def test_SimpleParametricJob():
job = Job()
job.setExecutable('myExec')
job.setLogLevel('DEBUG')
parList = [1, 2, 3]
job.setParameterSequence('JOB_ID', parList, addToWorkflow=True)
inputDataList = [['/lhcb/data/data1', '/lhcb/data/data2'],
['/lhcb/data/data3', '/lhcb/data/data4'],
['/lhcb/data/data5', '/lhcb/data/data6']]
job.setParameterSequence('InputData', inputDataList, addToWorkflow=True)
jdl = job._toJDL()
try:
with open('./DIRAC/Interfaces/API/test/testWF.jdl') as fd:
expected = fd.read()
except IOError:
with open('./Interfaces/API/test/testWF.jdl') as fd:
expected = fd.read()
assert jdl == expected
clad = ClassAd('[' + jdl + ']')
arguments = clad.getAttributeString('Arguments')
job_id = clad.getAttributeString('JOB_ID')
inputData = clad.getAttributeString('InputData')
assert job_id == '%(JOB_ID)s'
assert inputData == '%(InputData)s'
assert 'jobDescription.xml' in arguments
assert '-o LogLevel=DEBUG' in arguments
assert '-p JOB_ID=%(JOB_ID)s' in arguments
assert '-p InputData=%(InputData)s' in arguments
def test_SimpleParametricJob(self):
job = Job()
job.setExecutable('myExec')
job.setLogLevel('DEBUG')
parList = [1, 2, 3]
job.setParameterSequence('JOB_ID', parList, addToWorkflow=True)
inputDataList = [['/lhcb/data/data1', '/lhcb/data/data2'],
['/lhcb/data/data3', '/lhcb/data/data4'],
['/lhcb/data/data5', '/lhcb/data/data6']]
job.setParameterSequence('InputData',
inputDataList,
addToWorkflow=True)
jdl = job._toJDL()
print jdl
clad = ClassAd('[' + jdl + ']')
arguments = clad.getAttributeString('Arguments')
job_id = clad.getAttributeString('JOB_ID')
inputData = clad.getAttributeString('InputData')
print "arguments", arguments
self.assertEqual(job_id, '%(JOB_ID)s')
self.assertEqual(inputData, '%(InputData)s')
self.assertIn('jobDescription.xml', arguments)
self.assertIn('-o LogLevel=DEBUG', arguments)
self.assertIn('-p JOB_ID=%(JOB_ID)s', arguments)
self.assertIn('-p InputData=%(InputData)s', arguments)
def test_SimpleParametricJob():
job = Job()
job.setExecutable("myExec")
job.setLogLevel("DEBUG")
parList = [1, 2, 3]
job.setParameterSequence("JOB_ID", parList, addToWorkflow=True)
inputDataList = [
["/lhcb/data/data1", "/lhcb/data/data2"],
["/lhcb/data/data3", "/lhcb/data/data4"],
["/lhcb/data/data5", "/lhcb/data/data6"],
]
job.setParameterSequence("InputData", inputDataList, addToWorkflow=True)
jdl = job._toJDL()
with open(join(dirname(__file__), "testWF.jdl")) as fd:
expected = fd.read()
assert jdl == expected
clad = ClassAd("[" + jdl + "]")
arguments = clad.getAttributeString("Arguments")
job_id = clad.getAttributeString("JOB_ID")
inputData = clad.getAttributeString("InputData")
assert job_id == "%(JOB_ID)s"
assert inputData == "%(InputData)s"
assert "jobDescription.xml" in arguments
assert "-o LogLevel=DEBUG" in arguments
assert "-p JOB_ID=%(JOB_ID)s" in arguments
assert "-p InputData=%(InputData)s" in arguments
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
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):
resourceDict[name] = classAdAgent.getAttributeString(name)
# Check if a JobID is requested
if classAdAgent.lookupAttribute('JobID'):
resourceDict['JobID'] = classAdAgent.getAttributeInt('JobID')
if classAdAgent.lookupAttribute('DIRACVersion'):
resourceDict['DIRACVersion'] = classAdAgent.getAttributeString(
'DIRACVersion')
if classAdAgent.lookupAttribute('VirtualOrganization'):
resourceDict[
'VirtualOrganization'] = classAdAgent.getAttributeString(
'VirtualOrganization')
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']
if resourceDescription.has_key('DIRACVersion'):
resourceDict['DIRACVersion'] = resourceDescription[
'DIRACVersion']
if resourceDescription.has_key('VirtualOrganization'):
resourceDict['VirtualOrganization'] = resourceDescription[
'VirtualOrganization']
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
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
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 __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 ):
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
if classAdAgent.lookupAttribute( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = classAdAgent.getAttributeString( 'DIRACVersion' )
if classAdAgent.lookupAttribute( 'VirtualOrganization' ):
resourceDict['VirtualOrganization'] = classAdAgent.getAttributeString( 'VirtualOrganization' )
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']
if resourceDescription.has_key( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = resourceDescription['DIRACVersion']
if resourceDescription.has_key( 'VirtualOrganization' ):
resourceDict['VirtualOrganization'] = resourceDescription['VirtualOrganization']
return resourceDict
def matchJob(self, resourceJDL):
""" Use Matcher service to retrieve a MPI job from Task Queue.
Returns: JobID, NumProc required, JDL and MPI flavor
Input: resourceJDL
Output: result = {JobID, JobJDL, NumProc, MPIFlavor}
"""
print "S37"
matcher = RPCClient('WorkloadManagement/Matcher', timeout = 600)
dictMatchMPI = {'Setup':'EELA-Production', 'CPUTime':6000, 'JobType':'MPI'}
result = matcher.getMatchingTaskQueues(dictMatchMPI)
if not result['OK']:
print "S38"
gLogger.info("-------------------------------------------------------------------")
gLogger.error ("Here I have to call to get normal job")
gLogger.info("-------------------------------------------------------------------")
gLogger.error (("Match not found: %s") % (result['Message']))
gLogger.info("-------------------------------------------------------------------")
return S_ERROR()
else:
if result['Value'] == {}:
gLogger.info("-------------------------------------------------------------------")
gLogger.info("Value == Empty")
gLogger.info("-------------------------------------------------------------------")
return S_ERROR()
mpiTaskQueue = result['Value']
classAdAgent = ClassAd(resourceJDL)
classAdAgent.insertAttributeString('JobType', 'MPI')
resourceJDL = str(classAdAgent.asJDL())
result = matcher.requestJob(resourceJDL)
if not result['OK']:
gLogger.error (("Request Job Error: %s") % (result['Message']))
return S_ERROR()
elif result['OK']==False:
gLogger.error (("Request Job False: %s") % (result['Message']))
return S_ERROR()
else:
gLogger.error (("Request Job OK"))
jobJDL = result['Value']['JDL']
### Review how to optimize this part (Importante)
jobID1 = ClassAd(jobJDL)
jobID = jobID1.getAttributeString('JobID')
numProc = jobID1.getAttributeString('CPUNumber')
mpiFlavor = jobID1.getAttributeString('Flavor')
result = {'JobID':jobID, 'JobJDL':jobJDL, 'NumProc': numProc,
'MPIFlavor': mpiFlavor}
print "S39"
print result
return S_OK(result)
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 test_SimpleParametricJob():
job = Job()
job.setExecutable('myExec')
job.setLogLevel('DEBUG')
parList = [1, 2, 3]
job.setParameterSequence('JOB_ID', parList, addToWorkflow=True)
inputDataList = [
[
'/lhcb/data/data1',
'/lhcb/data/data2'
],
[
'/lhcb/data/data3',
'/lhcb/data/data4'
],
[
'/lhcb/data/data5',
'/lhcb/data/data6'
]
]
job.setParameterSequence('InputData', inputDataList, addToWorkflow=True)
jdl = job._toJDL()
try:
with open('./DIRAC/Interfaces/API/test/testWF.jdl') as fd:
expected = fd.read()
except IOError:
with open('./Interfaces/API/test/testWF.jdl') as fd:
expected = fd.read()
assert jdl == expected
clad = ClassAd('[' + jdl + ']')
arguments = clad.getAttributeString('Arguments')
job_id = clad.getAttributeString('JOB_ID')
inputData = clad.getAttributeString('InputData')
assert job_id == '%(JOB_ID)s'
assert inputData == '%(InputData)s'
assert 'jobDescription.xml' in arguments
assert '-o LogLevel=DEBUG' in arguments
assert'-p JOB_ID=%(JOB_ID)s' in arguments
assert'-p InputData=%(InputData)s' in arguments
def test_SimpleProgression(self):
clad = ClassAd(TEST_JDL_SIMPLE_PROGRESSION)
nParam = getNumberOfParameters(clad)
self.assertEqual(nParam, 3)
result = generateParametricJobs(clad)
self.assert_(result['OK'])
jobDescList = result['Value']
self.assertEqual(nParam, len(jobDescList))
# Check the definition of the 2nd job
jobClassAd = ClassAd(jobDescList[1])
self.assertEqual(jobClassAd.getAttributeString('Arguments'), '3')
self.assertEqual(jobClassAd.getAttributeString('JobName'), 'Test_1')
def test_SimpleProgression(self):
clad = ClassAd( TEST_JDL_SIMPLE_PROGRESSION )
nParam = getNumberOfParameters( clad )
self.assertEqual( nParam, 3 )
result = generateParametricJobs( clad )
self.assert_( result['OK'] )
jobDescList = result['Value']
self.assertEqual( nParam, len( jobDescList ) )
# Check the definition of the 2nd job
jobClassAd = ClassAd( jobDescList[1] )
self.assertEqual( jobClassAd.getAttributeString( 'Arguments' ), '3' )
self.assertEqual( jobClassAd.getAttributeString( 'JobName' ), 'Test_1' )
def test_SimpleBunch(self):
clad = ClassAd( TEST_JDL_SIMPLE_BUNCH )
result = getParameterVectorLength( clad )
self.assertTrue( result['OK'] )
nParam = result['Value']
self.assertEqual( nParam, 3 )
result = generateParametricJobs( clad )
self.assertTrue(result['OK'])
jobDescList = result['Value']
self.assertEqual( nParam, len( jobDescList ) )
# Check the definition of the 2nd job
jobClassAd = ClassAd( jobDescList[1] )
self.assertEqual( jobClassAd.getAttributeString( 'Arguments' ), '5' )
self.assertEqual( jobClassAd.getAttributeString( 'JobName' ), 'Test_1' )
def test_Simple(self):
clad = ClassAd(TEST_JDL_SIMPLE)
result = getParameterVectorLength(clad)
self.assertTrue(result['OK'])
nParam = result['Value']
self.assertEqual(nParam, 3)
result = generateParametricJobs(clad)
self.assertTrue(result['OK'])
jobDescList = result['Value']
self.assertEqual(nParam, len(jobDescList))
# Check the definition of the 2nd job
jobClassAd = ClassAd(jobDescList[1])
self.assertEqual(jobClassAd.getAttributeString('Arguments'), 'b')
self.assertEqual(jobClassAd.getAttributeString('JobName'), 'Test_1')
def _getProcessingType(self, jobID):
"""Get the Processing Type from the JDL, until it is promoted to a real Attribute"""
processingType = "unknown"
result = self.jobDB.getJobJDL(jobID, original=True)
if not result["OK"]:
return processingType
classAdJob = ClassAd(result["Value"])
if classAdJob.lookupAttribute("ProcessingType"):
processingType = classAdJob.getAttributeString("ProcessingType")
return processingType
def test_SimpleBunch(self):
clad = ClassAd(TEST_JDL_SIMPLE_BUNCH)
result = getParameterVectorLength(clad)
self.assertTrue(result["OK"])
nParam = result["Value"]
self.assertEqual(nParam, 3)
result = generateParametricJobs(clad)
self.assertTrue(result["OK"])
jobDescList = result["Value"]
self.assertEqual(nParam, len(jobDescList))
# Check the definition of the 2nd job
jobClassAd = ClassAd(jobDescList[1])
self.assertEqual(jobClassAd.getAttributeString("Arguments"), "5")
self.assertEqual(jobClassAd.getAttributeString("JobName"), "Test_1")
def __getProcessingType( self, jobID ):
""" Get the Processing Type from the JDL, until it is promoted to a real Attribute
"""
processingType = 'unknown'
result = self.jobDB.getJobJDL( jobID, original = True )
if not result['OK']:
return processingType
classAdJob = ClassAd( result['Value'] )
if classAdJob.lookupAttribute( 'ProcessingType' ):
processingType = classAdJob.getAttributeString( 'ProcessingType' )
return processingType
def __getProcessingType(self, jobID):
""" Get the Processing Type from the JDL, until it is promoted to a real Attribute
"""
processingType = 'unknown'
result = self.jobDB.getJobJDL(jobID, original=True)
if not result['OK']:
return processingType
classAdJob = ClassAd(result['Value'])
if classAdJob.lookupAttribute('ProcessingType'):
processingType = classAdJob.getAttributeString('ProcessingType')
return processingType
def __getProcessingType(self, jobID):
""" Get the Processing Type from the JDL, until it is promoted to a real Attribute
"""
processingType = "unknown"
result = self.jobDB.getJobJDL(jobID, original=True)
if not result["OK"]:
return processingType
classAdJob = ClassAd(result["Value"])
if classAdJob.lookupAttribute("ProcessingType"):
processingType = classAdJob.getAttributeString("ProcessingType")
return processingType
def test_SimpleParametricJob( self ):
job = Job()
job.setExecutable( 'myExec' )
job.setLogLevel( 'DEBUG' )
parList = [1,2,3]
job.setParameterSequence( 'JOB_ID', parList, addToWorkflow=True )
inputDataList = [
[
'/lhcb/data/data1',
'/lhcb/data/data2'
],
[
'/lhcb/data/data3',
'/lhcb/data/data4'
],
[
'/lhcb/data/data5',
'/lhcb/data/data6'
]
]
job.setParameterSequence( 'InputData', inputDataList, addToWorkflow=True )
jdl = job._toJDL()
print jdl
clad = ClassAd( '[' + jdl + ']' )
arguments = clad.getAttributeString( 'Arguments' )
job_id = clad.getAttributeString( 'JOB_ID' )
inputData = clad.getAttributeString( 'InputData' )
print "arguments", arguments
self.assertEqual( job_id, '%(JOB_ID)s' )
self.assertEqual( inputData, '%(InputData)s' )
self.assertIn( 'jobDescription.xml', arguments )
self.assertIn( '-o LogLevel=DEBUG', arguments )
self.assertIn( '-p JOB_ID=%(JOB_ID)s', arguments )
self.assertIn( '-p InputData=%(InputData)s', arguments )
def submitNewBigJob( self ):
result = jobDB.getJobJDL( str( self.__jobID ) , True )
classAdJob = ClassAd( result['Value'] )
executableFile = ""
if classAdJob.lookupAttribute( 'Executable' ):
executableFile = classAdJob.getAttributeString( 'Executable' )
tempPath = self.__tmpSandBoxDir
dirac = Dirac()
if not os.path.exists( tempPath ):
os.makedirs( tempPath )
settingJobSandBoxDir = dirac.getInputSandbox( self.__jobID, tempPath )
self.log.info( 'Writting temporal SandboxDir in Server', settingJobSandBoxDir )
moveData = self.__tmpSandBoxDir + "/InputSandbox" + str( self.__jobID )
HiveV1Cli = HiveV1Client( self.__User , self.__publicIP )
returned = HiveV1Cli.dataCopy( moveData, self.__tmpSandBoxDir )
self.log.info( 'Copy the job contain to the Hadoop Master with HIVE: ', returned )
jobInfo = jobDB.getJobAttributes( self.__jobID )
if not jobInfo['OK']:
return S_ERROR( jobInfo['Value'] )
proxy = ""
jobInfo = jobInfo['Value']
if gProxyManager.userHasProxy( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] ):
proxy = gProxyManager.downloadProxyToFile( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] )
else:
proxy = self.__requestProxyFromProxyManager( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] )
HiveJob = "InputSandbox" + str( self.__jobID ) + "/" + executableFile
HiveJobOutput = str( self.__jobID ) + "_" + executableFile + "_out"
returned = HiveV1Cli.jobSubmit( tempPath, HiveJob, proxy['chain'], HiveJobOutput )
self.log.info( 'Launch Hadoop-Hive job to the Master: ', returned )
if not returned['OK']:
return S_ERROR( returned['Message'] )
else:
self.log.info( 'Hadoop-Hive Job ID: ', returned['Value'] )
return S_OK( returned['Value'] )
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 submitNewBigJob( self ):
#1.- Creamos carpeta temporal
self.log.debug( 'Step1::: mkdir temp folder' )
tempPath = self.__tmpSandBoxDir + str( self.__jobID ) + "/"
dirac = Dirac()
if not os.path.exists( tempPath ):
os.makedirs( tempPath )
#2.- Introducimos el contenido del inputsandbox en la carpeta temporal
self.log.debug( 'Step2::: download inputsand to temp folder' )
settingJobSandBoxDir = dirac.getInputSandbox( self.__jobID, tempPath )
self.log.info( 'Writting temporal SandboxDir in Server', settingJobSandBoxDir )
moveData = tempPath + "/InputSandbox" + str( self.__jobID )
#3.- Move the data to client
self.log.debug( 'Step2::: download inputsandbox to temp folder' )
HadoopV1InteractiveCli = HadoopV1InteractiveClient( self.__User , self.__publicIP, self.__Port )
returned = HadoopV1InteractiveCli.dataCopy( tempPath, self.__tmpSandBoxDir )
self.log.debug( 'Returned of copy the job contain to the Hadoop Master with HadoopInteractive::: ', returned )
#3.- Get executable file
result = jobDB.getJobJDL( str( self.__jobID ) , True )
classAdJob = ClassAd( result['Value'] )
executableFile = ""
if classAdJob.lookupAttribute( 'Executable' ):
executableFile = classAdJob.getAttributeString( 'Executable' )
self.log.debug( 'Step3::: Get executable file: ', executableFile )
jobInfo = jobDB.getJobAttributes( self.__jobID )
if not jobInfo['OK']:
return S_ERROR( jobInfo['Value'] )
proxy = ""
jobInfo = jobInfo['Value']
if gProxyManager.userHasProxy( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] ):
proxy = gProxyManager.downloadProxyToFile( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] )
else:
proxy = self.__requestProxyFromProxyManager( jobInfo["OwnerDN"], jobInfo["OwnerGroup"] )
HadoopInteractiveJob = "InputSandbox" + str( self.__jobID ) + "/" + executableFile
HadoopInteractiveJobCommand = "InputSandbox" + str( self.__jobID ) + "/" + executableFile + " " + self.__JobName
HadoopInteractiveJobOutput = tempPath + str( self.__jobID ) + "_" + executableFile + "_out"
#4.- Creating second part of the job name
if ( len( re.split( " ", self.__JobName ) ) > 1 ):
#(name for random writter with -D)name_job = re.split( " ", self.__JobName )[0] + " " + re.split( " ", self.__JobName )[1] + " " + re.split( " ", self.__JobName )[2]
name_job = re.split( " ", self.__JobName )[0] + " " + re.split( " ", self.__JobName )[1]
#(name for random writter with -D)output_job = moveData + "/" + re.split( " ", self.__JobName )[3]
#(name for random writter with -D)cfg_job = ""
#(name for random writter with -D)if ( len( re.split( " ", self.__JobName ) ) > 4 ):
#(name for random writter with -D) cfg_job = moveData + "/" + re.split( " ", self.__JobName )[4]
#5.- Parsing execution command
#cmd = "hadoop jar " + tempPath + HadoopInteractiveJob + " " + name_job + " " + output_job + " " + cfg_job
cmd = "hadoop jar " + tempPath + HadoopInteractiveJob + " " + name_job + " " + tempPath + "/InputSandbox" + str( self.__jobID ) + "/" + "/dataset-USC-a-grep '[and]+'"
else:
dataset = re.split( "/", self.__Dataset )
count = 0
datasetname = ""
for dir in dataset:
count = count + 1
if ( count > 2 ):
datasetname = datasetname + "/" + dir
cmd = "hadoop jar " + tempPath + HadoopInteractiveJob + " " + self.__JobName + " " + datasetname + " " + tempPath + "/" + self.__JobName.replace( " ", "" ) + "_" + str( self.__jobID )
self.log.debug( 'Step4::: Making CMD for submission: ', cmd )
self.log.debug( 'Step5::: Submit file to hadoop: ' )
returned = HadoopV1InteractiveCli.jobSubmit( tempPath, HadoopInteractiveJob, proxy['chain'],
HadoopInteractiveJobOutput, cmd )
self.log.info( 'Launch Hadoop-HadoopInteractive job to the Master: ', returned )
if not returned['OK']:
return S_ERROR( returned['Message'] )
else:
self.log.info( 'Hadoop-HadoopInteractive Job ID: ', returned['Value'] )
return S_OK( returned['Value'] )
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 selectJob( self, resourceDescription ):
""" Main job selection function to find the highest priority job
matching the resource capacity
"""
startTime = time.time()
# 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 taskQueueDB.getSingleValueTQDefFields():
if classAdAgent.lookupAttribute( name ):
if name == 'CPUTime':
resourceDict[name] = classAdAgent.getAttributeInt( name )
else:
resourceDict[name] = classAdAgent.getAttributeString( name )
for name in taskQueueDB.getMultiValueMatchFields():
if classAdAgent.lookupAttribute( name ):
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
if classAdAgent.lookupAttribute( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = classAdAgent.getAttributeString( 'DIRACVersion' )
else:
for name in taskQueueDB.getSingleValueTQDefFields():
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
for name in taskQueueDB.getMultiValueMatchFields():
if resourceDescription.has_key( name ):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key( 'JobID' ):
resourceDict['JobID'] = resourceDescription['JobID']
if resourceDescription.has_key( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = resourceDescription['DIRACVersion']
# Check the pilot DIRAC version
if self.checkPilotVersion:
if 'DIRACVersion' in resourceDict:
if self.pilotVersion and resourceDict['DIRACVersion'] != self.pilotVersion:
return S_ERROR( 'Pilot version does not match the production version %s:%s' % \
( resourceDict['DIRACVersion'], self.pilotVersion ) )
# Get common site mask and check the agent site
result = jobDB.getSiteMask( siteState = 'Active' )
if result['OK']:
maskList = result['Value']
else:
return S_ERROR( 'Internal error: can not get site mask' )
if not 'Site' in resourceDict:
return S_ERROR( 'Missing Site Name in Resource JDL' )
siteName = resourceDict['Site']
if resourceDict['Site'] not in maskList:
if 'GridCE' in resourceDict:
del resourceDict['Site']
else:
return S_ERROR( 'Site not in mask and GridCE not specified' )
resourceDict['Setup'] = self.serviceInfoDict['clientSetup']
if DEBUG:
print "Resource description:"
for k, v in resourceDict.items():
print k.rjust( 20 ), v
# Check if Job Limits are imposed onto the site
extraConditions = {}
if self.siteJobLimits:
result = self.getExtraConditions( siteName )
if result['OK']:
extraConditions = result['Value']
if extraConditions:
gLogger.info( 'Job Limits for site %s are: %s' % ( siteName, str( extraConditions ) ) )
result = taskQueueDB.matchAndGetJob( resourceDict, extraConditions = extraConditions )
if DEBUG:
print result
if not result['OK']:
return result
result = result['Value']
if not result['matchFound']:
return S_ERROR( 'No match found' )
jobID = result['jobId']
resAtt = jobDB.getJobAttributes( jobID, ['OwnerDN', 'OwnerGroup', 'Status'] )
if not resAtt['OK']:
return S_ERROR( 'Could not retrieve job attributes' )
if not resAtt['Value']:
return S_ERROR( 'No attributes returned for job' )
if not resAtt['Value']['Status'] == 'Waiting':
gLogger.error( 'Job %s matched by the TQ is not in Waiting state' % str( jobID ) )
result = taskQueueDB.deleteJob( jobID )
result = jobDB.setJobStatus( jobID, status = 'Matched', minor = 'Assigned' )
result = jobLoggingDB.addLoggingRecord( jobID,
status = 'Matched',
minor = 'Assigned',
source = 'Matcher' )
result = jobDB.getJobJDL( jobID )
if not result['OK']:
return S_ERROR( 'Failed to get the job JDL' )
resultDict = {}
resultDict['JDL'] = result['Value']
resultDict['JobID'] = jobID
matchTime = time.time() - startTime
gLogger.info( "Match time: [%s]" % str( matchTime ) )
gMonitor.addMark( "matchTime", matchTime )
# Get some extra stuff into the response returned
resOpt = jobDB.getJobOptParameters( jobID )
if resOpt['OK']:
for key, value in resOpt['Value'].items():
resultDict[key] = value
resAtt = jobDB.getJobAttributes( jobID, ['OwnerDN', 'OwnerGroup'] )
if not resAtt['OK']:
return S_ERROR( 'Could not retrieve job attributes' )
if not resAtt['Value']:
return S_ERROR( 'No attributes returned for job' )
resultDict['DN'] = resAtt['Value']['OwnerDN']
resultDict['Group'] = resAtt['Value']['OwnerGroup']
return S_OK( resultDict )
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 selectJob( self, resourceDescription ):
""" Main job selection function to find the highest priority job
matching the resource capacity
"""
startTime = time.time()
# 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 ):
resourceDict[name] = classAdAgent.getAttributeString( name )
# Check if a JobID is requested
if classAdAgent.lookupAttribute( 'JobID' ):
resourceDict['JobID'] = classAdAgent.getAttributeInt( 'JobID' )
if classAdAgent.lookupAttribute( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = classAdAgent.getAttributeString( 'DIRACVersion' )
if classAdAgent.lookupAttribute( 'VirtualOrganization' ):
resourceDict['VirtualOrganization'] = classAdAgent.getAttributeString( 'VirtualOrganization' )
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']
if resourceDescription.has_key( 'DIRACVersion' ):
resourceDict['DIRACVersion'] = resourceDescription['DIRACVersion']
if resourceDescription.has_key( 'VirtualOrganization' ):
resourceDict['VirtualOrganization'] = resourceDescription['VirtualOrganization']
# Check the pilot DIRAC version
if self.checkPilotVersion:
if not 'DIRACVersion' in resourceDict:
return S_ERROR( 'Version check requested and not provided by Pilot' )
# Check if the matching Request provides a VirtualOrganization
if 'VirtualOrganization' in resourceDict:
voName = resourceDict['VirtualOrganization']
# Check if the matching Request provides an OwnerGroup
elif 'OwnerGroup' in resourceDict:
voName = getVOForGroup( resourceDict['OwnerGroup'] )
# else take the default VirtualOrganization for the installation
else:
voName = getVOForGroup( '' )
self.pilotVersion = gConfig.getValue( '/Operations/%s/%s/Versions/PilotVersion' % ( voName, self.setup ), '' )
if self.pilotVersion and resourceDict['DIRACVersion'] != self.pilotVersion:
return S_ERROR( 'Pilot version does not match the production version %s:%s' % \
( resourceDict['DIRACVersion'], self.pilotVersion ) )
# Get common site mask and check the agent site
result = gJobDB.getSiteMask( siteState = 'Active' )
if result['OK']:
maskList = result['Value']
else:
return S_ERROR( 'Internal error: can not get site mask' )
if not 'Site' in resourceDict:
return S_ERROR( 'Missing Site Name in Resource JDL' )
siteName = resourceDict['Site']
if resourceDict['Site'] not in maskList:
if 'GridCE' in resourceDict:
del resourceDict['Site']
else:
return S_ERROR( 'Site not in mask and GridCE not specified' )
resourceDict['Setup'] = self.serviceInfoDict['clientSetup']
if DEBUG:
print "Resource description:"
for key, value in resourceDict.items():
print key.rjust( 20 ), value
# Check if Job Limits are imposed onto the site
extraConditions = {}
if self.siteJobLimits:
result = self.getExtraConditions( siteName )
if result['OK']:
extraConditions = result['Value']
if extraConditions:
gLogger.info( 'Job Limits for site %s are: %s' % ( siteName, str( extraConditions ) ) )
result = gTaskQueueDB.matchAndGetJob( resourceDict, extraConditions = extraConditions )
if DEBUG:
print result
if not result['OK']:
return result
result = result['Value']
if not result['matchFound']:
return S_ERROR( 'No match found' )
jobID = result['jobId']
resAtt = gJobDB.getJobAttributes( jobID, ['OwnerDN', 'OwnerGroup', 'Status'] )
if not resAtt['OK']:
return S_ERROR( 'Could not retrieve job attributes' )
if not resAtt['Value']:
return S_ERROR( 'No attributes returned for job' )
if not resAtt['Value']['Status'] == 'Waiting':
gLogger.error( 'Job %s matched by the TQ is not in Waiting state' % str( jobID ) )
result = gTaskQueueDB.deleteJob( jobID )
result = gJobDB.setJobStatus( jobID, status = 'Matched', minor = 'Assigned' )
result = gJobLoggingDB.addLoggingRecord( jobID,
status = 'Matched',
minor = 'Assigned',
source = 'Matcher' )
result = gJobDB.getJobJDL( jobID )
if not result['OK']:
return S_ERROR( 'Failed to get the job JDL' )
resultDict = {}
resultDict['JDL'] = result['Value']
resultDict['JobID'] = jobID
matchTime = time.time() - startTime
gLogger.info( "Match time: [%s]" % str( matchTime ) )
gMonitor.addMark( "matchTime", matchTime )
# Get some extra stuff into the response returned
resOpt = gJobDB.getJobOptParameters( jobID )
if resOpt['OK']:
for key, value in resOpt['Value'].items():
resultDict[key] = value
resAtt = gJobDB.getJobAttributes( jobID, ['OwnerDN', 'OwnerGroup'] )
if not resAtt['OK']:
return S_ERROR( 'Could not retrieve job attributes' )
if not resAtt['Value']:
return S_ERROR( 'No attributes returned for job' )
resultDict['DN'] = resAtt['Value']['OwnerDN']
resultDict['Group'] = resAtt['Value']['OwnerGroup']
return S_OK( resultDict )
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 _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 selectJob(self, resourceDescription):
""" Main job selection function to find the highest priority job
matching the resource capacity
"""
startTime = time.time()
# 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 taskQueueDB.getSingleValueTQDefFields():
if classAdAgent.lookupAttribute(name):
if name == "CPUTime":
resourceDict[name] = classAdAgent.getAttributeInt(name)
else:
resourceDict[name] = classAdAgent.getAttributeString(name)
for name in taskQueueDB.getMultiValueMatchFields():
if classAdAgent.lookupAttribute(name):
resourceDict[name] = classAdAgent.getAttributeString(name)
# Check if a JobID is requested
if classAdAgent.lookupAttribute("JobID"):
resourceDict["JobID"] = classAdAgent.getAttributeInt("JobID")
if classAdAgent.lookupAttribute("DIRACVersion"):
resourceDict["DIRACVersion"] = classAdAgent.getAttributeString("DIRACVersion")
else:
for name in taskQueueDB.getSingleValueTQDefFields():
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
for name in taskQueueDB.getMultiValueMatchFields():
if resourceDescription.has_key(name):
resourceDict[name] = resourceDescription[name]
if resourceDescription.has_key("JobID"):
resourceDict["JobID"] = resourceDescription["JobID"]
if resourceDescription.has_key("DIRACVersion"):
resourceDict["DIRACVersion"] = resourceDescription["DIRACVersion"]
# Check the pilot DIRAC version
if self.checkPilotVersion:
if "DIRACVersion" in resourceDict:
if self.pilotVersion and resourceDict["DIRACVersion"] != self.pilotVersion:
return S_ERROR(
"Pilot version does not match the production version %s:%s"
% (resourceDict["DIRACVersion"], self.pilotVersion)
)
# Get common site mask and check the agent site
result = jobDB.getSiteMask(siteState="Active")
if result["OK"]:
maskList = result["Value"]
else:
return S_ERROR("Internal error: can not get site mask")
if not "Site" in resourceDict:
return S_ERROR("Missing Site Name in Resource JDL")
siteName = resourceDict["Site"]
if resourceDict["Site"] not in maskList:
if "GridCE" in resourceDict:
del resourceDict["Site"]
else:
return S_ERROR("Site not in mask and GridCE not specified")
resourceDict["Setup"] = self.serviceInfoDict["clientSetup"]
if DEBUG:
print "Resource description:"
for k, v in resourceDict.items():
print k.rjust(20), v
# Check if Job Limits are imposed onto the site
extraConditions = {}
if self.siteJobLimits:
result = self.getExtraConditions(siteName)
if result["OK"]:
extraConditions = result["Value"]
if extraConditions:
gLogger.info("Job Limits for site %s are: %s" % (siteName, str(extraConditions)))
result = taskQueueDB.matchAndGetJob(resourceDict, extraConditions=extraConditions)
if DEBUG:
print result
if not result["OK"]:
return result
result = result["Value"]
if not result["matchFound"]:
return S_ERROR("No match found")
jobID = result["jobId"]
resAtt = jobDB.getJobAttributes(jobID, ["OwnerDN", "OwnerGroup", "Status"])
if not resAtt["OK"]:
return S_ERROR("Could not retrieve job attributes")
if not resAtt["Value"]:
return S_ERROR("No attributes returned for job")
if not resAtt["Value"]["Status"] == "Waiting":
gLogger.error("Job %s matched by the TQ is not in Waiting state" % str(jobID))
result = taskQueueDB.deleteJob(jobID)
result = jobDB.setJobStatus(jobID, status="Matched", minor="Assigned")
result = jobLoggingDB.addLoggingRecord(jobID, status="Matched", minor="Assigned", source="Matcher")
result = jobDB.getJobJDL(jobID)
if not result["OK"]:
return S_ERROR("Failed to get the job JDL")
resultDict = {}
resultDict["JDL"] = result["Value"]
resultDict["JobID"] = jobID
matchTime = time.time() - startTime
gLogger.info("Match time: [%s]" % str(matchTime))
gMonitor.addMark("matchTime", matchTime)
# Get some extra stuff into the response returned
resOpt = jobDB.getJobOptParameters(jobID)
if resOpt["OK"]:
for key, value in resOpt["Value"].items():
resultDict[key] = value
resAtt = jobDB.getJobAttributes(jobID, ["OwnerDN", "OwnerGroup"])
if not resAtt["OK"]:
return S_ERROR("Could not retrieve job attributes")
if not resAtt["Value"]:
return S_ERROR("No attributes returned for job")
resultDict["DN"] = resAtt["Value"]["OwnerDN"]
resultDict["Group"] = resAtt["Value"]["OwnerGroup"]
return S_OK(resultDict)
