def insertJobInQueue( self, job, classAdJob ):
""" Check individual job and add to the Task Queue eventually.
"""
jobReq = classAdJob.get_expression( "JobRequirements" )
classAdJobReq = ClassAd( jobReq )
jobReqDict = {}
for name in self.taskQueueDB.getSingleValueTQDefFields():
if classAdJobReq.lookupAttribute( name ):
if name == 'CPUTime':
jobReqDict[name] = classAdJobReq.getAttributeInt( name )
else:
jobReqDict[name] = classAdJobReq.getAttributeString( name )
for name in self.taskQueueDB.getMultiValueTQDefFields():
if classAdJobReq.lookupAttribute( name ):
jobReqDict[name] = classAdJobReq.getListFromExpression( name )
jobPriority = classAdJobReq.getAttributeInt( 'UserPriority' )
result = self.taskQueueDB.insertJob( job, jobReqDict, jobPriority )
if not result[ 'OK' ]:
self.log.error( "Cannot insert job %s in task queue: %s" % ( job, result[ 'Message' ] ) )
# Force removing the job from the TQ if it was actually inserted
result = self.taskQueueDB.deleteJob( job )
if result['OK']:
if result['Value']:
self.log.info( "Job %s removed from the TQ" % job )
return S_ERROR( "Cannot insert in task queue" )
return S_OK()
def insertJobInQueue(self, job, classAdJob):
""" Check individual job and add to the Task Queue eventually.
"""
jobReq = classAdJob.get_expression("JobRequirements")
classAdJobReq = ClassAd(jobReq)
jobReqDict = {}
for name in self.taskQueueDB.getSingleValueTQDefFields():
if classAdJobReq.lookupAttribute(name):
if name == 'CPUTime':
jobReqDict[name] = classAdJobReq.getAttributeInt(name)
else:
jobReqDict[name] = classAdJobReq.getAttributeString(name)
for name in self.taskQueueDB.getMultiValueTQDefFields():
if classAdJobReq.lookupAttribute(name):
jobReqDict[name] = classAdJobReq.getListFromExpression(name)
jobPriority = classAdJobReq.getAttributeInt('UserPriority')
result = self.taskQueueDB.insertJob(job, jobReqDict, jobPriority)
if not result['OK']:
self.log.error("Cannot insert job %s in task queue: %s" %
(job, result['Message']))
# Force removing the job from the TQ if it was actually inserted
result = self.taskQueueDB.deleteJob(job)
if result['OK']:
if result['Value']:
self.log.info("Job %s removed from the TQ" % job)
return S_ERROR("Cannot insert in task queue")
return S_OK()
def __processResourceDescription(self, resourceDescription):
# Check and form the resource description dictionary
resourceDict = {}
if type(resourceDescription) in StringTypes:
classAdAgent = ClassAd(resourceDescription)
if not classAdAgent.isOK():
return S_ERROR('Illegal Resource JDL')
gLogger.verbose(classAdAgent.asJDL())
for name in gTaskQueueDB.getSingleValueTQDefFields():
if classAdAgent.lookupAttribute(name):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt(name)
else:
resourceDict[name] = classAdAgent.getAttributeString(
name)
for name in gTaskQueueDB.getMultiValueMatchFields():
if classAdAgent.lookupAttribute(name):
if name == 'SubmitPool':
resourceDict[
name] = classAdAgent.getListFromExpression(name)
else:
resourceDict[name] = classAdAgent.getAttributeString(
name)
# Check if a JobID is requested
if classAdAgent.lookupAttribute('JobID'):
resourceDict['JobID'] = classAdAgent.getAttributeInt('JobID')
for k in ('DIRACVersion', 'ReleaseVersion', 'ReleaseProject',
'VirtualOrganization'):
if classAdAgent.lookupAttribute(k):
resourceDict[k] = classAdAgent.getAttributeString(k)
else:
for name in gTaskQueueDB.getSingleValueTQDefFields():
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
for name in gTaskQueueDB.getMultiValueMatchFields():
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key('JobID'):
resourceDict['JobID'] = resourceDescription['JobID']
for k in ('DIRACVersion', 'ReleaseVersion', 'ReleaseProject',
'VirtualOrganization', 'PilotReference',
'PilotInfoReportedFlag', 'PilotBenchmark',
'LHCbPlatform'):
if k in resourceDescription:
resourceDict[k] = resourceDescription[k]
return resourceDict
def _processResourceDescription( self, resourceDescription ):
""" Check and form the resource description dictionary
resourceDescription is a ceDict coming from a JobAgent, for example.
"""
resourceDict = {}
if isinstance( resourceDescription, basestring ):
classAdAgent = ClassAd( resourceDescription )
if not classAdAgent.isOK():
raise ValueError( 'Illegal Resource JDL' )
self.log.verbose( classAdAgent.asJDL() )
for name in singleValueDefFields:
if classAdAgent.lookupAttribute( name ):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
for name in multiValueMatchFields:
if classAdAgent.lookupAttribute( name ):
if name == 'SubmitPool':
resourceDict[name] = classAdAgent.getListFromExpression( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization' ):
if classAdAgent.lookupAttribute( k ):
resourceDict[ k ] = classAdAgent.getAttributeString( k )
else:
for name in singleValueDefFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
for name in multiValueMatchFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key( 'JobID' ):
resourceDict['JobID'] = resourceDescription['JobID']
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization',
'PilotReference', 'PilotBenchmark', 'PilotInfoReportedFlag' ):
if k in resourceDescription:
resourceDict[ k ] = resourceDescription[ k ]
return resourceDict
def _processResourceDescription( self, resourceDescription ):
""" Check and form the resource description dictionary
resourceDescription is a ceDict coming from a JobAgent, for example.
"""
resourceDict = {}
if type( resourceDescription ) in StringTypes:
classAdAgent = ClassAd( resourceDescription )
if not classAdAgent.isOK():
raise ValueError( 'Illegal Resource JDL' )
self.log.verbose( classAdAgent.asJDL() )
for name in singleValueDefFields:
if classAdAgent.lookupAttribute( name ):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
for name in multiValueMatchFields:
if classAdAgent.lookupAttribute( name ):
if name == 'SubmitPool':
resourceDict[name] = classAdAgent.getListFromExpression( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization' ):
if classAdAgent.lookupAttribute( k ):
resourceDict[ k ] = classAdAgent.getAttributeString( k )
else:
for name in singleValueDefFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
for name in multiValueMatchFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key( 'JobID' ):
resourceDict['JobID'] = resourceDescription['JobID']
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization',
'PilotReference', 'PilotBenchmark', 'PilotInfoReportedFlag' ):
if k in resourceDescription:
resourceDict[ k ] = resourceDescription[ k ]
return resourceDict
def __processResourceDescription( self, resourceDescription ):
# Check and form the resource description dictionary
resourceDict = {}
if type( resourceDescription ) in StringTypes:
classAdAgent = ClassAd( resourceDescription )
if not classAdAgent.isOK():
return S_ERROR( 'Illegal Resource JDL' )
gLogger.verbose( classAdAgent.asJDL() )
for name in gTaskQueueDB.getSingleValueTQDefFields():
if classAdAgent.lookupAttribute( name ):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
for name in gTaskQueueDB.getMultiValueMatchFields():
if classAdAgent.lookupAttribute( name ):
if name == 'SubmitPool':
resourceDict[name] = classAdAgent.getListFromExpression( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization' ):
if classAdAgent.lookupAttribute( k ):
resourceDict[ k ] = classAdAgent.getAttributeString( k )
else:
for name in gTaskQueueDB.getSingleValueTQDefFields():
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
for name in gTaskQueueDB.getMultiValueMatchFields():
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key( 'JobID' ):
resourceDict['JobID'] = resourceDescription['JobID']
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization',
'PilotReference', 'PilotInfoReportedFlag', 'PilotBenchmark' ):
if k in resourceDescription:
resourceDict[ k ] = resourceDescription[ k ]
return resourceDict
def export_submitJob( self, jobDesc ):
""" Submit a single job to DIRAC WMS
"""
if self.peerUsesLimitedProxy:
return S_ERROR( "Can't submit using a limited proxy! (bad boy!)" )
# Check job submission permission
result = self.jobPolicy.getJobPolicy()
if not result['OK']:
return S_ERROR( 'Failed to get job policies' )
policyDict = result['Value']
if not policyDict[ RIGHT_SUBMIT ]:
return S_ERROR( 'Job submission not authorized' )
#jobDesc is JDL for now
jobDesc = jobDesc.strip()
if jobDesc[0] != "[":
jobDesc = "[%s" % jobDesc
if jobDesc[-1] != "]":
jobDesc = "%s]" % jobDesc
# Check if the job is a parameteric one
jobClassAd = ClassAd( jobDesc )
parametricJob = False
if jobClassAd.lookupAttribute( 'Parameters' ):
parametricJob = True
if jobClassAd.isAttributeList( 'Parameters' ):
parameterList = jobClassAd.getListFromExpression( 'Parameters' )
else:
pStep = 0
pFactor = 1
pStart = 1
nParameters = jobClassAd.getAttributeInt( 'Parameters' )
if not nParameters:
value = jobClassAd.get_expression( 'Parameters' )
return S_ERROR( 'Illegal value for Parameters JDL field: %s' % value )
if jobClassAd.lookupAttribute( 'ParameterStart' ):
value = jobClassAd.get_expression( 'ParameterStart' ).replace( '"', '' )
try:
pStart = int( value )
except:
try:
pStart = float( value )
except:
return S_ERROR( 'Illegal value for ParameterStart JDL field: %s' % value )
if jobClassAd.lookupAttribute( 'ParameterStep' ):
pStep = jobClassAd.getAttributeInt( 'ParameterStep' )
if not pStep:
pStep = jobClassAd.getAttributeFloat( 'ParameterStep' )
if not pStep:
value = jobClassAd.get_expression( 'ParameterStep' )
return S_ERROR( 'Illegal value for ParameterStep JDL field: %s' % value )
if jobClassAd.lookupAttribute( 'ParameterFactor' ):
pFactor = jobClassAd.getAttributeInt( 'ParameterFactor' )
if not pFactor:
pFactor = jobClassAd.getAttributeFloat( 'ParameterFactor' )
if not pFactor:
value = jobClassAd.get_expression( 'ParameterFactor' )
return S_ERROR( 'Illegal value for ParameterFactor JDL field: %s' % value )
parameterList = list()
parameterList.append( pStart )
for i in range( nParameters - 1 ):
parameterList.append( parameterList[i] * pFactor + pStep )
if len( parameterList ) > self.maxParametricJobs:
return S_ERROR( 'The number of parametric jobs exceeded the limit of %d' % self.maxParametricJobs )
jobDescList = []
nParam = len(parameterList) - 1
for n,p in enumerate(parameterList):
newJobDesc = jobDesc.replace('%s',str(p)).replace('%n',str(n).zfill(len(str(nParam))))
newClassAd = ClassAd(newJobDesc)
for attr in ['Parameters','ParameterStep','ParameterFactor']:
newClassAd.deleteAttribute(attr)
if type( p ) == type ( ' ' ) and p.startswith('{'):
newClassAd.insertAttributeInt( 'Parameter',str(p) )
else:
newClassAd.insertAttributeString( 'Parameter', str( p ) )
newClassAd.insertAttributeInt( 'ParameterNumber', n )
newJDL = newClassAd.asJDL()
jobDescList.append( newJDL )
else:
jobDescList = [ jobDesc ]
jobIDList = []
for jobDescription in jobDescList:
result = gJobDB.insertNewJobIntoDB( jobDescription, self.owner, self.ownerDN, self.ownerGroup, self.diracSetup )
if not result['OK']:
return result
jobID = result['JobID']
gLogger.info( 'Job %s added to the JobDB for %s/%s' % ( jobID, self.ownerDN, self.ownerGroup ) )
gJobLoggingDB.addLoggingRecord( jobID, result['Status'], result['MinorStatus'], source = 'JobManager' )
jobIDList.append( jobID )
#Set persistency flag
retVal = gProxyManager.getUserPersistence( self.ownerDN, self.ownerGroup )
if 'Value' not in retVal or not retVal[ 'Value' ]:
gProxyManager.setPersistency( self.ownerDN, self.ownerGroup, True )
if parametricJob:
result = S_OK( jobIDList )
else:
result = S_OK( jobIDList[0] )
result['JobID'] = result['Value']
result[ 'requireProxyUpload' ] = self.__checkIfProxyUploadIsRequired()
self.__sendNewJobsToMind( jobIDList )
return result
def export_submitJob(self, jobDesc):
""" Submit a single job to DIRAC WMS
"""
if self.peerUsesLimitedProxy:
return S_ERROR("Can't submit using a limited proxy! (bad boy!)")
# Check job submission permission
result = self.jobPolicy.getJobPolicy()
if not result['OK']:
return S_ERROR('Failed to get job policies')
policyDict = result['Value']
if not policyDict[RIGHT_SUBMIT]:
return S_ERROR('Job submission not authorized')
#jobDesc is JDL for now
jobDesc = jobDesc.strip()
if jobDesc[0] != "[":
jobDesc = "[%s" % jobDesc
if jobDesc[-1] != "]":
jobDesc = "%s]" % jobDesc
# Check if the job is a parameteric one
jobClassAd = ClassAd(jobDesc)
parametricJob = False
if jobClassAd.lookupAttribute('Parameters'):
parametricJob = True
if jobClassAd.isAttributeList('Parameters'):
parameterList = jobClassAd.getListFromExpression('Parameters')
else:
pStep = 0
pFactor = 1
pStart = 1
nParameters = jobClassAd.getAttributeInt('Parameters')
if not nParameters:
value = jobClassAd.get_expression('Parameters')
return S_ERROR(
'Illegal value for Parameters JDL field: %s' % value)
if jobClassAd.lookupAttribute('ParameterStart'):
value = jobClassAd.get_expression(
'ParameterStart').replace('"', '')
try:
pStart = int(value)
except:
try:
pStart = float(value)
except:
return S_ERROR(
'Illegal value for ParameterStart JDL field: %s'
% value)
if jobClassAd.lookupAttribute('ParameterStep'):
pStep = jobClassAd.getAttributeInt('ParameterStep')
if not pStep:
pStep = jobClassAd.getAttributeFloat('ParameterStep')
if not pStep:
value = jobClassAd.get_expression('ParameterStep')
return S_ERROR(
'Illegal value for ParameterStep JDL field: %s'
% value)
if jobClassAd.lookupAttribute('ParameterFactor'):
pFactor = jobClassAd.getAttributeInt('ParameterFactor')
if not pFactor:
pFactor = jobClassAd.getAttributeFloat(
'ParameterFactor')
if not pFactor:
value = jobClassAd.get_expression(
'ParameterFactor')
return S_ERROR(
'Illegal value for ParameterFactor JDL field: %s'
% value)
parameterList = list()
parameterList.append(pStart)
for i in range(nParameters - 1):
parameterList.append(parameterList[i] * pFactor + pStep)
if len(parameterList) > self.maxParametricJobs:
return S_ERROR(
'The number of parametric jobs exceeded the limit of %d' %
self.maxParametricJobs)
jobDescList = []
nParam = len(parameterList) - 1
for n, p in enumerate(parameterList):
newJobDesc = jobDesc.replace('%s', str(p)).replace(
'%n',
str(n).zfill(len(str(nParam))))
newClassAd = ClassAd(newJobDesc)
for attr in ['Parameters', 'ParameterStep', 'ParameterFactor']:
newClassAd.deleteAttribute(attr)
if type(p) == type(' ') and p.startswith('{'):
newClassAd.insertAttributeInt('Parameter', str(p))
else:
newClassAd.insertAttributeString('Parameter', str(p))
newClassAd.insertAttributeInt('ParameterNumber', n)
newJDL = newClassAd.asJDL()
jobDescList.append(newJDL)
else:
jobDescList = [jobDesc]
jobIDList = []
for jobDescription in jobDescList:
result = gJobDB.insertNewJobIntoDB(jobDescription, self.owner,
self.ownerDN, self.ownerGroup,
self.diracSetup)
if not result['OK']:
return result
jobID = result['JobID']
gLogger.info('Job %s added to the JobDB for %s/%s' %
(jobID, self.ownerDN, self.ownerGroup))
gJobLoggingDB.addLoggingRecord(jobID,
result['Status'],
result['MinorStatus'],
source='JobManager')
jobIDList.append(jobID)
#Set persistency flag
retVal = gProxyManager.getUserPersistence(self.ownerDN,
self.ownerGroup)
if 'Value' not in retVal or not retVal['Value']:
gProxyManager.setPersistency(self.ownerDN, self.ownerGroup, True)
if parametricJob:
result = S_OK(jobIDList)
else:
result = S_OK(jobIDList[0])
result['JobID'] = result['Value']
result['requireProxyUpload'] = self.__checkIfProxyUploadIsRequired()
self.__sendNewJobsToMind(jobIDList)
return result
def matchQueue(jobJDL, queueDict, fullMatch=False):
"""
Match the job description to the queue definition
:param str job: JDL job description
:param bool fullMatch: test matching on all the criteria
:param dict queueDict: queue parameters dictionary
:return: S_OK/S_ERROR, Value - result of matching, S_OK if matched or
S_ERROR with the reason for no match
"""
# Check the job description validity
job = ClassAd(jobJDL)
if not job.isOK():
return S_ERROR("Invalid job description")
noMatchReasons = []
# Check job requirements to resource
# 1. CPUTime
cpuTime = job.getAttributeInt("CPUTime")
if not cpuTime:
cpuTime = 84600
if cpuTime > int(queueDict.get("CPUTime", 0)):
noMatchReasons.append("Job CPUTime requirement not satisfied")
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 2. Multi-value match requirements
for parameter in ["Site", "GridCE", "Platform", "JobType"]:
if parameter in queueDict:
valueSet = set(job.getListFromExpression(parameter))
if not valueSet:
valueSet = set(job.getListFromExpression("%ss" % parameter))
queueSet = set(fromChar(queueDict[parameter]))
if valueSet and queueSet and not valueSet.intersection(queueSet):
valueToPrint = ",".join(valueSet)
if len(valueToPrint) > 20:
valueToPrint = "%s..." % valueToPrint[:20]
noMatchReasons.append("Job %s %s requirement not satisfied" % (parameter, valueToPrint))
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 3. Banned multi-value match requirements
for par in ["Site", "GridCE", "Platform", "JobType"]:
parameter = "Banned%s" % par
if par in queueDict:
valueSet = set(job.getListFromExpression(parameter))
if not valueSet:
valueSet = set(job.getListFromExpression("%ss" % parameter))
queueSet = set(fromChar(queueDict[par]))
if valueSet and queueSet and valueSet.issubset(queueSet):
valueToPrint = ",".join(valueSet)
if len(valueToPrint) > 20:
valueToPrint = "%s..." % valueToPrint[:20]
noMatchReasons.append("Job %s %s requirement not satisfied" % (parameter, valueToPrint))
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 4. Tags
tags = set(job.getListFromExpression("Tag"))
nProc = job.getAttributeInt("NumberOfProcessors")
if nProc and nProc > 1:
tags.add("MultiProcessor")
wholeNode = job.getAttributeString("WholeNode")
if wholeNode:
tags.add("WholeNode")
queueTags = set(queueDict.get("Tag", []))
if not tags.issubset(queueTags):
noMatchReasons.append("Job Tag %s not satisfied" % ",".join(tags))
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 4. MultiProcessor requirements
if nProc and nProc > int(queueDict.get("NumberOfProcessors", 1)):
noMatchReasons.append("Job NumberOfProcessors %d requirement not satisfied" % nProc)
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 5. RAM
ram = job.getAttributeInt("RAM")
# If MaxRAM is not specified in the queue description, assume 2GB
if ram and ram > int(queueDict.get("MaxRAM", 2048) / 1024):
noMatchReasons.append("Job RAM %d requirement not satisfied" % ram)
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# Check resource requirements to job
# 1. OwnerGroup - rare case but still
if "OwnerGroup" in queueDict:
result = getProxyInfo(disableVOMS=True)
if not result["OK"]:
return S_ERROR("No valid proxy available")
ownerGroup = result["Value"]["group"]
if ownerGroup != queueDict["OwnerGroup"]:
noMatchReasons.append("Resource OwnerGroup %s requirement not satisfied" % queueDict["OwnerGroup"])
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 2. Required tags
requiredTags = set(queueDict.get("RequiredTags", []))
if not requiredTags.issubset(tags):
noMatchReasons.append("Resource RequiredTags %s not satisfied" % ",".join(requiredTags))
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
# 3. RunningLimit
site = queueDict["Site"]
ce = queueDict.get("GridCE")
opsHelper = Operations()
result = opsHelper.getSections("JobScheduling/RunningLimit")
if result["OK"] and site in result["Value"]:
result = opsHelper.getSections("JobScheduling/RunningLimit/%s" % site)
if result["OK"]:
for parameter in result["Value"]:
value = job.getAttributeString(parameter)
if (
value
and (
opsHelper.getValue("JobScheduling/RunningLimit/%s/%s/%s" % (site, parameter, value), 1)
or opsHelper.getValue(
"JobScheduling/RunningLimit/%s/CEs/%s/%s/%s" % (site, ce, parameter, value), 1
)
)
== 0
):
noMatchReasons.append("Resource operational %s requirement not satisfied" % parameter)
if not fullMatch:
return S_OK({"Match": False, "Reason": noMatchReasons[0]})
return S_OK({"Match": not bool(noMatchReasons), "Reason": noMatchReasons})
def _processResourceDescription( self, resourceDescription ):
""" Check and form the resource description dictionary
resourceDescription is a ceDict coming from a JobAgent, for example.
"""
resourceDict = {}
if isinstance( resourceDescription, basestring ):
classAdAgent = ClassAd( resourceDescription )
if not classAdAgent.isOK():
raise ValueError( 'Illegal Resource JDL' )
self.log.verbose( classAdAgent.asJDL() )
for name in singleValueDefFields:
if classAdAgent.lookupAttribute( name ):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
for name in multiValueMatchFields:
if classAdAgent.lookupAttribute( name ):
if name == 'SubmitPool':
resourceDict[name] = classAdAgent.getListFromExpression( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization' ):
if classAdAgent.lookupAttribute( k ):
resourceDict[ k ] = classAdAgent.getAttributeString( k )
else:
for name in singleValueDefFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
for name in multiValueMatchFields:
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
if 'JobID' in resourceDescription:
resourceDict['JobID'] = resourceDescription['JobID']
# Convert MaxRAM and NumberOfCores parameters into a list of tags
maxRAM = resourceDescription.get( 'MaxRAM' )
nCores = resourceDescription.get( 'NumberOfProcessors' )
for param, key in [ ( maxRAM, 'GB' ), ( nCores, 'Cores' ) ]:
if param:
try:
intValue = int( param )/1000
if intValue <= 128:
paramList = range( 1, intValue + 1 )
paramTags = [ '%d%s' % ( par, key ) for par in paramList ]
resourceDict.setdefault( "Tag", [] ).extend( paramTags )
except ValueError:
pass
if 'Tag' in resourceDict:
resourceDict['Tag'] = list( set( resourceDict['Tag'] ) )
for k in ( 'DIRACVersion', 'ReleaseVersion', 'ReleaseProject', 'VirtualOrganization',
'PilotReference', 'PilotBenchmark', 'PilotInfoReportedFlag' ):
if k in resourceDescription:
resourceDict[ k ] = resourceDescription[ k ]
return resourceDict
def matchQueue(jobJDL, queueDict, fullMatch=False):
"""
Match the job description to the queue definition
:param str job: JDL job description
:param bool fullMatch: test matching on all the criteria
:param dict queueDict: queue parameters dictionary
:return: S_OK/S_ERROR, Value - result of matching, S_OK if matched or
S_ERROR with the reason for no match
"""
# Check the job description validity
job = ClassAd(jobJDL)
if not job.isOK():
return S_ERROR('Invalid job description')
noMatchReasons = []
# Check job requirements to resource
# 1. CPUTime
cpuTime = job.getAttributeInt('CPUTime')
if not cpuTime:
cpuTime = 84600
if cpuTime > queueDict.get('CPUTime', 0.):
noMatchReasons.append('Job CPUTime requirement not satisfied')
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 2. Multi-value match requirements
for parameter in [
'Site', 'GridCE', 'Platform', 'GridMiddleware', 'PilotType',
'SubmitPool', 'JobType'
]:
if parameter in queueDict:
valueSet = set(job.getListFromExpression(parameter))
if not valueSet:
valueSet = set(job.getListFromExpression('%ss' % parameter))
queueSet = set(fromChar(queueDict[parameter]))
if valueSet and queueSet and not valueSet.intersection(queueSet):
valueToPrint = ','.join(valueSet)
if len(valueToPrint) > 20:
valueToPrint = "%s..." % valueToPrint[:20]
noMatchReasons.append('Job %s %s requirement not satisfied' %
(parameter, valueToPrint))
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 3. Banned multi-value match requirements
for par in [
'Site', 'GridCE', 'Platform', 'GridMiddleware', 'PilotType',
'SubmitPool', 'JobType'
]:
parameter = "Banned%s" % par
if par in queueDict:
valueSet = set(job.getListFromExpression(parameter))
if not valueSet:
valueSet = set(job.getListFromExpression('%ss' % parameter))
queueSet = set(fromChar(queueDict[par]))
if valueSet and queueSet and valueSet.issubset(queueSet):
valueToPrint = ','.join(valueSet)
if len(valueToPrint) > 20:
valueToPrint = "%s..." % valueToPrint[:20]
noMatchReasons.append('Job %s %s requirement not satisfied' %
(parameter, valueToPrint))
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 4. Tags
tags = set(job.getListFromExpression('Tag'))
nProc = job.getAttributeInt('NumberOfProcessors')
if nProc and nProc > 1:
tags.add('MultiProcessor')
wholeNode = job.getAttributeString('WholeNode')
if wholeNode:
tags.add('WholeNode')
queueTags = set(queueDict.get('Tags', []))
if not tags.issubset(queueTags):
noMatchReasons.append('Job Tag %s not satisfied' % ','.join(tags))
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 4. MultiProcessor requirements
if nProc and nProc > int(queueDict.get('NumberOfProcessors', 1)):
noMatchReasons.append(
'Job NumberOfProcessors %d requirement not satisfied' % nProc)
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 5. RAM
ram = job.getAttributeInt('RAM')
# If MaxRAM is not specified in the queue description, assume 2GB
if ram and ram > int(queueDict.get('MaxRAM', 2048)) / 1024:
noMatchReasons.append('Job RAM %d requirement not satisfied' % ram)
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# Check resource requirements to job
# 1. OwnerGroup - rare case but still
if "OwnerGroup" in queueDict:
result = getProxyInfo(disableVOMS=True)
if not result['OK']:
return S_ERROR('No valid proxy available')
ownerGroup = result['Value']['group']
if ownerGroup != queueDict['OwnerGroup']:
noMatchReasons.append(
'Resource OwnerGroup %s requirement not satisfied' %
queueDict['OwnerGroup'])
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 2. Required tags
requiredTags = set(queueDict.get('RequiredTags', []))
if not requiredTags.issubset(tags):
noMatchReasons.append('Resource RequiredTags %s not satisfied' %
','.join(requiredTags))
if not fullMatch:
return S_OK({'Match': False, 'Reason': noMatchReasons[0]})
# 3. RunningLimit
site = queueDict['Site']
opsHelper = Operations()
result = opsHelper.getSections('JobScheduling/RunningLimit')
if result['OK'] and site in result['Value']:
result = opsHelper.getSections('JobScheduling/RunningLimit/%s' % site)
if result['OK']:
for parameter in result['Value']:
value = job.getAttributeString(parameter)
if value and opsHelper.getValue(
'JobScheduling/RunningLimit/%s/%s/%s' %
(site, parameter, value), 1) == 0:
noMatchReasons.append(
'Resource operational %s requirement not satisfied' %
parameter)
if not fullMatch:
return S_OK({
'Match': False,
'Reason': noMatchReasons[0]
})
return S_OK({'Match': not bool(noMatchReasons), 'Reason': noMatchReasons})
def _processResourceDescription(self, resourceDescription):
""" Check and form the resource description dictionary
resourceDescription is a ceDict coming from a JobAgent, for example.
"""
resourceDict = {}
if isinstance(resourceDescription, basestring):
classAdAgent = ClassAd(resourceDescription)
if not classAdAgent.isOK():
raise ValueError('Illegal Resource JDL')
self.log.verbose(classAdAgent.asJDL())
for name in singleValueDefFields:
if classAdAgent.lookupAttribute(name):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt(name)
else:
resourceDict[name] = classAdAgent.getAttributeString(
name)
for name in multiValueMatchFields:
if classAdAgent.lookupAttribute(name):
if name == 'SubmitPool':
resourceDict[
name] = classAdAgent.getListFromExpression(name)
else:
resourceDict[name] = classAdAgent.getAttributeString(
name)
# Check if a JobID is requested
if classAdAgent.lookupAttribute('JobID'):
resourceDict['JobID'] = classAdAgent.getAttributeInt('JobID')
for k in ('DIRACVersion', 'ReleaseVersion', 'ReleaseProject',
'VirtualOrganization'):
if classAdAgent.lookupAttribute(k):
resourceDict[k] = classAdAgent.getAttributeString(k)
else:
for name in singleValueDefFields:
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
for name in multiValueMatchFields:
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
if 'JobID' in resourceDescription:
resourceDict['JobID'] = resourceDescription['JobID']
# Convert MaxRAM and NumberOfCores parameters into a list of tags
maxRAM = resourceDescription.get('MaxRAM')
nCores = resourceDescription.get('NumberOfProcessors')
for param, key in [(maxRAM, 'GB'), (nCores, 'Cores')]:
if param:
try:
intValue = int(param) / 1000
if intValue <= 128:
paramList = range(1, intValue + 1)
paramTags = [
'%d%s' % (par, key) for par in paramList
]
resourceDict.setdefault("Tag",
[]).extend(paramTags)
except ValueError:
pass
if 'Tag' in resourceDict:
resourceDict['Tag'] = list(set(resourceDict['Tag']))
for k in ('DIRACVersion', 'ReleaseVersion', 'ReleaseProject',
'VirtualOrganization', 'PilotReference',
'PilotBenchmark', 'PilotInfoReportedFlag'):
if k in resourceDescription:
resourceDict[k] = resourceDescription[k]
return resourceDict
def export_submitJob( self, jobDesc ):
""" Submit a single job to DIRAC WMS
"""
if self.peerUsesLimitedProxy:
return S_ERROR( "Can't submit using a limited proxy! (bad boy!)" )
# Check job submission permission
result = self.jobPolicy.getJobPolicy()
if not result['OK']:
return S_ERROR( 'Failed to get job policies' )
policyDict = result['Value']
if not policyDict[ RIGHT_SUBMIT ]:
return S_ERROR('Job submission not authorized')
#jobDesc is JDL for now
jobDesc = jobDesc.strip()
if jobDesc[0] != "[":
jobDesc = "[%s" % jobDesc
if jobDesc[-1] != "]":
jobDesc = "%s]" % jobDesc
# Check if the job is a parameteric one
jobClassAd = ClassAd(jobDesc)
parametricJob = False
if jobClassAd.lookupAttribute('Parameters'):
parametricJob = True
if jobClassAd.isAttributeList('Parameters'):
parameterList = jobClassAd.getListFromExpression('Parameters')
else:
nParameters = jobClassAd.getAttributeInt('Parameters')
if not nParameters:
value = jobClassAd.get_expression('Parameters')
return S_ERROR('Illegal value for Parameters JDL field: %s' % value)
if jobClassAd.lookupAttribute('ParameterStart'):
pStart = jobClassAd.getAttributeInt('ParameterStart')
else:
return S_ERROR('Missing JDL field ParameterStart')
if jobClassAd.lookupAttribute('ParameterStep'):
pStep = jobClassAd.getAttributeInt('ParameterStep')
if not pStep:
value = jobClassAd.get_expression('ParameterStep')
return S_ERROR('Illegal value for ParameterStep JDL field: %s' % value)
else:
return S_ERROR('Missing JDL field ParameterStep')
parameterList = list( range(pStart,pStart+pStep*nParameters,pStep) )
if len(parameterList) > MAX_PARAMETRIC_JOBS:
return S_ERROR('The number of parametric jobs exceeded the limit of %d' % MAX_PARAMETRIC_JOBS )
jobDescList = []
for p in parameterList:
jobDescList.append( jobDesc.replace('%s',str(p)) )
else:
jobDescList = [ jobDesc ]
jobIDList = []
for jobDescription in jobDescList:
result = gJobDB.insertNewJobIntoDB( jobDescription, self.owner, self.ownerDN, self.ownerGroup, self.diracSetup )
if not result['OK']:
return result
jobID = result['JobID']
gLogger.info( 'Job %s added to the JobDB for %s/%s' % ( jobID, self.ownerDN, self.ownerGroup ) )
gJobLoggingDB.addLoggingRecord( jobID, result['Status'], result['MinorStatus'], source = 'JobManager' )
jobIDList.append(jobID)
#Set persistency flag
retVal = gProxyManager.getUserPersistence( self.ownerDN, self.ownerGroup )
if 'Value' not in retVal or not retVal[ 'Value' ]:
gProxyManager.setPersistency( self.ownerDN, self.ownerGroup, True )
if parametricJob:
result = S_OK(jobIDList)
else:
result = S_OK(jobIDList[0])
result['JobID'] = result['Value']
result[ 'requireProxyUpload' ] = self.__checkIfProxyUploadIsRequired()
return result
