def submitBigDataJobs( self, endpoint, numBigDataJobsAllowed, runningSiteName, NameNode,
BigDataSoftware, BigDataSoftwareVersion, HLLName, HLLVersion,
PublicIP, Port, jobIds , runningEndPointName, JobName, User, dataset, UsePilot, IsInteractive ):
"""
Big Data job submission with all the parameters of SITE and Job
"""
self.log.info( 'Director:submitBigDataJobs:JobSubmisionProcess' )
if ( numBigDataJobsAllowed <= 0 ):
return S_ERROR( "Number of slots reached for %s in the NameNode " % runningSiteName, NameNode )
if NameNode not in self.runningEndPoints[endpoint]['NameNode']:
return S_ERROR( 'Unknown NameNode: %s' % NameNode )
newJob = BigDataDB.insertBigDataJob( jobIds, JobName, Time.toString(), NameNode,
runningSiteName, PublicIP, "", "",
"", BigDataSoftware, BigDataSoftwareVersion, HLLName,
HLLVersion, "Submitted" )
self.log.info( 'Director:submitBigDataJobs:SubmitJob' )
dictBDJobSubmitted = self._submitBigDataJobs( NameNode, Port, jobIds, PublicIP,
runningEndPointName, User, JobName, dataset, UsePilot, IsInteractive )
if not dictBDJobSubmitted[ 'OK' ]:
return dictBDJobSubmitted
bdjobID = dictBDJobSubmitted['Value']
result = BigDataDB.setHadoopID( jobIds, bdjobID )
if not result[ 'OK' ]:
S_ERROR( "BigData ID not updated" )
result = BigDataDB.setIntoJobDBStatus( jobIds, "Submitted", "",
runningSiteName, bdjobID )
if not result[ 'OK' ]:
S_ERROR( "JobDB of BigData Soft not updated" )
self.log.info( 'Director:submitBigDataJobs:JobSubmitted' )
return S_OK( "OK" )
def __updateSandBox( self, jobid, output ):
jobInfo = BigDataDB.getJobIDInfo( jobid )
result = self.sshConnect.scpCall( 100, output, output, False )
if not result['OK']:
self.log.error( 'Error to get the data from BigData Software DFS:', result )
file_paths = []
file_paths.append( output )
outputSandbox = file_paths
resolvedSandbox = self.__resolveOutputSandboxFiles( outputSandbox )
if not resolvedSandbox['OK']:
self.log.warn( 'Output sandbox file resolution failed:' )
self.log.warn( resolvedSandbox['Message'] )
self.__report( 'Failed', 'Resolving Output Sandbox' )
fileList = resolvedSandbox['Value']['Files']
missingFiles = resolvedSandbox['Value']['Missing']
if missingFiles:
self.jobReport.setJobParameter( 'OutputSandboxMissingFiles', ', '.join( missingFiles ), sendFlag = False )
if fileList and jobid:
self.outputSandboxSize = getGlobbedTotalSize( fileList )
self.log.info( 'Attempting to upload Sandbox with limit:', self.sandboxSizeLimit )
result = self.sandboxClient.uploadFilesAsSandboxForJob( fileList, jobid,
'Output', self.sandboxSizeLimit ) # 1024*1024*10
if not result['OK']:
self.log.error( 'Output sandbox upload failed with message', result['Message'] )
if result.has_key( 'SandboxFileName' ):
outputSandboxData = result['SandboxFileName']
self.log.info( 'Attempting to upload %s as output data' % ( outputSandboxData ) )
outputData.append( outputSandboxData )
self.jobReport.setJobParameter( 'OutputSandbox', 'Sandbox uploaded to grid storage', sendFlag = False )
self.jobReport.setJobParameter( 'OutputSandboxLFN',
self.__getLFNfromOutputFile( outputSandboxData )[0], sendFlag = False )
else:
self.log.info( 'Could not get SandboxFileName to attempt upload to Grid storage' )
return S_ERROR( 'Output sandbox upload failed and no file name supplied for failover to Grid storage' )
else:
# Do not overwrite in case of Error
if not self.failedFlag:
self.__report( 'Completed', 'Output Sandbox Uploaded' )
self.log.info( 'Sandbox uploaded successfully' )
return "OK"
def configureFromSection( self, mySection ):
"""
reload from CS
"""
self.log.debug( 'Configuring from %s' % mySection )
self.errorMailAddress = gConfig.getValue( mySection + '/ErrorMailAddress' , self.errorMailAddress )
self.alarmMailAddress = gConfig.getValue( mySection + '/AlarmMailAddress' , self.alarmMailAddress )
self.mailFromAddress = gConfig.getValue( mySection + '/MailFromAddress' , self.mailFromAddress )
# following will do something only when call from reload including SubmitPool as mySection
requestedRunningEndPoints = gConfig.getValue( mySection + '/RunningEndPoints', self.runningEndPoints.keys() )
for runningEndPointName in requestedRunningEndPoints:
self.log.verbose( 'Trying to configure RunningEndPoint:', runningEndPointName )
if runningEndPointName in self.runningEndPoints:
continue
RunningBDEndPointDict = BigDataDB.getRunningEnPointDict( runningEndPointName )
if not RunningBDEndPointDict['OK']:
self.log.error( 'Error in RunningBDEndPointDict: %s' % RunningBDEndPointDict['Message'] )
return RunningBDEndPointDict
self.log.verbose( 'Trying to configure RunningBDEndPointDict:', RunningBDEndPointDict )
RunningBDEndPointDict = RunningBDEndPointDict[ 'Value' ]
for option in ['NameNode', 'Port', 'SiteName', 'BigDataSoftware',
'BigDataSoftwareVersion', 'HighLevelLanguage',
'LimitQueueJobsEndPoint', 'URL', 'PublicIP']:
if option not in RunningBDEndPointDict.keys():
self.log.error( 'Missing option in "%s" EndPoint definition:%s' % ( runningEndPointName, option ) )
continue
self.runningEndPoints[runningEndPointName] = {}
self.runningEndPoints[runningEndPointName]['NameNode'] = RunningBDEndPointDict['NameNode']
self.runningEndPoints[runningEndPointName]['Port'] = int ( RunningBDEndPointDict['Port'] )
self.runningEndPoints[runningEndPointName]['SiteName'] = RunningBDEndPointDict['SiteName']
self.runningEndPoints[runningEndPointName]['BigDataSoftware'] = RunningBDEndPointDict['BigDataSoftware']
self.runningEndPoints[runningEndPointName]['BigDataSoftwareVersion'] = RunningBDEndPointDict['BigDataSoftwareVersion']
self.runningEndPoints[runningEndPointName]['LimitQueueJobsEndPoint'] = int( RunningBDEndPointDict['LimitQueueJobsEndPoint'] )
self.runningEndPoints[runningEndPointName]['URL'] = RunningBDEndPointDict['URL']
self.runningEndPoints[runningEndPointName]['User'] = RunningBDEndPointDict['User']
self.runningEndPoints[runningEndPointName]['PublicIP'] = RunningBDEndPointDict['PublicIP']
self.runningEndPoints[runningEndPointName]['UsePilot'] = RunningBDEndPointDict['UsePilot']
self.runningEndPoints[runningEndPointName]['IsInteractive'] = RunningBDEndPointDict['IsInteractive']
self.runningEndPoints[runningEndPointName]['HighLevelLanguage'] = RunningBDEndPointDict['HighLevelLanguage']
self.runningEndPoints[runningEndPointName]['Requirements'] = RunningBDEndPointDict['Requirements']
self.runningEndPoints[runningEndPointName]['Requirements']['CPUTime'] = int ( self.runningEndPoints[runningEndPointName]['Requirements']['CPUTime'] )
def configureFromSection(self, mySection, endPoint):
"""
get CS for monitoring endpoints
"""
self.log.debug("Configuring from %s" % mySection)
monitoringBDEndPointDict = BigDataDB.getRunningEnPointDict(endPoint)
if not monitoringBDEndPointDict["OK"]:
self.log.error("Error in RunninggBDEndPointDict: %s" % monitoringBDEndPointDict["Message"])
return monitoringBDEndPointDict
self.log.verbose("Trying to configure RunningBDEndPointDict:", monitoringBDEndPointDict)
monitoringBDEndPointDict = monitoringBDEndPointDict["Value"]
for option in [
"NameNode",
"Port",
"SiteName",
"BigDataSoftware",
"BigDataSoftwareVersion",
"HighLevelLanguage",
"LimitQueueJobsEndPoint",
"URL",
"PublicIP",
]:
if option not in monitoringBDEndPointDict.keys():
self.log.error('Missing option in "%s" EndPoint definition:' % endPoint, option)
continue
self.monitoringEndPoints[endPoint] = {}
self.monitoringEndPoints[endPoint]["NameNode"] = monitoringBDEndPointDict["NameNode"]
self.monitoringEndPoints[endPoint]["Port"] = int(monitoringBDEndPointDict["Port"])
self.monitoringEndPoints[endPoint]["SiteName"] = monitoringBDEndPointDict["SiteName"]
self.monitoringEndPoints[endPoint]["BigDataSoftware"] = monitoringBDEndPointDict["BigDataSoftware"]
self.monitoringEndPoints[endPoint]["BigDataSoftwareVersion"] = monitoringBDEndPointDict[
"BigDataSoftwareVersion"
]
self.monitoringEndPoints[endPoint]["LimitQueueJobsEndPoint"] = int(
monitoringBDEndPointDict["LimitQueueJobsEndPoint"]
)
self.monitoringEndPoints[endPoint]["URL"] = monitoringBDEndPointDict["URL"]
self.monitoringEndPoints[endPoint]["User"] = monitoringBDEndPointDict["User"]
self.monitoringEndPoints[endPoint]["PublicIP"] = monitoringBDEndPointDict["PublicIP"]
self.monitoringEndPoints[endPoint]["IsInteractive"] = monitoringBDEndPointDict["IsInteractive"]
self.monitoringEndPoints[endPoint]["HighLevelLanguage"] = monitoringBDEndPointDict["HighLevelLanguage"]
def monitoring( self, loop, parentthread, output ):
self.initialTiming = os.times()
accountingReport = AccountingJob()
accountingReport.setStartTime()
numberJobsFlag = True
numberJobs = 0
numberStartedJobsDict = {}
numberEndingJobsDict = {}
job_pattern = re.compile( 'Job =.*?,' )
job_pattern_2 = re.compile( 'Job =.*?\n' )
jobid = int( re.split( "_", re.split( "/", output )[int( len( re.split( "/", output ) ) - 1 )] )[0] )
cmd = '/bin/chmod 555 ' + self.getinfo
returned = self.commandLaunch( cmd )
while parentthread.isAlive():
time.sleep( loop )
if numberJobsFlag:
cmd = self.getinfo + ' -c step1'
returned = self.commandLaunch( cmd )
self.log.info( 'InteractiveJobMonitorThread:step1:numJobs:', returned )
if returned != None:
if ( returned['Value'][1] != "" ):
if re.split( "=", returned['Value'][1] )[1].strip().isdigit():
numberJobs = int( re.split( "=", returned['Value'][1] )[1] )
if ( numberJobs != 0 ):
numberJobsFlag = False
BigDataDB.setJobStatus( jobid, "Running" )
else:
cmd = self.getinfo + ' -c step2'
returned = self.commandLaunch( cmd )
self.log.info( 'InteractiveJobMonitorThread:step2:startedJobs:', returned )
if returned != "":
if ( returned['Value'][1] != "" ):
startedJobs = job_pattern.findall( returned['Value'][1] )
self.log.info( 'step2:startedJobs:', startedJobs )
cmd = self.getinfo + ' -c step3'
returned = self.commandLaunch( cmd )
self.log.info( 'InteractiveJobMonitorThread:step3:endedJobs:', returned )
if returned != "":
if ( returned['Value'][1] != "" ):
finishedJobs = job_pattern_2.findall( returned['Value'][1] )
self.log.info( 'step3:finishedJobs:', finishedJobs )
if ( len( finishedJobs ) == numberJobs ):
BigDataDB.setJobStatus( jobid, "Done" )
BigDataDB.setHadoopID( jobid, finishedJobs )
self.__updateSandBox( jobid, output )
#Update Accounting
EXECUTION_RESULT = {}
EXECUTION_RESULT['CPU'] = []
finalStat = os.times()
for i in range( len( finalStat ) ):
EXECUTION_RESULT['CPU'].append( finalStat[i] - self.initialTiming[i] )
utime, stime, cutime, cstime, elapsed = EXECUTION_RESULT['CPU']
cpuTime = utime + stime + cutime + cstime
execTime = elapsed
result = jobDB.getJobAttributes( jobid )
getting = result['Value']
acData = {
'User' : getting['Owner'],
'UserGroup' : getting['OwnerGroup'],
'JobGroup' : 'cesga',
'JobType' : 'User',
'JobClass' : 'unknown',
'ProcessingType' : 'unknown',
'FinalMajorStatus' : getting['Status'],
'FinalMinorStatus' : getting['MinorStatus'],
'CPUTime' : cpuTime,
'Site' : getting['Site'],
# Based on the factor to convert raw CPU to Normalized units (based on the CPU Model)
'NormCPUTime' : 0,
'ExecTime' : cpuTime,
'InputDataSize' : 0,
'OutputDataSize' : 0,
'InputDataFiles' : 0,
'OutputDataFiles' : 0,
'DiskSpace' : 0,
'InputSandBoxSize' : 0,
'OutputSandBoxSize' : 0,
'ProcessedEvents' : 0
}
accountingReport.setEndTime()
accountingReport.setValuesFromDict( acData )
result = accountingReport.commit()
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()
def execute(self):
"""Main Agent code:
1.- Query BigDataDB for existing Running, Queue, or Submitted jobs
2.- Ask about the status
3.- Change the status into DB in the case of had changed
"""
self.pendingJobs["Submitted"] = BigDataDB.getBigDataJobsByStatus("Submitted")
self.pendingJobs["Running"] = BigDataDB.getBigDataJobsByStatus("Running")
self.pendingJobs["Unknown"] = BigDataDB.getBigDataJobsByStatus("Unknown")
self.__getMonitoringPools()
self.log.verbose("monitoring pools", self.monitoringEndPoints)
for status in self.pendingJobs:
self.log.verbose("Analizing %s jobs" % status)
JobStatus = 0
if self.pendingJobs[status]["OK"]:
for jobId in self.pendingJobs[status]["Value"]:
self.log.verbose("Analizing job %s" % jobId)
getSoftIdAndSiteName = BigDataDB.getSoftwareJobIDByJobID(jobId[0])
self.log.verbose("Site and SoftID:", getSoftIdAndSiteName)
for runningEndPoint in self.monitoringEndPoints:
if (self.monitoringEndPoints[runningEndPoint]["NameNode"] == getSoftIdAndSiteName[0][1]) and (
getSoftIdAndSiteName[0][0] != ""
):
# Depending on the BigData Software the Query should be different
if self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"] == "hadoop":
if self.monitoringEndPoints[runningEndPoint]["BigDataSoftwareVersion"] == "hdv1":
if (
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"]["HLLName"]
== "none"
):
self.log.info(
"Hadoop V.1 Monitoring submmission command with Hadoop jobID: ",
getSoftIdAndSiteName[0][0],
)
from BigDataDIRAC.WorkloadManagementSystem.Client.HadoopV1Client import (
HadoopV1Client,
)
HadoopV1cli = HadoopV1Client(
self.monitoringEndPoints[runningEndPoint]["User"],
self.monitoringEndPoints[runningEndPoint]["PublicIP"],
self.monitoringEndPoints[runningEndPoint]["Port"],
)
JobStatus = HadoopV1cli.jobStatus(
getSoftIdAndSiteName[0][0],
self.monitoringEndPoints[runningEndPoint]["User"],
self.monitoringEndPoints[runningEndPoint]["PublicIP"],
)
if (JobStatus["OK"] == True) and (
self.monitoringEndPoints[runningEndPoint]["IsInteractive"] == "1"
):
if JobStatus["Value"][1].strip() == "Succeded":
result = HadoopV1cli.newJob(
self.__tmpSandBoxDir, jobId[0], getSoftIdAndSiteName[0][0]
)
if result["OK"] == True:
result = BigDataDB.updateHadoopIDAndJobStatus(
jobId[0], result["Value"]
)
BigDataDB.setJobStatus(jobId[0], "Running")
JobStatus["OK"] = False
else:
self.log.info("New result from new Job", result)
if JobStatus["OK"] == True:
if JobStatus["Value"][1].strip() == "Succeded":
BigDataDB.setJobStatus(jobId[0], "Done")
if self.monitoringEndPoints[runningEndPoint]["IsInteractive"] == "1":
self.__updateInteractiveSandBox(
jobId[0],
self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"],
self.monitoringEndPoints[runningEndPoint][
"BigDataSoftwareVersion"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLName"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLVersion"
],
HadoopV1cli,
)
else:
self.__updateSandBox(
jobId[0],
self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"],
self.monitoringEndPoints[runningEndPoint][
"BigDataSoftwareVersion"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLName"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLVersion"
],
HadoopV1cli,
)
getStatus = HadoopV1cli.jobCompleteStatus(getSoftIdAndSiteName[0][0])
if getStatus["OK"]:
result = self.getJobFinalStatusInfo(getStatus["Value"][1])
if result["OK"]:
self.sendJobAccounting(result["Value"], jobId[0])
if self.cleanDataAfterFinish:
self.__deleteData(jobId[0], HadoopV1cli)
if JobStatus["Value"][1].strip() == "Unknown":
BigDataDB.setJobStatus(jobId[0], "Submitted")
if JobStatus["Value"][1].strip() == "Running":
BigDataDB.setJobStatus(jobId[0], "Running")
if self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"] == "hadoop":
if self.monitoringEndPoints[runningEndPoint]["BigDataSoftwareVersion"] == "hdv2":
if (
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"]["HLLName"]
== "none"
):
self.log.info(
"Hadoop V.2 Monitoring submmission command with Hadoop jobID: ",
getSoftIdAndSiteName[0][0],
)
from BigDataDIRAC.WorkloadManagementSystem.Client.HadoopV2Client import (
HadoopV2Client,
)
HadoopV2cli = HadoopV2Client(
self.monitoringEndPoints[runningEndPoint]["User"],
self.monitoringEndPoints[runningEndPoint]["PublicIP"],
)
JobStatus = HadoopV2cli.jobStatus(
getSoftIdAndSiteName[0][0],
self.monitoringEndPoints[runningEndPoint]["User"],
self.monitoringEndPoints[runningEndPoint]["PublicIP"],
)
if (JobStatus["OK"] == True) and (
self.monitoringEndPoints[runningEndPoint]["IsInteractive"] == "1"
):
if JobStatus["Value"].strip() == "Succeded":
result = HadoopV2cli.newJob(
self.__tmpSandBoxDir, jobId[0], getSoftIdAndSiteName[0][0]
)
if result["OK"] == True:
result = BigDataDB.updateHadoopIDAndJobStatus(
jobId[0], result["Value"]
)
BigDataDB.setJobStatus(jobId[0], "Running")
JobStatus["OK"] = False
else:
self.log.info("New result from new Job", result)
if JobStatus["OK"] == True:
if JobStatus["Value"] == "Succeded":
BigDataDB.setJobStatus(jobId[0], "Done")
if self.monitoringEndPoints[runningEndPoint]["IsInteractive"] == "1":
self.__updateInteractiveSandBox(
jobId[0],
self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"],
self.monitoringEndPoints[runningEndPoint][
"BigDataSoftwareVersion"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLName"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLVersion"
],
HadoopV2cli,
)
else:
self.__updateSandBox(
jobId[0],
self.monitoringEndPoints[runningEndPoint]["BigDataSoftware"],
self.monitoringEndPoints[runningEndPoint][
"BigDataSoftwareVersion"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLName"
],
self.monitoringEndPoints[runningEndPoint]["HighLevelLanguage"][
"HLLVersion"
],
HadoopV2cli,
)
getStatus = HadoopV2cli.jobCompleteStatus(getSoftIdAndSiteName[0][0])
if getStatus["OK"]:
result = self.getJobFinalStatusInfo(getStatus["Value"][1])
if result["OK"]:
self.sendJobAccounting(result["Value"], jobId[0])
# if self.cleanDataAfterFinish:
# self.__deleteData( jobId[0], HadoopV2cli )
if JobStatus["Value"] == "Unknown":
BigDataDB.setJobStatus(jobId[0], "Submitted")
if JobStatus["Value"] == "Running":
BigDataDB.setJobStatus(jobId[0], "Running")
return DIRAC.S_OK()
def __updateSandBox(self, jobid, software, version, hll, hllversion, cli):
jobInfo = BigDataDB.getJobIDInfo(jobid)
source = (
self.__tmpSandBoxDir
+ str(jobid)
+ "/InputSandbox"
+ str(jobid)
+ "/"
+ self.__getJobName(jobInfo[0][0]).replace(" ", "")
+ "_"
+ str(jobid)
)
dest = (
self.__tmpSandBoxDir
+ str(jobid)
+ "/"
+ self.__getJobName(jobInfo[0][0]).replace(" ", "")
+ "_"
+ str(jobid)
)
result = 0
if (software == "hadoop") and (version == "hdv1") and (hll == "none"):
result = cli.getData(source, dest)
if (software == "hadoop") and (version == "hdv2") and (hll == "none"):
result = cli.getData(source, dest)
if not result["OK"]:
self.log.error("Error to get the data from BigData Software DFS:", result)
result = cli.getdata(dest, dest)
if not result["OK"]:
self.log.error("Error to get the data from BigData Cluster to DIRAC:", result)
outputSandbox = self.get_filepaths(dest)
resolvedSandbox = self.__resolveOutputSandboxFiles(outputSandbox)
if not resolvedSandbox["OK"]:
self.log.warn("Output sandbox file resolution failed:")
self.log.warn(resolvedSandbox["Message"])
self.__report("Failed", "Resolving Output Sandbox")
self.fileList = resolvedSandbox["Value"]["Files"]
missingFiles = resolvedSandbox["Value"]["Missing"]
if missingFiles:
self.jobReport.setJobParameter("OutputSandboxMissingFiles", ", ".join(missingFiles), sendFlag=False)
if self.fileList and jobid:
self.outputSandboxSize = getGlobbedTotalSize(self.fileList)
self.log.info("Attempting to upload Sandbox with limit:", self.sandboxSizeLimit)
result = self.sandboxClient.uploadFilesAsSandboxForJob(
self.fileList, jobid, "Output", self.sandboxSizeLimit
) # 1024*1024*10
if not result["OK"]:
self.log.error("Output sandbox upload failed with message", result["Message"])
if result.has_key("SandboxFileName"):
outputSandboxData = result["SandboxFileName"]
self.log.info("Attempting to upload %s as output data" % (outputSandboxData))
outputData.append(outputSandboxData)
self.jobReport.setJobParameter("OutputSandbox", "Sandbox uploaded to grid storage", sendFlag=False)
self.jobReport.setJobParameter(
"OutputSandboxLFN", self.__getLFNfromOutputFile(outputSandboxData)[0], sendFlag=False
)
else:
self.log.info("Could not get SandboxFileName to attempt upload to Grid storage")
return S_ERROR(
"Output sandbox upload failed and no file name supplied for failover to Grid storage"
)
else:
# Do not overwrite in case of Error
if not self.failedFlag:
self.__report("Completed", "Output Sandbox Uploaded")
self.log.info("Sandbox uploaded successfully")
return "OK"
def __updateInteractiveSandBox(self, jobid, software, version, hll, hllversion, cli):
# Detele content of InputSandbox
jobInfo = BigDataDB.getJobIDInfo(jobid)
source = self.__tmpSandBoxDir + str(jobid) + "/*_out"
dest = self.__tmpSandBoxDir + str(jobid)
result = 0
result = cli.delHadoopData(self.__tmpSandBoxDir + str(jobid) + "/InputSandbox" + str(jobid))
self.log.debug("ATENTION::Deleting InputSandBox Contain:", result)
result = cli.getdata(dest, source)
self.log.debug("Step 0:getting data from hadoop:", result)
if not result["OK"]:
self.log.error("Error to get the data from BigData Cluster to DIRAC:", result)
self.log.debug("Step:1:GetFilePaths:")
outputSandbox = self.get_filepaths(self.__tmpSandBoxDir + str(jobid))
self.log.debug("Step:2:OutputSandBox:", self.__tmpSandBoxDir + str(jobid))
self.log.debug("Step:2:OutputSandBox:", outputSandbox)
resolvedSandbox = self.__resolveOutputSandboxFiles(outputSandbox)
self.log.debug("Step:3:ResolveSandbox:", resolvedSandbox)
if not resolvedSandbox["OK"]:
self.log.warn("Output sandbox file resolution failed:")
self.log.warn(resolvedSandbox["Message"])
self.__report("Failed", "Resolving Output Sandbox")
self.fileList = resolvedSandbox["Value"]["Files"]
missingFiles = resolvedSandbox["Value"]["Missing"]
if missingFiles:
self.jobReport.setJobParameter("OutputSandboxMissingFiles", ", ".join(missingFiles), sendFlag=False)
if self.fileList and jobid:
self.outputSandboxSize = getGlobbedTotalSize(self.fileList)
self.log.info("Attempting to upload Sandbox with limit:", self.sandboxSizeLimit)
result = self.sandboxClient.uploadFilesAsSandboxForJob(
self.fileList, jobid, "Output", self.sandboxSizeLimit
) # 1024*1024*10
if not result["OK"]:
self.log.error("Output sandbox upload failed with message", result["Message"])
if result.has_key("SandboxFileName"):
outputSandboxData = result["SandboxFileName"]
self.log.info("Attempting to upload %s as output data" % (outputSandboxData))
outputData.append(outputSandboxData)
self.jobReport.setJobParameter("OutputSandbox", "Sandbox uploaded to grid storage", sendFlag=False)
self.jobReport.setJobParameter(
"OutputSandboxLFN", self.__getLFNfromOutputFile(outputSandboxData)[0], sendFlag=False
)
else:
self.log.info("Could not get SandboxFileName to attempt upload to Grid storage")
return S_ERROR(
"Output sandbox upload failed and no file name supplied for failover to Grid storage"
)
else:
# Do not overwrite in case of Error
if not self.failedFlag:
self.__report("Completed", "Output Sandbox Uploaded")
self.log.info("Sandbox uploaded successfully")
return "OK"
