def __init__(self, cfg_params, parsed_range, val):
self.cfg_params = cfg_params
self.limitJobs = True
# get user request
self.nsjobs = -1
self.chosenJobsList = None
if val:
if val=='range': # for Resubmitter
self.chosenJobsList = parsed_range
elif val=='all':
pass
elif (type(eval(val)) is int) and eval(val) > 0:
# positive number
self.nsjobs = eval(val)
elif (type(eval(val)) is tuple)or( type(eval(val)) is int and eval(val)<0 ) :
self.chosenJobsList = parsed_range
self.nsjobs = len(self.chosenJobsList)
else:
msg = 'Bad submission option <'+str(val)+'>\n'
msg += ' Must be an integer or "all"'
msg += ' Generic range is not allowed"'
raise CrabException(msg)
pass
self.seWhiteList = parseIntoList(cfg_params.get('GRID.se_white_list',[]))
self.seBlackList = parseIntoList(cfg_params.get('GRID.se_black_list',[]))
self.datasetPath = cfg_params['CMSSW.datasetpath']
if string.lower(self.datasetPath)=='none':
self.datasetPath = None
self.scram = Scram.Scram(cfg_params)
return
class ServerCommunicator:
"""
Common interface for the interaction between the Crab client and the server Web Service
"""
def __init__(self, serverName, serverPort, cfg_params, proxyPath=None):
"""
Open the communication with an Analysis Server by passing the server URL and the port
"""
self.ServerTwiki = 'https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabServerForUsers#Server_available_for_users'
self.asSession = C_AS_Session(serverName, serverPort)
self.cfg_params = cfg_params
self.userSubj = ''
self.serverName = serverName
credentialType = 'Proxy'
if common.scheduler.name().upper() in ['CAF','LSF']:
credentialType = 'Token'
CliServerParams(self)
self.crab_task_name = common.work_space.topDir().split('/')[-2] # nice task name "crab_0_..."
configAPI = {'credential' : credentialType, \
'logger' : common.logger() }
CredAPI = CredentialAPI( configAPI )
try:
self.userSubj = CredAPI.getSubject()
except Exception, err:
common.logger.debug("Getting Credential Subject: " +str(traceback.format_exc()))
raise CrabException("Error Getting Credential Subject")
self.scram=Scram.Scram(cfg_params)
def __init__(self, cfg_params, template):
# pattern -> action
self.actions = {
'title': self.title_,
'untar_software': self.untarSoftware_,
'initial_environment': self.initialEnvironment_,
'setup_scheduler_environment': self.setupSchedulerEnvironment_,
'setup_jobtype_environment': self.setupJobTypeEnvironment_,
'copy_input': self.copyInput_,
'rewrite_cmssw_cfg': self.rewriteCMSSWcfg_,
'build_executable': self.buildExe_,
'run_executable': self.runExe_,
'rename_output': self.renameOutput_,
'copy_output': self.copyOutput_,
'parse_report': self.parseReport_,
'modify_report': self.modifyReport_,
'func_exit': self.func_exit_
}
if os.path.isfile("./" + template):
self.template = "./" + template
elif os.getenv('CRABDIR') and os.path.isfile(
os.getenv('CRABDIR') + '/python/' + template):
self.template = os.getenv('CRABDIR') + '/python/' + template
else:
raise CrabException("No crab_template.sh found!")
self.nj = -1 # current job number
try:
self.scram = Scram.Scram(None)
self.CMSSWversion = self.scram.getSWVersion()
parts = self.CMSSWversion.split('_')
self.CMSSW_major = int(parts[1])
self.CMSSW_minor = int(parts[2])
if parts[3] == 'DEVEL': parts[3] = '0'
self.CMSSW_patch = int(parts[3])
except:
raise CrabException("Could not determine CMSSW version")
self.debug_wrapper = ''
debug = int(cfg_params.get('USER.debug_wrapper', 0))
if debug == 1: self.debug_wrapper = '--debug'
self.scriptName = cfg_params.get('CRAB.jobtype').upper() + '.sh'
self.pset = cfg_params.get('CMSSW.pset', 'None')
return
def sched_parameter(self, i, task):
"""
Return scheduler-specific parameters. Used at crab -submit time
by $CRABPYTHON/Scheduler.py
"""
#SB paste from crab ScheduerGlidein
jobParams = ""
(self.rcondorHost,
self.rcondorUserHost) = self.pickRcondorSubmissionHost(task)
seDest = task.jobs[i - 1]['dlsDestination']
if seDest == ['']:
seDest = self.blackWhiteListParser.expandList("T") # all of SiteDB
seString = self.blackWhiteListParser.cleanForBlackWhiteList(seDest)
jobParams += '+DESIRED_SEs = "' + seString + '"; '
scram = Scram.Scram(None)
cmsVersion = scram.getSWVersion()
scramArch = scram.getArch()
cmsver = re.split('_', cmsVersion)
numericCmsVersion = "%s%.2d%.2d" % (cmsver[1], int(
cmsver[2]), int(cmsver[3]))
jobParams += '+DESIRED_CMSVersion ="' + cmsVersion + '";'
jobParams += '+DESIRED_CMSVersionNr ="' + numericCmsVersion + '";'
jobParams += '+DESIRED_CMSScramArch ="' + scramArch + '";'
myscheddName = self.rcondorHost
jobParams += '+Glidein_MonitorID = "https://' + myscheddName + '//$(Cluster).$(Process)"; '
if (self.EDG_clock_time):
jobParams += '+MaxWallTimeMins = ' + self.EDG_clock_time + '; '
else:
jobParams += '+MaxWallTimeMins = %d; ' % (60 * 24)
common._db.updateTask_({'jobType': jobParams})
return jobParams
def listMatch(self, seList, full, onlyOSG=True):
"""
Check the compatibility of available resources
"""
scram = Scram.Scram(None)
versionCMSSW = scram.getSWVersion()
arch = scram.getArch()
if self.selectNoInput:
availCEs = listAllCEs(versionCMSSW, arch, onlyOSG=onlyOSG)
else:
seDest = self.blackWhiteListParser.cleanForBlackWhiteList(seList, "list")
availCEs = getJobManagerList(seDest, versionCMSSW,
arch, onlyOSG=onlyOSG)
uniqCEs = []
for ce in availCEs:
if ce not in uniqCEs:
uniqCEs.append(ce)
ceDest = self.ceBlackWhiteListParser.cleanForBlackWhiteList(uniqCEs, "list")
return ceDest
def __init__(self, cfg_params, ncjobs):
# parameter
self.cfg_params = cfg_params
self.ncjobs = ncjobs
# dataset
try:
self.Dataset = cfg_params['SKIM.dataset']
except:
msg = 'Please specify a dataset'
raise SkimException(msg)
# pset
try:
self.pset = cfg_params['SKIM.pset']
if (not os.path.exists(self.pset)):
raise SkimException("User defined PSet file " + self.pset +
" does not exist")
except KeyError:
raise SkimException("PSet file missing. Cannot run cmsRun ")
# scram
self.scram = Scram.Scram(cfg_params)
scramArea = ''
self.additional_input_files = []
self.version = self.scram.getSWVersion()
# name of tarball
self.tgz_name = 'default.tgz'
# additional input files
try:
tmpAddFiles = string.split(
cfg_params['SKIM.additional_input_files'], ',')
for tmp in tmpAddFiles:
if not os.path.exists(tmp):
raise SkimException("Additional input file not found: " +
tmp)
self.additional_input_files.append(string.strip(tmp))
pass
pass
except KeyError:
pass
# files per job
try:
if (cfg_params['SKIM.files_per_jobs']):
self.filesPerJob = int(cfg_params['SKIM.files_per_job'])
self.selectFilesPerJob = 1
except KeyError:
self.filesPerJob = -1
self.selectFilesPerJob = 0
# Events per job
try:
self.eventsPerJob = int(cfg_params['SKIM.events_per_job'])
self.selectEventsPerJob = 1
except KeyError:
self.eventsPerJob = -1
self.selectEventsPerJob = 0
# number of jobs
try:
self.theNumberOfJobs = int(cfg_params['SKIM.number_of_jobs'])
self.selectNumberOfJobs = 1
except KeyError:
self.theNumberOfJobs = 0
self.selectNumberOfJobs = 0
# total number of events
try:
self.total_number_of_events = int(
cfg_params['SKIM.total_number_of_events'])
self.selectTotalNumberEvents = 1
except KeyError:
self.total_number_of_events = 0
self.selectTotalNumberEvents = 0
# total number of files
try:
self.total_number_of_files = int(
cfg_params['SKIM.total_number_of_files'])
self.selectTotalNumberFiles = 1
except KeyError:
self.total_number_of_files = 0
self.selectTotalNumberFiles = 0
if (self.selectFilesPerJob + self.selectEventsPerJob == 2):
msg = 'Must define either files per job or events per job.'
raise SkimException(msg)
if (self.selectEventsPerJob):
if ((self.selectTotalNumberEvents + self.selectEventsPerJob +
self.selectNumberOfJobs) != 2):
msg = 'Must define exactly two of total_number_of_events, events_per_job, or number_of_jobs.'
raise SkimException(msg)
if (self.selectFilesPerJob):
if ((self.selectTotalNumberFiles + self.selectNumberOfJobs +
self.selectFilesPerJob) != 2):
msg = 'Must define exactly two of total_number_of_events, events_per_job, or number_of_jobs.'
raise SkimException(msg)
return
def sched_parameter(self, i, task):
"""
Return scheduler-specific parameters. Used at crab -submit time
by $CRABPYTHON/Scheduler.py
"""
#SB paste from crab SchedulerGlidein
jobParams = ""
(self.remoteHost,
self.remoteUserHost) = self.pickRemoteSubmissionHost(task)
seDest = task.jobs[i - 1]['dlsDestination']
if seDest == ['']:
seDest = self.blackWhiteListParser.expandList("T") # all of SiteDB
seString = self.blackWhiteListParser.cleanForBlackWhiteList(seDest)
# beware SiteDB V2 API, explicely cast to string in case it is unicode
seString = str(seString)
jobParams += '+DESIRED_SEs = "' + seString + '"; '
scram = Scram.Scram(None)
cmsVersion = scram.getSWVersion()
scramArch = scram.getArch()
cmsver = re.split('_', cmsVersion)
numericCmsVersion = "%s%.2d%.2d" % (cmsver[1], int(
cmsver[2]), int(cmsver[3]))
# protect against datasetPath being None
jobParams += '+DESIRED_CMSDataset ="' + str(self.datasetPath) + '";'
jobParams += '+DESIRED_CMSVersion ="' + cmsVersion + '";'
jobParams += '+DESIRED_CMSVersionNr ="' + numericCmsVersion + '";'
jobParams += '+DESIRED_CMSScramArch ="' + scramArch + '";'
userName = gethnUserNameFromSiteDB()
jobParams += '+AccountingGroup ="' + userName + '";'
myscheddName = self.remoteHost
jobParams += '+Glidein_MonitorID = "https://'+ myscheddName + \
'//' + self.submissionDay + '//$(Cluster).$(Process)"; '
if (self.EDG_clock_time):
glideinTime = "%d" % (int(self.EDG_clock_time) + 20
) # 20 min to wrapup
jobParams += '+MaxWallTimeMins = ' + glideinTime + '; '
else:
jobParams += '+MaxWallTimeMins = %d; ' % (
21 * 60 + 55) # 21:55h (unit = min)
if self.max_rss:
jobParams += 'request_memory = ' + self.max_rss + ';'
if self.allowOverflow == "0":
jobParams += '+CMS_ALLOW_OVERFLOW = False; '
if self.EDG_addJdlParam:
if self.EDG_addJdlParam[-1] == '':
self.EDG_addJdlParam = self.EDG_addJdlParam[:-1]
for p in self.EDG_addJdlParam:
jobParams += p.strip() + ';\n'
common._db.updateTask_({'jobType': jobParams})
return jobParams
def sched_parameter(self, i, task):
"""
Return scheduler-specific parameters. Used at crab -submit time
by $CRABPYTHON/Scheduler.py
"""
#SB paste from crab SchedulerGlidein
jobParams = ""
(self.remoteHost,
self.remoteUserHost) = self.pickRemoteSubmissionHost(task)
psnDest = task.jobs[i - 1]['dlsDestination']
if psnDest == ['']:
# crab.cfg had datasetpath = None
pnn2psn = getMapOfPhedexNodeName2ProcessingNodeNameFromSiteDB()
allPSNs = set(pnn2psn.values()) # set removes duplicates
psnDest = allPSNs
blackList = parseIntoList(self.cfg_params.get("GRID.se_black_list",
[]))
whiteList = parseIntoList(self.cfg_params.get("GRID.se_white_list",
[]))
psnDest = cleanPsnListForBlackWhiteLists(psnDest, blackList, whiteList)
if not psnDest or psnDest == [] or psnDest == ['']:
msg = "No Processing Site Name after applying black/white list."
msg += " Can't submit"
msg += "\nUsed Black List: %s" % blackList
msg += "\nUsed White List: %s" % whiteList
common.logger.info(msg)
raise CrabException(msg)
# condor's JDL needs a string of CSV's, not a list
psnDest = ",".join(psnDest)
msg = "list of PSN's for submission: %s" % psnDest
common.logger.info(msg)
jobParams += '+DESIRED_Sites = "%s";' % psnDest
scram = Scram.Scram(None)
cmsVersion = scram.getSWVersion()
scramArch = scram.getArch()
cmsver = re.split('_', cmsVersion)
if cmsver[3] == 'DEVEL': cmsver[3] = '0'
numericCmsVersion = "%s%.2d%.2d" % (cmsver[1], int(
cmsver[2]), int(cmsver[3]))
if "slc5" in scramArch:
cmsOpSys = "LINUX"
cmsOpSysMajorVer = "5,6" # SL5 exe's also run on SL6
if "slc6" in scramArch:
cmsOpSys = "LINUX"
cmsOpSysMajorVer = "6"
if "ia32" in scramArch:
cmsArch = "INTEL,X86_64" # 32bit exe's also run on 64bit
if "amd64" in scramArch:
cmsArch = "X86_64"
# protect against datasetPath being None
jobParams += '+DESIRED_CMSDataset ="' + str(self.datasetPath) + '";'
jobParams += '+DESIRED_CMSVersion ="' + cmsVersion + '";'
jobParams += '+DESIRED_CMSVersionNr ="' + numericCmsVersion + '";'
jobParams += '+DESIRED_CMSScramArch ="' + scramArch + '";'
jobParams += '+DESIRED_OpSyses ="' + cmsOpSys + '";'
jobParams += '+DESIRED_OpSysMajorVers ="' + cmsOpSysMajorVer + '";'
jobParams += '+DESIRED_Archs ="' + cmsArch + '";'
userName = gethnUserNameFromSiteDB()
jobParams += '+AccountingGroup = "analysis.%s";' % userName
myscheddName = self.remoteHost
jobParams += '+Glidein_MonitorID = "https://'+ myscheddName + \
'//' + self.submissionDay + '//$(Cluster).$(Process)"; '
if (self.EDG_clock_time):
glideinTime = "%d" % (int(self.EDG_clock_time) + 20
) # 20 min to wrapup
jobParams += '+MaxWallTimeMins = ' + glideinTime + '; '
else:
jobParams += '+MaxWallTimeMins = %d; ' % (
21 * 60 + 55) # 21:55h (unit = min)
if self.max_rss:
jobParams += 'request_memory = ' + self.max_rss + ';'
if self.allowOverflow == "0":
jobParams += '+CMS_ALLOW_OVERFLOW = False; '
if self.EDG_addJdlParam:
if self.EDG_addJdlParam[-1] == '':
self.EDG_addJdlParam = self.EDG_addJdlParam[:-1]
for p in self.EDG_addJdlParam:
jobParams += p.strip() + ';\n'
common._db.updateTask_({'jobType': jobParams})
return jobParams