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')
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()
if __name__ == '__main__':
url='cmsweb.cern.ch'
parser = optparse.OptionParser()
#parser.add_option('--do',default=False,action='store_true')
parser.add_option('--test', dest='do', default=True,action='store_false')
parser.add_option('--leave',dest='ass',default=True,action='store_false')
parser.add_option('--go',default=False,action='store_true',help="override possible blocking conditions")
(options,args) = parser.parse_args()
spec=None
if len(args)!=0:
spec = args[0]
if not options.do: options.ass=False
recoveror(url,spec,options=options)
fdb = closeoutInfo()
fdb.html()
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
if __name__ == '__main__':
url = reqmgr_url
parser = optparse.OptionParser()
parser.add_option('--test', dest='do', default=True, action='store_false')
parser.add_option('--leave',
dest='ass',
default=True,
action='store_false')
parser.add_option('--go',
default=False,
action='store_true',
help="override possible blocking conditions")
parser.add_option('--new', default=False, action='store_true')
(options, args) = parser.parse_args()
spec = None
if len(args) != 0:
spec = args[0]
if not options.do: options.ass = False
if options.new:
new_recoveror(url, spec, options=options)
else:
recoveror(url, spec, options=options)
fdb = closeoutInfo()
fdb.html()
#from showError import parse_all
#parse_all(url)
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 rulor(spec=None, options=None):
mlock = moduleLock()
if mlock(): return
up = componentInfo(soft=['mcm', 'wtc'])
if not up.check(): return
if spec:
wfs = session.query(Workflow).filter(
Workflow.status.contains('manual')).filter(
Workflow.name.contains(spec)).all()
else:
wfs = session.query(Workflow).filter(
Workflow.status.contains('manual')).all()
COI = closeoutInfo()
RI = reportInfo()
WC = wtcClient()
JC = JIRAClient()
## a list of function with a given trace ( wfi, record, report) => (action dict list)
rules = [
majority_of_139_nanoaod,
majority_of_71104,
]
for wfo in wfs:
wfi = workflowInfo(reqmgr_url, wfo.name)
record = COI.get(wfo.name)
report = RI.get(wfo.name)
if not record:
print "no information to look at"
continue
print "close out information as in the assistance page"
print json.dumps(record, indent=2)
print "report information as in the unified report"
print json.dumps(report, indent=2)
## parse the information and produce an action document
### a rule for on-going issue with memory in campaign ...
acted = False
for condition in rules:
acts = condition(wfi, record, report)
if acts:
print "list of actions being taken for", wfo.name
for a in acts:
print json.dumps(a, indent=2)
if not options.test:
acted = True
WC.set_actions(acts)
wfo.status = wfo.status.replace('manual', 'acting')
session.commit()
break
if acted: continue
if "some conditions":
action_doc = {
'workflow': wfo.name,
'name': "a task name",
'parameters': {
'action': 'acdc',
'memory': 5000
}
}
acted = True
if acted: continue
if "majority of 139":
pass
if acted: continue
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 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 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 closor(url, specific=None, options=None):
if userLock(): return
mlock = moduleLock()
if mlock(): return
if not componentInfo().check(): return
UC = unifiedConfiguration()
CI = campaignInfo()
BI = batchInfo()
CloseI = closeoutInfo()
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()
CloseI.pop(wfo.name)
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')
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)
def checkor(url, spec=None, options=None):
fDB = closeoutInfo()
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 = 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
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.*60.
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:
print wfo.name,"on hold"
continue
if wfo.name in holdings:
wfo.status = 'assistance-onhold'
print "setting",wfo.name,"on hold"
session.commit()
continue
if wfo.wm_status != 'completed':
## 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
## do the closed-out checks one by one
## 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
# 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']
fractions_pass = {}
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 )
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.
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 = ['MINIAODSIM']
out_worth_checking = [out for out in custodial_locations.keys() if out.split('/')[-1] not in vetoed_custodial_tier]
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]
## 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
## 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 not custodial and pick_custodial:
## pick one at random
custodial = SI.pick_SE()
if custodial and ((not sub_assistance and not acdc) or by_pass_checks):
## 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"
else:
print "cannot find a custodial for",wfo.name
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:
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 lumi count:\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
