
if not done_by_input[need] and need in list(secondaries):
wfi.sendLog('stagor',"assuming it is OK for secondary %s to have no attached transfers"% need)
readys[need] = True
done_by_input[need] = { "fake" : True }
continue
if len(done_by_input[need]) and all(done_by_input[need].values()):
wfi.sendLog('stagor',"%s is ready"%need)
print json.dumps( done_by_input[need] , indent=2)
readys[need] = True
else:
wfi.sendLog('stagor',"%s is not ready \n%s"%(need,json.dumps( done_by_input[need] , indent=2)))
readys[need] = False
if readys and all(readys.values()):
if wfo.status == 'staging':
wfi.sendLog('stagor',"all needs are fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
wfi.sendLog('stagor',"all needs are fullfilled, already")
print json.dumps( readys, indent=2 )
else:
wfi.sendLog('stagor',"missing requirements")
copies_needed,_ = wfi.getNCopies()
jump_ahead = False
re_transfer = False
## there is missing input let's do something more elaborated
for need in list(primaries):#+list(secondaries):
if endpoint_in_downtime[need] == endpoint_incompleted[need]:
#print need,"is going to an end point in downtime"
wfi.sendLog('stagor',"%s has only incomplete endpoint in downtime"%need)
re_transfer=True
if not se_allowed_key in available_cache[need]:
available_cache[need][se_allowed_key] = getDatasetBlocksFraction( url , need, sites=se_allowed )
if available_cache[need][se_allowed_key] >= copies_needed:
wfi.sendLog('stagor',"assuming it is OK to move on like this already for %s"%need)
jump_ahead = True
else:
wfi.sendLog('stagor',"Available %s times"% available_cache[need][se_allowed_key])
missing_and_downtime = list(set(endpoint_in_downtime[need]) & set(endpoint_incompleted[need]))
if missing_and_downtime:
wfi.sendLog('stagor',"%s is incomplete at %s which is in downtime, trying to move along"%(need, ','.join(missing_and_downtime)))
jump_ahead = True
else:
wfi.sendLog('stagor',"continue waiting for transfers for optimum production performance.")
## compute a time since staging to filter jump starting ?
# check whether the inputs is already in the stuck list ...
for need in list(primaries)+list(secondaries):
if need in already_stuck:
wfi.sendLog('stagor',"%s is stuck, so try to jump ahead"%need)
jump_ahead = True
if jump_ahead or re_transfer:
details_text = "checking on availability for %s to jump ahead"%wfo.name
details_text += '\n%s wants %s copies'%(wfo.name,copies_needed)
copies_needed = max(1,copies_needed-1)
details_text += '\nlowering by one unit to %s'%copies_needed
wfi.sendLog('stagor', details_text)
all_check = True
prim_where = set()
for need in list(primaries):
if not se_allowed_key in presence_cache[need]:
presence_cache[need][se_allowed_key] = getDatasetPresence( url, need , within_sites=se_allowed)
presence = presence_cache[need][se_allowed_key]
prim_where.update( presence.keys() )
available = available_cache[need][se_allowed_key]
this_check = (available >= copies_needed)
wfi.sendLog('stagor', "%s is available %s times (%s), at %s"%( need, available, this_check, se_allowed_key))
all_check &= this_check
if not all_check: break
for need in list(secondaries):
## I do not want to check on the secon
## this below does not function because the primary could be all available, and the secondary not complete at a certain site that does not matter at that point
this_check = all(done_by_input[need].values())
wfi.sendLog('stagor',"%s is this much transfered %s"%(need, json.dumps(done_by_input[need], indent=2)))
all_check&= this_check
#if not se_allowed_key in presence_cache[need]:
# presence_cache[need][se_allowed_key] = getDatasetPresence( url, need , within_sites=se_allowed)
## restrict to where the primary is
#presence = dict([(k,v) for (k,v) in presence_cache[need][se_allowed_key].items() if k in prim_where])
#this_check = all([there for (there,frac) in presence.values()])
#print need,"is present at all sites:",this_check
#all_check&= this_check
if all_check and not re_transfer:
wfi.sendLog('stagor',"needs are sufficiently fullfilled, setting staged")
wfo.status = 'staged'
session.commit()
else:
print wfo.name,"has to wait a bit more"
wfi.sendLog('stagor',"needs to wait a bit more")
else:
wfi.sendLog('stagor',"not checking availability")
if re_transfer:
wfi.sendLog('stagor',"Sending back to considered because of endpoint in downtime")
if wfo.status == 'staging':
wfo.status = 'considered'
session.commit()
send_back_to_considered.add( wfo.name )
time_point("Checked affected workflows")
if send_back_to_considered:
#sendEmail("transfer to endpoint in downtime","sending back to considered the following workflows \n%s"%('\n'.join( send_back_to_considered)))
sendLog('stagor', "sending back to considered the following workflows \n%s"%('\n'.join( send_back_to_considered)), level='critical')
print "-"*10,"Checking on non-available datasets","-"*10
## now check on those that are not fully available
for dsname in available_cache.keys():
## squash the se_allowed_key key
available_cache[dsname] = min( available_cache[dsname].values() )
really_stuck_dataset = set()
for dsname,available in available_cache.items():
using_its = getWorkflowByInput(url, dsname)
#print using_its
using_wfos = []
for using_it in using_its:
wf = session.query(Workflow).filter(Workflow.name == using_it).first()
if wf:
using_wfos.append( wf )
if not len(done_by_input[dsname]):
print "For dataset",dsname,"there are no transfer report. That's an issue."
for wf in using_wfos:
if wf.status == 'staging':
if UC.get("stagor_sends_back"):
print "sending",wf.name,"back to considered"
wf.status = 'considered'
session.commit()
#sendEmail( "send back to considered","%s was send back and might be trouble"% wf.name)
sendLog('stagor', "%s was send back and might be trouble"% wf.name, level='critical')
else:
print "would send",wf.name,"back to considered"
#sendEmail( "subscription lagging behind","susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name)
sendLog('stagor', "susbscriptions to get %s running are not appearing in phedex. I would have send it back to considered but that's not good."% wf.name, level='critical')
continue
## not compatible with checking on secondary availability
#if all([wf.status != 'staging' for wf in using_wfos]):
# ## means despite all checks that input is not needed
# continue
if available < 1.:
print "incomplete",dsname
## there is a problem in the method below that it does not account for files stuck in T1*Buffer only
lost_blocks,lost_files = findLostBlocksFiles( url, dsname ) if (not dsname.endswith('/RAW')) else ([],[])
lost_block_names = [item['name'] for item in lost_blocks]
lost_file_names = [item['name'] for item in lost_files]
if lost_blocks:
#print json.dumps( lost , indent=2 )
## estimate for how much !
fraction_loss,_,n_missing = getDatasetBlockFraction(dsname, lost_block_names)
print "We have lost",len(lost_block_names),"blocks",lost_block_names,"for %f%%"%(100.*fraction_loss)
if fraction_loss > 0.05: ## 95% completion mark
#sendEmail('we have lost too many blocks','%s is missing %d blocks, for %d events, %f %% loss'%(dsname, len(lost_block_names), n_missing, fraction_loss))
sendLog('stagor', '%s is missing %d blocks, for %d events, %3.2f %% loss'%(dsname, len(lost_block_names), n_missing, 100*fraction_loss), level='critical')
## the workflow should be rejected !
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
sendLog('stagor', '%s has too much loss on the input dataset %s. Missing %d blocks, for %d events, %3.2f %% loss'%(wf.name, dsname, len(lost_block_names), n_missing, 100*fraction_loss), level='critical')
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_blocks:
#sendEmail('we have lost a few blocks', '%s is missing %d blocks, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_block_names), n_missing, fraction_loss, '\n'.join( lost_block_names ) ))
sendLog('stagor', '%s is missing %d blocks, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_block_names), n_missing, fraction_loss, '\n'.join( lost_block_names ) ), level='critical')
known_lost_blocks[dsname] = [i['name'] for i in lost_blocks]
really_stuck_dataset.add( dsname )
if lost_files:
fraction_loss,_,n_missing = getDatasetFileFraction(dsname, lost_file_names)
print "We have lost",len(lost_file_names),"files",lost_file_names,"for %f%%"%fraction_loss
if fraction_loss > 0.05:
#sendEmail('we have lost too many files','%s is missing %d files, for %d events, %f %% loss'%(dsname, len(lost_file_names),n_missing, fraction_loss))
sendLog('stagor', '%s is missing %d files, for %d events, %f %% loss'%(dsname, len(lost_file_names),n_missing, fraction_loss), level='critical')
for wf in using_wfos:
if wf.status == 'staging':
print wf.name,"is doomed. setting to trouble"
wf.status = 'trouble'
session.commit()
else:
## probably enough to make a ggus and remove
if not dsname in known_lost_files:
#sendEmail('we have lost a few files','%s is missing %d files, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_file_names),n_missing, fraction_loss, '\n'.join(lost_file_names)))
sendLog('stagor', '%s is missing %d files, for %d events, %f %% loss\n\n%s'%(dsname, len(lost_file_names),n_missing, fraction_loss, '\n'.join(lost_file_names)), level='critical')
known_lost_files[dsname] = [i['name'] for i in lost_files]
## should the status be change to held-staging and pending on a ticket
missings = [pid for (pid,d) in done_by_input[dsname].items() if d==False]
print "\t",done_by_input[dsname]
print "\tneeds",len(done_by_input[dsname])
print "\tgot",done_by_input[dsname].values().count(True)
print "\tmissing",missings
missing_in_action[dsname].extend( missings )
rr= open('%s/lost_blocks_datasets.json'%monitor_dir,'w')
rr.write( json.dumps( known_lost_blocks, indent=2))
rr.close()
rr= open('%s/lost_files_datasets.json'%monitor_dir,'w')
rr.write( json.dumps( known_lost_files, indent=2))
rr.close()
open('%s/incomplete_transfers.json'%monitor_dir,'w').write( json.dumps(missing_in_action, indent=2) )
print "Stuck transfers and datasets"
print json.dumps( missing_in_action, indent=2 )
datasets_by_phid = defaultdict(set)
for dataset in missing_in_action:
for phid in missing_in_action[dataset]:
#print dataset,"stuck through",phid
datasets_by_phid[phid].add( dataset )
for k in datasets_by_phid:
datasets_by_phid[k] = list(datasets_by_phid[k])
open('datasets_by_phid.json','w').write( json.dumps(datasets_by_phid, indent=2 ))
open('really_stuck_dataset.json','w').write( json.dumps(list(really_stuck_dataset), indent=2 ))
print '\n'*2,"Datasets really stuck"
print '\n'.join( really_stuck_dataset )
#############
## not going further for what matters
#############
return