def execute( self ):
"""Main Agent code:
1.- Query TaskQueueDB for existing TQs
2.- Count Pending Jobs
3.- Submit Jobs
"""
self.__checkSubmitPools()
bigDataJobsToSubmit = {}
bigDataJobIdsToSubmit = {}
for directorName, directorDict in self.directors.items():
self.log.verbose( 'Checking Director:', directorName )
self.log.verbose( 'RunningEndPoints:', directorDict['director'].runningEndPoints )
for runningEndPointName in directorDict['director'].runningEndPoints:
runningEndPointDict = directorDict['director'].runningEndPoints[runningEndPointName]
NameNode = runningEndPointDict['NameNode']
jobsByEndPoint = 0
result = BigDataDB.getBigDataJobsByStatusAndEndpoint( 'Submitted', NameNode )
if result['OK']:
jobsByEndPoint += len( result['Value'] )
result = BigDataDB.getBigDataJobsByStatusAndEndpoint( 'Running', NameNode )
if result['OK']:
jobsByEndPoint += len( result['Value'] )
self.log.verbose( 'Checking Jobs By EndPoint %s:' % jobsByEndPoint )
jobLimitsEndPoint = runningEndPointDict['LimitQueueJobsEndPoint']
bigDataJobs = 0
if jobsByEndPoint >= jobLimitsEndPoint:
self.log.info( '%s >= %s Running jobs reach job limits: %s, skipping' % ( jobsByEndPoint, jobLimitsEndPoint, runningEndPointName ) )
continue
else:
bigDataJobs = jobLimitsEndPoint - jobsByEndPoint
requirementsDict = runningEndPointDict['Requirements']
self.log.info( 'Requirements Dict: ', requirementsDict )
result = taskQueueDB.getMatchingTaskQueues( requirementsDict )
if not result['OK']:
self.log.error( 'Could not retrieve TaskQueues from TaskQueueDB', result['Message'] )
return result
taskQueueDict = result['Value']
self.log.info( 'Task Queues Dict: ', taskQueueDict )
jobs = 0
priority = 0
cpu = 0
jobsID = 0
self.log.info( 'Pending Jobs from TaskQueue, which not matching before: ', self.pendingTaskQueueJobs )
for tq in taskQueueDict:
jobs += taskQueueDict[tq]['Jobs']
priority += taskQueueDict[tq]['Priority']
cpu += taskQueueDict[tq]['Jobs'] * taskQueueDict[tq]['CPUTime']
#Matching of Jobs with BigData Softwares
#This process is following the sequence:
#Retrieve a job from taskqueueDict
#Get job name and try to match with the resources
#If not match store the var pendingTaskQueueJobs for the
#next iteration
#
#This matching is doing with the following JobName Pattern
# NameSoftware _ SoftwareVersion _ HighLanguageName _ HighLanguageVersion _ DataSetName
#extract a job from the TaskQueue
if tq not in self.pendingTaskQueueJobs.keys():
self.pendingTaskQueueJobs[tq] = {}
getJobFromTaskQueue = taskQueueDB.matchAndGetJob( taskQueueDict[tq] )
if not getJobFromTaskQueue['OK']:
self.log.error( 'Could not get Job and FromTaskQueue', getJobFromTaskQueue['Message'] )
return getJobFromTaskQueue
jobInfo = getJobFromTaskQueue['Value']
jobID = jobInfo['jobId']
jobAttrInfo = jobDB.getJobAttributes( jobID )
if not jobAttrInfo['OK']:
self.log.error( 'Could not get Job Attributes', jobAttrInfo['Message'] )
return jobAttrInfo
jobInfoUniq = jobAttrInfo['Value']
jobName = jobInfoUniq['JobName']
self.pendingTaskQueueJobs[tq][jobID] = jobName
result = jobDB.getJobJDL( jobID, True )
classAdJob = ClassAd( result['Value'] )
arguments = 0
if classAdJob.lookupAttribute( 'Arguments' ):
arguments = classAdJob.getAttributeString( 'Arguments' )
#if not classAdJob.lookupAttribute( 'Arguments' ):
# continue
jobsToSubmit = self.matchingJobsForBDSubmission( arguments,
runningEndPointName,
runningEndPointDict['BigDataSoftware'],
runningEndPointDict['BigDataSoftwareVersion'],
runningEndPointDict['HighLevelLanguage']['HLLName'],
runningEndPointDict['HighLevelLanguage']['HLLVersion'],
jobID )
if ( jobsToSubmit == "OK" ):
if directorName not in bigDataJobsToSubmit:
bigDataJobsToSubmit[directorName] = {}
if runningEndPointName not in bigDataJobsToSubmit[directorName]:
bigDataJobsToSubmit[directorName][runningEndPointName] = {}
bigDataJobsToSubmit[directorName][runningEndPointName] = { 'JobId': jobID,
'JobName': jobName,
'TQPriority': priority,
'CPUTime': cpu,
'BigDataEndpoint': runningEndPointName,
'BigDataEndpointNameNode': runningEndPointDict['NameNode'],
'BdSoftware': runningEndPointDict['BigDataSoftware'],
'BdSoftwareVersion': runningEndPointDict['BigDataSoftwareVersion'],
'HLLName' : runningEndPointDict['HighLevelLanguage']['HLLName'],
'HLLVersion' : runningEndPointDict['HighLevelLanguage']['HLLVersion'],
'NumBigDataJobsAllowedToSubmit': bigDataJobs,
'SiteName': runningEndPointDict['SiteName'],
'PublicIP': runningEndPointDict['PublicIP'],
'User': runningEndPointDict['User'],
'Port': runningEndPointDict['Port'],
'UsePilot': runningEndPointDict['UsePilot'],
'IsInteractive': runningEndPointDict['IsInteractive'],
'Arguments': arguments }
del self.pendingTaskQueueJobs[tq][jobID]
else:
self.log.error( jobsToSubmit )
self.log.info( 'Pending Jobs from TaskQueue, which not matching after: ', self.pendingTaskQueueJobs )
for tq in self.pendingTaskQueueJobs.keys():
for jobid in self.pendingTaskQueueJobs[tq].keys():
result = jobDB.getJobJDL( jobid, True )
classAdJob = ClassAd( result['Value'] )
arguments = 0
if classAdJob.lookupAttribute( 'Arguments' ):
arguments = classAdJob.getAttributeString( 'Arguments' )
#if not classAdJob.lookupAttribute( 'Arguments' ):
# continue
#do the match with the runningEndPoint
jobsToSubmit = self.matchingJobsForBDSubmission( arguments,
runningEndPointName,
runningEndPointDict['BigDataSoftware'],
runningEndPointDict['BigDataSoftwareVersion'],
runningEndPointDict['HighLevelLanguage']['HLLName'],
runningEndPointDict['HighLevelLanguage']['HLLVersion'],
jobid )
if ( jobsToSubmit == "OK" ):
if directorName not in bigDataJobsToSubmit:
bigDataJobsToSubmit[directorName] = {}
if runningEndPointName not in bigDataJobsToSubmit[directorName]:
bigDataJobsToSubmit[directorName][runningEndPointName] = {}
bigDataJobsToSubmit[directorName][runningEndPointName] = { 'JobId': jobid,
'JobName': self.pendingTaskQueueJobs[tq][jobid],
'TQPriority': priority,
'CPUTime': cpu,
'BigDataEndpoint': runningEndPointName,
'BigDataEndpointNameNode': runningEndPointDict['NameNode'],
'BdSoftware': runningEndPointDict['BigDataSoftware'],
'BdSoftwareVersion': runningEndPointDict['BigDataSoftwareVersion'],
'HLLName' : runningEndPointDict['HighLevelLanguage']['HLLName'],
'HLLVersion' : runningEndPointDict['HighLevelLanguage']['HLLVersion'],
'NumBigDataJobsAllowedToSubmit': bigDataJobs,
'SiteName': runningEndPointDict['SiteName'],
'PublicIP': runningEndPointDict['PublicIP'],
'User': runningEndPointDict['User'],
'Port': runningEndPointDict['Port'],
'UsePilot': runningEndPointDict['UsePilot'],
'IsInteractive': runningEndPointDict['IsInteractive'],
'Arguments': arguments }
del self.pendingTaskQueueJobs[tq][jobid]
else:
self.log.error( jobsToSubmit )
if not jobs and not self.pendingTaskQueueJobs:
self.log.info( 'No matching jobs for %s found, skipping' % NameNode )
continue
self.log.info( '___BigDataJobsTo Submit:', bigDataJobsToSubmit )
for directorName, JobsToSubmitDict in bigDataJobsToSubmit.items():
for runningEndPointName, jobsToSubmitDict in JobsToSubmitDict.items():
if self.directors[directorName]['isEnabled']:
self.log.info( 'Requesting submission to %s of %s' % ( runningEndPointName, directorName ) )
director = self.directors[directorName]['director']
pool = self.pools[self.directors[directorName]['pool']]
jobIDs = JobsToSubmitDict[runningEndPointName]['JobId']
jobName = JobsToSubmitDict[runningEndPointName]['JobName']
endpoint = JobsToSubmitDict[runningEndPointName]['BigDataEndpoint']
runningSiteName = JobsToSubmitDict[runningEndPointName]['SiteName']
NameNode = JobsToSubmitDict[runningEndPointName]['BigDataEndpointNameNode']
BigDataSoftware = JobsToSubmitDict[runningEndPointName]['BdSoftware']
BigDataSoftwareVersion = JobsToSubmitDict[runningEndPointName]['BdSoftwareVersion']
HLLName = JobsToSubmitDict[runningEndPointName]['HLLName']
HLLVersion = JobsToSubmitDict[runningEndPointName]['HLLVersion']
PublicIP = JobsToSubmitDict[runningEndPointName]['PublicIP']
User = JobsToSubmitDict[runningEndPointName]['User']
Port = JobsToSubmitDict[runningEndPointName]['Port']
UsePilot = JobsToSubmitDict[runningEndPointName]['UsePilot']
IsInteractive = JobsToSubmitDict[runningEndPointName]['IsInteractive']
Arguments = JobsToSubmitDict[runningEndPointName]['Arguments']
numBigDataJobsAllowed = JobsToSubmitDict[runningEndPointName]['NumBigDataJobsAllowedToSubmit']
ret = pool.generateJobAndQueueIt( director.submitBigDataJobs,
args = ( endpoint, numBigDataJobsAllowed, runningSiteName, NameNode,
BigDataSoftware, BigDataSoftwareVersion, HLLName, HLLVersion,
PublicIP, Port, jobIDs, runningEndPointName, jobName, User, self.jobDataset, UsePilot, IsInteractive ),
oCallback = self.callBack,
oExceptionCallback = director.exceptionCallBack,
blocking = False )
if not ret['OK']:
# Disable submission until next iteration
self.directors[directorName]['isEnabled'] = False
else:
time.sleep( self.am_getOption( 'ThreadStartDelay' ) )
if 'Default' in self.pools:
# only for those in "Default' thread Pool
# for pool in self.pools:
self.pools['Default'].processAllResults()
return DIRAC.S_OK()
