def test_CreateAndSubmit( self ):
jobParams = {'JobID': '1',
'JobType': 'Merge',
'CPUTime': '1000000',
'Executable': '$DIRACROOT/scripts/dirac-jobexec',
'Arguments': "helloWorld.xml -o LogLevel=DEBUG pilot.cfg",
'ExtraOptions': 'pilot.cfg',
'InputSandbox': ['helloWorld.xml', 'exe-script.py']}
resourceParams = {}
optimizerParams = {}
# res = createJobWrapper( 1, jobParams, resourceParams, optimizerParams, logLevel = 'DEBUG' )
# self.assert_( res['OK'] )
# wrapperFile = res['Value']
ceFactory = ComputingElementFactory()
ceInstance = ceFactory.getCE( 'InProcess' )
self.assert_( ceInstance['OK'] )
computingElement = ceInstance['Value']
# res = computingElement.submitJob( wrapperFile, self.payloadProxy )
# self.assert_( res['OK'] )
res = createJobWrapper( 2, jobParams, resourceParams, optimizerParams, extraOptions = 'pilot.cfg', logLevel = 'DEBUG' )
self.assert_( res['OK'] )
wrapperFile = res['Value']
res = computingElement.submitJob( wrapperFile, self.payloadProxy )
self.assert_( res['OK'] )
def _setUpworkloadCE(self):
"""Get application queue and configure it
:return: a ComputingElement instance
"""
# Get CE parameters
workloadSite = gConfig.getValue("/LocalSite/Site")
workloadCE = gConfig.getValue("/LocalSite/GridCE")
workloadQueue = gConfig.getValue("/LocalSite/CEQueue")
result = getQueue(workloadSite, workloadCE, workloadQueue)
if not result["OK"]:
return result
ceType = result["Value"]["CEType"]
ceParams = result["Value"]
# Build CE
ceFactory = ComputingElementFactory()
result = ceFactory.getCE(ceName=workloadCE, ceType=ceType, ceParametersDict=ceParams)
if not result["OK"]:
return result
workloadCE = result["Value"]
# Add a proxy to the CE
result = getProxyInfo()
if not result["OK"] and not result["Value"]["chain"]:
return result
proxy = result["Value"]["chain"]
result = proxy.getRemainingSecs()
if not result["OK"]:
return result
lifetime_secs = result["Value"]
workloadCE.setProxy(proxy, lifetime_secs)
return S_OK(workloadCE)
def export_killPilot(self, pilotRefList ):
""" Kill the specified pilots
"""
# Make a list if it is not yet
pilotRefs = list( pilotRefList )
if type( pilotRefList ) in StringTypes:
pilotRefs = [pilotRefList]
# Regroup pilots per site and per owner
pilotRefDict = {}
for pilotReference in pilotRefs:
result = pilotDB.getPilotInfo(pilotReference)
if not result['OK'] or not result[ 'Value' ]:
return S_ERROR('Failed to get info for pilot ' + pilotReference)
pilotDict = result['Value'][pilotReference]
owner = pilotDict['OwnerDN']
group = pilotDict['OwnerGroup']
queue = '@@@'.join( [owner, group, pilotDict['GridSite'], pilotDict['DestinationSite'], pilotDict['Queue']] )
gridType = pilotDict['GridType']
pilotRefDict.setdefault( queue, {} )
pilotRefDict[queue].setdefault( 'PilotList', [] )
pilotRefDict[queue]['PilotList'].append( pilotReference )
pilotRefDict[queue]['GridType'] = gridType
# Do the work now queue by queue
ceFactory = ComputingElementFactory()
failed = []
for key, pilotDict in pilotRefDict.items():
owner,group,site,ce,queue = key.split( '@@@' )
result = getQueue( site, ce, queue )
if not result['OK']:
return result
queueDict = result['Value']
gridType = pilotDict['GridType']
result = ceFactory.getCE( gridType, ce, queueDict )
if not result['OK']:
return result
ce = result['Value']
if gridType in ["LCG","gLite","CREAM"]:
group = getGroupOption(group,'VOMSRole',group)
ret = gProxyManager.getPilotProxyFromVOMSGroup( owner, group )
if not ret['OK']:
gLogger.error( ret['Message'] )
gLogger.error( 'Could not get proxy:', 'User "%s", Group "%s"' % ( owner, group ) )
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = ret['Value']
ce.setProxy( proxy )
pilotList = pilotDict['PilotList']
result = ce.killJob( pilotList )
if not result['OK']:
failed.extend( pilotList )
if failed:
return S_ERROR('Failed to kill at least some pilots')
return S_OK()
def initialize(self, loops=0):
"""Sets default parameters and creates CE instance
"""
#Disable monitoring
self.am_setOption('MonitoringEnabled', False)
# self.log.setLevel('debug') #temporary for debugging
self.am_setOption('MaxCycles', loops)
ceUniqueID = self.am_getOption('CEUniqueID', 'InProcess')
localCE = gConfig.getValue('/LocalSite/LocalCE', '')
if localCE:
self.log.info('Defining CE from local configuration = %s' %
localCE)
ceUniqueID = localCE
ceFactory = ComputingElementFactory()
self.ceName = ceUniqueID
ceInstance = ceFactory.getCE(ceUniqueID)
if not ceInstance['OK']:
self.log.warn(ceInstance['Message'])
return ceInstance
self.computingElement = ceInstance['Value']
self.diracRoot = os.path.dirname(
os.path.dirname(os.path.dirname(os.path.dirname(__file__))))
#Localsite options
self.siteRoot = gConfig.getValue('/LocalSite/Root', os.getcwd())
self.siteName = gConfig.getValue('/LocalSite/Site', 'Unknown')
self.pilotReference = gConfig.getValue('/LocalSite/PilotReference',
'Unknown')
self.defaultProxyLength = gConfig.getValue(
'/Registry/DefaultProxyLifeTime', 86400 * 5)
#Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue('/LocalSite/CPUNormalizationFactor',
0.0)
defaultWrapperLocation = 'DIRAC/WorkloadManagementSystem/JobWrapper/JobWrapperTemplate.py'
self.jobWrapperTemplate = os.path.join(
self.diracRoot,
self.am_getOption('JobWrapperTemplate', defaultWrapperLocation))
self.jobSubmissionDelay = self.am_getOption('SubmissionDelay', 10)
self.defaultLogLevel = self.am_getOption('DefaultLogLevel', 'info')
self.fillingMode = self.am_getOption('FillingModeFlag', False)
self.stopOnApplicationFailure = self.am_getOption(
'StopOnApplicationFailure', True)
self.stopAfterFailedMatches = self.am_getOption(
'StopAfterFailedMatches', 10)
self.jobCount = 0
self.matchFailedCount = 0
#Timeleft
self.timeLeftUtil = TimeLeft()
self.timeLeft = gConfig.getValue(
'/Resources/Computing/CEDefaults/MaxCPUTime', 0.0)
self.gridCEQueue = gConfig.getValue(
'/Resources/Computing/CEDefaults/GridCEQueue', '')
self.timeLeftError = ''
self.scaledCPUTime = 0.0
self.pilotInfoReportedFlag = False
return S_OK()
def initialize(self, loops=0):
"""Sets default parameters and creates CE instance
"""
# Disable monitoring
self.am_setOption('MonitoringEnabled', False)
# self.log.setLevel('debug') #temporary for debugging
self.am_setOption('MaxCycles', loops)
ceType = self.am_getOption('CEType', 'InProcess')
localCE = gConfig.getValue('/LocalSite/LocalCE', '')
if localCE:
self.log.info('Defining CE from local configuration = %s' %
localCE)
ceType = localCE
# Create backend Computing Element
ceFactory = ComputingElementFactory()
self.ceName = ceType
ceInstance = ceFactory.getCE(ceType)
if not ceInstance['OK']:
self.log.warn(ceInstance['Message'])
return ceInstance
self.computingElement = ceInstance['Value']
result = self.computingElement.getDescription()
if not result['OK']:
self.log.warn("Can not get the CE description")
return result
ceDict = result['Value']
self.timeLeft = ceDict.get('CPUTime', self.timeLeft)
self.timeLeft = gConfig.getValue(
'/Resources/Computing/CEDefaults/MaxCPUTime', self.timeLeft)
self.initTimes = os.times()
# Localsite options
self.siteName = gConfig.getValue('/LocalSite/Site', self.siteName)
self.pilotReference = gConfig.getValue('/LocalSite/PilotReference',
self.pilotReference)
self.defaultProxyLength = gConfig.getValue(
'/Registry/DefaultProxyLifeTime', self.defaultProxyLength)
# Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue('/LocalSite/CPUNormalizationFactor',
self.cpuFactor)
self.jobSubmissionDelay = self.am_getOption('SubmissionDelay',
self.jobSubmissionDelay)
self.fillingMode = self.am_getOption('FillingModeFlag',
self.fillingMode)
self.minimumTimeLeft = self.am_getOption('MinimumTimeLeft',
self.minimumTimeLeft)
self.stopOnApplicationFailure = self.am_getOption(
'StopOnApplicationFailure', self.stopOnApplicationFailure)
self.stopAfterFailedMatches = self.am_getOption(
'StopAfterFailedMatches', self.stopAfterFailedMatches)
self.extraOptions = gConfig.getValue(
'/AgentJobRequirements/ExtraOptions', self.extraOptions)
# Timeleft
self.timeLeftUtil = TimeLeft()
return S_OK()
def getQueues( self, resourceDict ):
""" Get the list of relevant CEs and their descriptions
"""
self.queueDict = {}
ceFactory = ComputingElementFactory()
for site in resourceDict:
for ce in resourceDict[site]:
ceDict = resourceDict[site][ce]
qDict = ceDict.pop( 'Queues' )
for queue in qDict:
queueName = '%s_%s' % ( ce, queue )
self.queueDict[queueName] = {}
self.queueDict[queueName]['ParametersDict'] = qDict[queue]
self.queueDict[queueName]['ParametersDict']['Queue'] = queue
self.queueDict[queueName]['ParametersDict']['Site'] = site
self.queueDict[queueName]['ParametersDict']['GridEnv'] = self.gridEnv
self.queueDict[queueName]['ParametersDict']['Setup'] = gConfig.getValue( '/DIRAC/Setup', 'unknown' )
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if "maxCPUTime" in self.queueDict[queueName]['ParametersDict'] and \
"SI00" in self.queueDict[queueName]['ParametersDict']:
maxCPUTime = float( self.queueDict[queueName]['ParametersDict']['maxCPUTime'] )
# For some sites there are crazy values in the CS
maxCPUTime = max( maxCPUTime, 0 )
maxCPUTime = min( maxCPUTime, 86400 * 12.5 )
si00 = float( self.queueDict[queueName]['ParametersDict']['SI00'] )
queueCPUTime = 60. / 250. * maxCPUTime * si00
self.queueDict[queueName]['ParametersDict']['CPUTime'] = int( queueCPUTime )
qwDir = os.path.join( self.workingDirectory, queue )
if not os.path.exists( qwDir ):
os.makedirs( qwDir )
self.queueDict[queueName]['ParametersDict']['WorkingDirectory'] = qwDir
ceQueueDict = dict( ceDict )
ceQueueDict.update( self.queueDict[queueName]['ParametersDict'] )
result = ceFactory.getCE( ceName = ce,
ceType = ceDict['CEType'],
ceParametersDict = ceQueueDict )
if not result['OK']:
return result
self.queueDict[queueName]['CE'] = result['Value']
self.queueDict[queueName]['CEName'] = ce
self.queueDict[queueName]['CEType'] = ceDict['CEType']
self.queueDict[queueName]['Site'] = site
self.queueDict[queueName]['QueueName'] = queue
result = self.queueDict[queueName]['CE'].isValid()
if not result['OK']:
self.log.fatal( result['Message'] )
return result
if 'BundleProxy' in self.queueDict[queueName]['ParametersDict']:
self.queueDict[queueName]['BundleProxy'] = True
return S_OK()
def __createCE( self, ceName ):
"""
return a CE object for the given ceName
"""
self.log.info( "Creating %s CE" % ( ceName ) )
ceFactory = ComputingElementFactory()
ret = ceFactory.getCE( ceName )
if not ret['OK']:
self.log.warn( ret['Message'] )
return ret
return ret
def initialize( self, loops = 0 ):
"""Sets default parameters and creates CE instance
"""
# Disable monitoring
self.am_setOption( 'MonitoringEnabled', False )
# self.log.setLevel('debug') #temporary for debugging
self.am_setOption( 'MaxCycles', loops )
ceType = self.am_getOption( 'CEType', 'InProcess' )
localCE = gConfig.getValue( '/LocalSite/LocalCE', '' )
if localCE:
self.log.info( 'Defining CE from local configuration = %s' % localCE )
ceType = localCE
# Create backend Computing Element
ceFactory = ComputingElementFactory()
self.ceName = ceType
ceInstance = ceFactory.getCE( ceType )
if not ceInstance['OK']:
self.log.warn( ceInstance['Message'] )
return ceInstance
self.computingElement = ceInstance['Value']
result = self.computingElement.getDescription()
if not result['OK']:
self.log.warn( "Can not get the CE description" )
return result
ceDict = result['Value']
self.timeLeft = ceDict.get( 'CPUTime', 0.0 )
self.timeLeft = gConfig.getValue( '/Resources/Computing/CEDefaults/MaxCPUTime', self.timeLeft )
self.initTimes = os.times()
# Localsite options
self.siteName = gConfig.getValue( '/LocalSite/Site', 'Unknown' )
self.pilotReference = gConfig.getValue( '/LocalSite/PilotReference', 'Unknown' )
self.defaultProxyLength = gConfig.getValue( '/Registry/DefaultProxyLifeTime', 86400 * 5 )
# Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue( '/LocalSite/CPUNormalizationFactor', 0.0 )
self.jobSubmissionDelay = self.am_getOption( 'SubmissionDelay', 10 )
self.fillingMode = self.am_getOption( 'FillingModeFlag', False )
self.minimumTimeLeft = self.am_getOption( 'MinimumTimeLeft', 1000 )
self.stopOnApplicationFailure = self.am_getOption( 'StopOnApplicationFailure', True )
self.stopAfterFailedMatches = self.am_getOption( 'StopAfterFailedMatches', 10 )
self.jobCount = 0
self.matchFailedCount = 0
self.extraOptions = gConfig.getValue( '/AgentJobRequirements/ExtraOptions', '' )
# Timeleft
self.timeLeftUtil = TimeLeft()
self.timeLeftError = ''
self.scaledCPUTime = 0.0
self.pilotInfoReportedFlag = False
return S_OK()
def initialize(self, loops=0):
"""Sets default parameters and creates CE instance
"""
# Disable monitoring
self.am_setOption('MonitoringEnabled', False)
# self.log.setLevel('debug') #temporary for debugging
self.am_setOption('MaxCycles', loops)
ceType = self.am_getOption('CEType', 'InProcess')
localCE = gConfig.getValue('/LocalSite/LocalCE', '')
if localCE:
self.log.info('Defining CE from local configuration = %s' %
localCE)
ceType = localCE
ceFactory = ComputingElementFactory()
self.ceName = ceType
ceInstance = ceFactory.getCE(ceType)
if not ceInstance['OK']:
self.log.warn(ceInstance['Message'])
return ceInstance
self.initTimes = os.times()
self.computingElement = ceInstance['Value']
#Localsite options
self.siteName = gConfig.getValue('/LocalSite/Site', 'Unknown')
self.pilotReference = gConfig.getValue('/LocalSite/PilotReference',
'Unknown')
self.defaultProxyLength = gConfig.getValue(
'/Registry/DefaultProxyLifeTime', 86400 * 5)
#Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue('/LocalSite/CPUNormalizationFactor',
0.0)
self.jobSubmissionDelay = self.am_getOption('SubmissionDelay', 10)
self.fillingMode = self.am_getOption('FillingModeFlag', False)
self.stopOnApplicationFailure = self.am_getOption(
'StopOnApplicationFailure', True)
self.stopAfterFailedMatches = self.am_getOption(
'StopAfterFailedMatches', 10)
self.jobCount = 0
self.matchFailedCount = 0
self.extraOptions = gConfig.getValue(
'/AgentJobRequirements/ExtraOptions', '')
#Timeleft
self.timeLeftUtil = TimeLeft()
self.timeLeft = gConfig.getValue(
'/Resources/Computing/CEDefaults/MaxCPUTime', 0.0)
self.timeLeftError = ''
self.scaledCPUTime = 0.0
self.pilotInfoReportedFlag = False
return S_OK()
def __createCE( self, ceName ):
"""
return a CE object for the given ceName
"""
self.log.info( "Creating %s CE" % ( ceName ) )
ceFactory = ComputingElementFactory()
ret = ceFactory.getCE( ceName )
if not ret['OK']:
self.log.warn( ret['Message'] )
return ret
return ret
def initialize( self, loops = 0 ):
"""Sets default parameters and creates CE instance
"""
# Disable monitoring
self.am_setOption( 'MonitoringEnabled', False )
# self.log.setLevel('debug') #temporary for debugging
self.am_setOption( 'MaxCycles', loops )
ceType = self.am_getOption( 'CEType', 'InProcess' )
localCE = gConfig.getValue( '/LocalSite/LocalCE', '' )
if localCE:
self.log.info( 'Defining CE from local configuration = %s' % localCE )
ceType = localCE
ceFactory = ComputingElementFactory()
self.ceName = ceType
ceInstance = ceFactory.getCE( ceType )
if not ceInstance['OK']:
self.log.warn( ceInstance['Message'] )
return ceInstance
self.initTimes = os.times()
self.computingElement = ceInstance['Value']
self.diracRoot = os.path.dirname( os.path.dirname( os.path.dirname( os.path.dirname( __file__ ) ) ) )
#Localsite options
self.siteRoot = gConfig.getValue( '/LocalSite/Root', os.getcwd() )
self.siteName = gConfig.getValue( '/LocalSite/Site', 'Unknown' )
self.pilotReference = gConfig.getValue( '/LocalSite/PilotReference', 'Unknown' )
self.defaultProxyLength = gConfig.getValue( '/Registry/DefaultProxyLifeTime', 86400 * 5 )
#Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue( '/LocalSite/CPUNormalizationFactor', 0.0 )
defaultWrapperLocation = 'DIRAC/WorkloadManagementSystem/JobWrapper/JobWrapperTemplate.py'
self.jobWrapperTemplate = os.path.join( self.diracRoot,
self.am_getOption( 'JobWrapperTemplate',
defaultWrapperLocation ) )
self.jobSubmissionDelay = self.am_getOption( 'SubmissionDelay', 10 )
self.defaultLogLevel = self.am_getOption( 'DefaultLogLevel', 'info' )
self.fillingMode = self.am_getOption( 'FillingModeFlag', False )
self.stopOnApplicationFailure = self.am_getOption( 'StopOnApplicationFailure', True )
self.stopAfterFailedMatches = self.am_getOption( 'StopAfterFailedMatches', 10 )
self.jobCount = 0
self.matchFailedCount = 0
#Timeleft
self.timeLeftUtil = TimeLeft()
self.timeLeft = gConfig.getValue( '/Resources/Computing/CEDefaults/MaxCPUTime', 0.0 )
self.gridCEQueue = gConfig.getValue( '/Resources/Computing/CEDefaults/GridCEQueue', '' )
self.timeLeftError = ''
self.scaledCPUTime = 0.0
self.pilotInfoReportedFlag = False
return S_OK()
def addComputingElement(self, ceList):
"""
Check if a CE object for the current CE is available,
instantiate one if necessary
"""
for CE in ceList:
if CE not in self.computingElementDict:
ceFactory = ComputingElementFactory( )
ceInstance = ceFactory.getCE( ceName = CE )
if not ceInstance['OK']:
self.log.error('Can not create CE object:', ceInstance['Message'])
return
self.computingElementDict[CE] = ceInstance['Value'].ceConfigDict
# add the 'CE' instance at the end to avoid being overwritten
self.computingElementDict[CE]['CE'] = ceInstance['Value']
def addComputingElement(self, ceList):
"""
Check if a CE object for the current CE is available,
instantiate one if necessary
"""
for CE in ceList:
if CE not in self.computingElementDict:
ceFactory = ComputingElementFactory( )
ceInstance = ceFactory.getCE( ceName = CE )
if not ceInstance['OK']:
self.log.error('Can not create CE object:', ceInstance['Message'])
return
self.computingElementDict[CE] = ceInstance['Value'].ceConfigDict
# add the 'CE' instance at the end to avoid being overwritten
self.computingElementDict[CE]['CE'] = ceInstance['Value']
def _initializeComputingElement(self, localCE):
"""Generate a ComputingElement and configure it"""
ceFactory = ComputingElementFactory()
self.ceName = localCE.split("/")[
0] # It might be "Pool/Singularity", or simply "Pool"
self.innerCESubmissionType = (localCE.split("/")[1] if len(
localCE.split("/")) == 2 else self.innerCESubmissionType)
ceInstance = ceFactory.getCE(self.ceName)
if not ceInstance["OK"]:
self.log.warn("Can't instantiate a CE", ceInstance["Message"])
return ceInstance
self.computingElement = ceInstance["Value"]
self.computingElement.setParameters(
{"InnerCESubmissionType": self.innerCESubmissionType})
return S_OK()
def getCREAMPilotOutput(proxy,pilotRef,pilotStamp):
"""
"""
gridEnv = getGridEnv()
tmpdir = mkdtemp()
result = ComputingElementFactory().getCE(ceName = 'CREAMSite',ceType = 'CREAM',
ceParametersDict = {'GridEnv':gridEnv,
'Queue':'Qeuue',
'OutputURL':"gsiftp://localhost",
'WorkingDirectory':tmpdir})
if not result['OK']:
shutil.rmtree(tmpdir)
return result
ce = result['Value']
ce.reset()
ce.setProxy(proxy)
fullPilotRef = ":::".join([pilotRef,pilotStamp])
result = ce.getJobOutput( fullPilotRef )
shutil.rmtree(tmpdir)
if not result['OK']:
return S_ERROR( 'Failed to get pilot output: %s' % result['Message'] )
output, error = result['Value']
fileList = outputSandboxFiles
result = S_OK()
result['FileList'] = fileList
result['StdOut'] = output
result['StdErr'] = error
return result
def getPilotCE(pilotDict):
"""Instantiate and return a CE bound to a pilot"""
ceFactory = ComputingElementFactory()
result = getQueue(pilotDict["GridSite"], pilotDict["DestinationSite"], pilotDict["Queue"])
if not result["OK"]:
return result
queueDict = result["Value"]
gridEnv = getGridEnv()
queueDict["GridEnv"] = gridEnv
queueDict["WorkingDirectory"] = mkdtemp()
result = ceFactory.getCE(pilotDict["GridType"], pilotDict["DestinationSite"], queueDict)
if not result["OK"]:
shutil.rmtree(queueDict["WorkingDirectory"])
return result
ce = result["Value"]
return S_OK(ce)
def test_CreateAndSubmit(self):
jobParams = {
"JobID": "1",
"JobType": "Merge",
"CPUTime": "1000000",
"Executable": "dirac-jobexec",
"Arguments": "helloWorld.xml -o LogLevel=DEBUG --cfg pilot.cfg",
"InputSandbox": ["helloWorld.xml", "exe-script.py"],
}
resourceParams = {}
optimizerParams = {}
# res = createJobWrapper( 1, jobParams, resourceParams, optimizerParams, logLevel = 'DEBUG' )
# self.assertTrue( res['OK'] )
# wrapperFile = res['Value']
ceFactory = ComputingElementFactory()
ceInstance = ceFactory.getCE("InProcess")
self.assertTrue(ceInstance["OK"])
computingElement = ceInstance["Value"]
# res = computingElement.submitJob( wrapperFile, self.payloadProxy )
# self.assertTrue( res['OK'] )
if "pilot.cfg" in os.listdir("."):
jobParams.setdefault("ExtraOptions", "pilot.cfg")
res = createJobWrapper(2,
jobParams,
resourceParams,
optimizerParams,
extraOptions="pilot.cfg",
logLevel="DEBUG")
else:
res = createJobWrapper(2,
jobParams,
resourceParams,
optimizerParams,
logLevel="DEBUG")
self.assertTrue(res["OK"], res.get("Message"))
wrapperFile = res["Value"][0]
res = computingElement.submitJob(wrapperFile, self.payloadProxy)
self.assertTrue(res["OK"], res.get("Message"))
def killPilotsInQueues(pilotRefDict):
"""kill pilots queue by queue
:params dict pilotRefDict: a dict of pilots in queues
"""
ceFactory = ComputingElementFactory()
failed = []
for key, pilotDict in pilotRefDict.items():
owner, group, site, ce, queue = key.split("@@@")
result = getQueue(site, ce, queue)
if not result["OK"]:
return result
queueDict = result["Value"]
gridType = pilotDict["GridType"]
result = ceFactory.getCE(gridType, ce, queueDict)
if not result["OK"]:
return result
ce = result["Value"]
# FIXME: quite hacky. Should be either removed, or based on some flag
if gridType in ["CREAM", "ARC", "Globus", "HTCondorCE"]:
group = getGroupOption(group, "VOMSRole", group)
ret = gProxyManager.getPilotProxyFromVOMSGroup(owner, group)
if not ret["OK"]:
gLogger.error(
"Could not get proxy:", 'User "%s" Group "%s" : %s' %
(owner, group, ret["Message"]))
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = ret["Value"]
ce.setProxy(proxy)
pilotList = pilotDict["PilotList"]
result = ce.killJob(pilotList)
if not result["OK"]:
failed.extend(pilotList)
return failed
def test__getCEDict(mocker, ceType, expectedType, expectedNumberElement):
"""Test JobAgent()._getCEDict()"""
mocker.patch(
"DIRAC.WorkloadManagementSystem.Agent.JobAgent.AgentModule.__init__")
jobAgent = JobAgent("Test", "Test1")
jobAgent.log = gLogger
jobAgent.log.setLevel("DEBUG")
result = ComputingElementFactory().getCE(ceType)
assert result["OK"]
ce = result["Value"]
ce.ceParameters["MultiProcessorStrategy"] = True
ce.ceParameters["NumberOfProcessors"] = 4
result = jobAgent._getCEDict(ce)
assert result["OK"]
ceDict = result["Value"]
assert isinstance(ceDict, expectedType)
assert len(ceDict) == expectedNumberElement
def test__checkCEAvailability(mocker, ceType, mockCEReply, expectedResult):
"""Test JobAgent()._checkAvailability()"""
mocker.patch(
"DIRAC.WorkloadManagementSystem.Agent.JobAgent.AgentModule.__init__")
mocker.patch(
"DIRAC.Resources.Computing.ComputingElement.ComputingElement.available",
return_value=mockCEReply)
jobAgent = JobAgent("Test", "Test1")
jobAgent.log = gLogger
jobAgent.log.setLevel("DEBUG")
result = ComputingElementFactory().getCE(ceType)
assert result["OK"]
ce = result["Value"]
result = jobAgent._checkCEAvailability(ce)
assert result["OK"] == expectedResult["OK"]
if "Value" in expectedResult:
assert result["Value"] == expectedResult["Value"]
if "Message" in expectedResult:
assert result["Message"] == expectedResult["Message"]
def getQueues(self, resourceDict):
""" Get the list of relevant CEs and their descriptions
"""
self.queueDict = {}
ceFactory = ComputingElementFactory()
for site in resourceDict:
for ce in resourceDict[site]:
ceDict = resourceDict[site][ce]
qDict = ceDict.pop("Queues")
for queue in qDict:
queueName = "%s_%s" % (ce, queue)
self.queueDict[queueName] = {}
self.queueDict[queueName]["ParametersDict"] = qDict[queue]
self.queueDict[queueName]["ParametersDict"]["Queue"] = queue
self.queueDict[queueName]["ParametersDict"]["Site"] = site
self.queueDict[queueName]["ParametersDict"]["GridEnv"] = self.gridEnv
self.queueDict[queueName]["ParametersDict"]["Setup"] = gConfig.getValue("/DIRAC/Setup", "unknown")
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if (
"maxCPUTime" in self.queueDict[queueName]["ParametersDict"]
and "SI00" in self.queueDict[queueName]["ParametersDict"]
):
maxCPUTime = float(self.queueDict[queueName]["ParametersDict"]["maxCPUTime"])
# For some sites there are crazy values in the CS
maxCPUTime = max(maxCPUTime, 0)
maxCPUTime = min(maxCPUTime, 86400 * 12.5)
si00 = float(self.queueDict[queueName]["ParametersDict"]["SI00"])
queueCPUTime = 60.0 / 250.0 * maxCPUTime * si00
self.queueDict[queueName]["ParametersDict"]["CPUTime"] = int(queueCPUTime)
qwDir = os.path.join(self.workingDirectory, queue)
if not os.path.exists(qwDir):
os.makedirs(qwDir)
self.queueDict[queueName]["ParametersDict"]["WorkingDirectory"] = qwDir
platform = ""
if "Platform" in self.queueDict[queueName]["ParametersDict"]:
platform = self.queueDict[queueName]["ParametersDict"]["Platform"]
elif "Platform" in ceDict:
platform = ceDict["Platform"]
elif "OS" in ceDict:
architecture = ceDict.get("architecture", "x86_64")
OS = ceDict["OS"]
platform = "_".join([architecture, OS])
if platform and not platform in self.platforms:
self.platforms.append(platform)
if not "Platform" in self.queueDict[queueName]["ParametersDict"] and platform:
result = Resources.getDIRACPlatform(platform)
if result["OK"]:
self.queueDict[queueName]["ParametersDict"]["Platform"] = result["Value"]
ceQueueDict = dict(ceDict)
ceQueueDict.update(self.queueDict[queueName]["ParametersDict"])
result = ceFactory.getCE(ceName=ce, ceType=ceDict["CEType"], ceParametersDict=ceQueueDict)
if not result["OK"]:
return result
self.queueDict[queueName]["CE"] = result["Value"]
self.queueDict[queueName]["CEName"] = ce
self.queueDict[queueName]["CEType"] = ceDict["CEType"]
self.queueDict[queueName]["Site"] = site
self.queueDict[queueName]["QueueName"] = queue
result = self.queueDict[queueName]["CE"].isValid()
if not result["OK"]:
self.log.fatal(result["Message"])
return result
if "BundleProxy" in self.queueDict[queueName]["ParametersDict"]:
self.queueDict[queueName]["BundleProxy"] = True
elif "BundleProxy" in ceDict:
self.queueDict[queueName]["BundleProxy"] = True
if site not in self.sites:
self.sites.append(site)
return S_OK()
def getQueues( self, resourceDict ):
""" Get the list of relevant CEs and their descriptions
"""
self.queueDict = {}
ceFactory = ComputingElementFactory()
for site in resourceDict:
for ce in resourceDict[site]:
ceDict = resourceDict[site][ce]
ceTags = ceDict.get( 'Tag' )
if isinstance( ceTags, basestring ):
ceTags = fromChar( ceTags )
qDict = ceDict.pop( 'Queues' )
for queue in qDict:
queueName = '%s_%s' % ( ce, queue )
self.queueDict[queueName] = {}
self.queueDict[queueName]['ParametersDict'] = qDict[queue]
self.queueDict[queueName]['ParametersDict']['Queue'] = queue
self.queueDict[queueName]['ParametersDict']['Site'] = site
self.queueDict[queueName]['ParametersDict']['GridEnv'] = self.gridEnv
self.queueDict[queueName]['ParametersDict']['Setup'] = gConfig.getValue( '/DIRAC/Setup', 'unknown' )
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if "maxCPUTime" in self.queueDict[queueName]['ParametersDict'] and \
"SI00" in self.queueDict[queueName]['ParametersDict']:
maxCPUTime = float( self.queueDict[queueName]['ParametersDict']['maxCPUTime'] )
# For some sites there are crazy values in the CS
maxCPUTime = max( maxCPUTime, 0 )
maxCPUTime = min( maxCPUTime, 86400 * 12.5 )
si00 = float( self.queueDict[queueName]['ParametersDict']['SI00'] )
queueCPUTime = 60. / 250. * maxCPUTime * si00
self.queueDict[queueName]['ParametersDict']['CPUTime'] = int( queueCPUTime )
queueTags = self.queueDict[queueName]['ParametersDict'].get( 'Tag' )
if queueTags and isinstance( queueTags, basestring ):
queueTags = fromChar( queueTags )
self.queueDict[queueName]['ParametersDict']['Tag'] = queueTags
if ceTags:
if queueTags:
allTags = list( set( ceTags + queueTags ) )
self.queueDict[queueName]['ParametersDict']['Tag'] = allTags
else:
self.queueDict[queueName]['ParametersDict']['Tag'] = ceTags
maxMemory = self.queueDict[queueName]['ParametersDict'].get( 'MaxRAM', None )
if maxMemory:
# MaxRAM value is supposed to be in MB
maxMemoryList = range( 1, int( maxMemory )/1000 + 1 )
memoryTags = [ '%dGB' % mem for mem in maxMemoryList ]
if memoryTags:
self.queueDict[queueName]['ParametersDict'].setdefault( 'Tag', [] )
self.queueDict[queueName]['ParametersDict']['Tag'] += memoryTags
qwDir = os.path.join( self.workingDirectory, queue )
if not os.path.exists( qwDir ):
os.makedirs( qwDir )
self.queueDict[queueName]['ParametersDict']['WorkingDirectory'] = qwDir
platform = ''
if "Platform" in self.queueDict[queueName]['ParametersDict']:
platform = self.queueDict[queueName]['ParametersDict']['Platform']
elif "Platform" in ceDict:
platform = ceDict['Platform']
elif "OS" in ceDict:
architecture = ceDict.get( 'architecture', 'x86_64' )
OS = ceDict['OS']
platform = '_'.join( [architecture, OS] )
if platform and not platform in self.platforms:
self.platforms.append( platform )
if not "Platform" in self.queueDict[queueName]['ParametersDict'] and platform:
result = Resources.getDIRACPlatform( platform )
if result['OK']:
self.queueDict[queueName]['ParametersDict']['Platform'] = result['Value'][0]
ceQueueDict = dict( ceDict )
ceQueueDict.update( self.queueDict[queueName]['ParametersDict'] )
# Generate the CE object for the queue or pick the already existing one
# if the queue definition did not change
queueHash = self.__generateQueueHash( ceQueueDict )
if queueName in self.queueCECache and self.queueCECache[queueName]['Hash'] == queueHash:
queueCE = self.queueCECache[queueName]['CE']
else:
result = ceFactory.getCE( ceName = ce,
ceType = ceDict['CEType'],
ceParametersDict = ceQueueDict )
if not result['OK']:
return result
self.queueCECache.setdefault( queueName, {} )
self.queueCECache[queueName]['Hash'] = queueHash
self.queueCECache[queueName]['CE'] = result['Value']
queueCE = self.queueCECache[queueName]['CE']
self.queueDict[queueName]['CE'] = queueCE
self.queueDict[queueName]['CEName'] = ce
self.queueDict[queueName]['CEType'] = ceDict['CEType']
self.queueDict[queueName]['Site'] = site
self.queueDict[queueName]['QueueName'] = queue
self.queueDict[queueName]['Platform'] = platform
result = self.queueDict[queueName]['CE'].isValid()
if not result['OK']:
self.log.fatal( result['Message'] )
return result
if 'BundleProxy' in self.queueDict[queueName]['ParametersDict']:
if self.queueDict[queueName]['ParametersDict']['BundleProxy'].lower() in ['true','yes','1']:
self.queueDict[queueName]['BundleProxy'] = True
elif 'BundleProxy' in ceDict:
if ceDict['BundleProxy'].lower() in ['true','yes','1']:
self.queueDict[queueName]['BundleProxy'] = True
if site not in self.sites:
self.sites.append( site )
return S_OK()
def getQueues( self ):
""" Get the list of relevant CEs and their descriptions
"""
ceFactory = ComputingElementFactory()
ceTypes = self.am_getOption( 'CETypes', [] )
ceConfList = self.am_getOption( 'CEs', [] )
for siteName in self.siteNames:
# Look up CE definitions in the site CS description
ceList = []
gridType = siteName.split( '.' )[0]
result = gConfig.getSections( '/Resources/Sites/%s/%s/CEs' % ( gridType, siteName ) )
if not result['OK']:
return S_ERROR( 'Failed to look up the CS for the site %s CEs' % siteName )
if not result['Value']:
return S_ERROR( 'No CEs found for site %s' % siteName )
ceTotalList = result['Value']
for ce in ceTotalList:
if ( ceConfList and ce in ceConfList ) or not ceConfList:
ceType = gConfig.getValue( '/Resources/Sites/%s/%s/CEs/%s/CEType' % ( gridType, siteName, ce ), 'Unknown' )
result = gConfig.getOptionsDict( '/Resources/Sites/%s/%s/CEs/%s' % ( gridType, siteName, ce ) )
if not result['OK']:
return S_ERROR( 'Failed to look up the CS for ce %s' % ce )
ceDict = result['Value']
if "SubmissionMode" in ceDict and ceDict['SubmissionMode'].lower() == "direct":
if ceType in ceTypes:
ceList.append( ( ce, ceType, ceDict ) )
for ce, ceType, ceDict in ceList:
section = '/Resources/Sites/%s/%s/CEs/%s/Queues' % ( gridType, siteName, ce )
result = gConfig.getSections( section )
if not result['OK']:
return S_ERROR( 'Failed to look up the CS for queues' )
if not result['Value']:
return S_ERROR( 'No Queues found for site %s, ce %s' % ( siteName, ce ) )
queues = result['Value']
for queue in queues:
result = gConfig.getOptionsDict( '%s/%s' % ( section, queue ) )
if not result['OK']:
return S_ERROR( 'Failed to look up the CS for ce,queue %s,%s' % ( ce, queue ) )
queueName = '%s_%s' % ( ce, queue )
self.queueDict[queueName] = {}
self.queueDict[queueName]['ParametersDict'] = result['Value']
self.queueDict[queueName]['ParametersDict']['Queue'] = queue
self.queueDict[queueName]['ParametersDict']['Site'] = siteName
self.queueDict[queueName]['ParametersDict']['GridEnv'] = self.gridEnv
self.queueDict[queueName]['ParametersDict']['Setup'] = gConfig.getValue( '/DIRAC/Setup', 'unknown' )
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if "maxCPUTime" in self.queueDict[queueName]['ParametersDict'] and \
"SI00" in self.queueDict[queueName]['ParametersDict']:
maxCPUTime = float( self.queueDict[queueName]['ParametersDict']['maxCPUTime'] )
# For some sites there are crazy values in the CS
maxCPUTime = max( maxCPUTime, 0 )
maxCPUTime = min( maxCPUTime, 86400 * 12.5 )
si00 = float( self.queueDict[queueName]['ParametersDict']['SI00'] )
queueCPUTime = 60. / 250. * maxCPUTime * si00
self.queueDict[queueName]['ParametersDict']['CPUTime'] = int( queueCPUTime )
qwDir = os.path.join( self.workingDirectory, queue )
if not os.path.exists( qwDir ):
os.mkdir( qwDir )
self.queueDict[queueName]['ParametersDict']['WorkingDirectory'] = qwDir
queueDict = dict( ceDict )
queueDict.update( self.queueDict[queueName]['ParametersDict'] )
result = ceFactory.getCE( ceName = ce,
ceType = ceType,
ceParametersDict = queueDict )
if not result['OK']:
return result
self.queueDict[queueName]['CE'] = result['Value']
self.queueDict[queueName]['CEName'] = ce
self.queueDict[queueName]['CEType'] = ceType
self.queueDict[queueName]['Site'] = siteName
self.queueDict[queueName]['QueueName'] = queue
result = self.queueDict[queueName]['CE'].isValid()
if not result['OK']:
self.log.fatal( result['Message'] )
return result
if 'BundleProxy' in self.queueDict[queueName]['ParametersDict']:
self.queueDict[queueName]['BundleProxy'] = True
return S_OK()
def initialize(self, loops=0):
"""Sets default parameters and creates CE instance
"""
# Disable monitoring, logLevel INFO, limited cycles
self.am_setOption('MonitoringEnabled', False)
self.am_setOption('MaxCycles', loops)
ceType = self.am_getOption('CEType', self.ceName)
localCE = gConfig.getValue('/LocalSite/LocalCE', '')
if localCE:
self.log.info('Defining CE from local configuration',
'= %s' % localCE)
ceType = localCE
# Create backend Computing Element
ceFactory = ComputingElementFactory()
self.ceName = ceType.split('/')[
0] # It might be "Pool/Singularity", or simply "Pool"
self.innerCESubmissionType = ceType.split('/')[1] if len(
ceType.split('/')) == 2 else self.innerCESubmissionType
ceInstance = ceFactory.getCE(self.ceName)
if not ceInstance['OK']:
self.log.warn("Can't instantiate a CE", ceInstance['Message'])
return ceInstance
self.computingElement = ceInstance['Value']
self.computingElement.ceParameters[
'InnerCESubmissionType'] = self.innerCESubmissionType
result = self.computingElement.getDescription()
if not result['OK']:
self.log.warn("Can not get the CE description")
return result
if isinstance(result['Value'], list):
ceDict = result['Value'][0]
else:
ceDict = result['Value']
self.timeLeft = ceDict.get('CPUTime', self.timeLeft)
self.timeLeft = gConfig.getValue(
'/Resources/Computing/CEDefaults/MaxCPUTime', self.timeLeft)
self.initTimes = os.times()
# Localsite options
self.siteName = gConfig.getValue('/LocalSite/Site', self.siteName)
self.pilotReference = gConfig.getValue('/LocalSite/PilotReference',
self.pilotReference)
self.defaultProxyLength = gConfig.getValue(
'/Registry/DefaultProxyLifeTime', self.defaultProxyLength)
# Agent options
# This is the factor to convert raw CPU to Normalized units (based on the CPU Model)
self.cpuFactor = gConfig.getValue('/LocalSite/CPUNormalizationFactor',
self.cpuFactor)
self.jobSubmissionDelay = self.am_getOption('SubmissionDelay',
self.jobSubmissionDelay)
self.fillingMode = self.am_getOption('FillingModeFlag',
self.fillingMode)
self.minimumTimeLeft = self.am_getOption('MinimumTimeLeft',
self.minimumTimeLeft)
self.stopOnApplicationFailure = self.am_getOption(
'StopOnApplicationFailure', self.stopOnApplicationFailure)
self.stopAfterFailedMatches = self.am_getOption(
'StopAfterFailedMatches', self.stopAfterFailedMatches)
self.extraOptions = gConfig.getValue(
'/AgentJobRequirements/ExtraOptions', self.extraOptions)
# Timeleft
self.timeLeftUtil = TimeLeft()
return S_OK()
def __getGridJobOutput(pilotReference):
""" Get the pilot job standard output and standard error files for the Grid
job reference
"""
result = pilotDB.getPilotInfo(pilotReference)
if not result['OK'] or not result['Value']:
return S_ERROR('Failed to get info for pilot ' + pilotReference)
pilotDict = result['Value'][pilotReference]
owner = pilotDict['OwnerDN']
group = pilotDict['OwnerGroup']
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
result = pilotDB.getPilotOutput(pilotReference)
if result['OK']:
stdout = result['Value']['StdOut']
error = result['Value']['StdErr']
if stdout or error:
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
return S_OK(resultDict)
else:
gLogger.warn('Empty pilot output found for %s' %
pilotReference)
gridType = pilotDict['GridType']
if gridType == "gLite":
result = getWMSPilotOutput(pilotReference,
proxyUserDN=owner,
proxyUserGroup=group) #pylint: disable=unexpected-keyword-arg
if not result['OK']:
return S_ERROR('Failed to get pilot output: ' +
result['Message'])
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
stdout = result['StdOut']
error = result['StdErr']
fileList = result['FileList']
if stdout:
result = pilotDB.storePilotOutput(pilotReference, stdout,
error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',
result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = fileList
return S_OK(resultDict)
else:
# Instantiate the appropriate CE
ceFactory = ComputingElementFactory()
result = getQueue(pilotDict['GridSite'],
pilotDict['DestinationSite'], pilotDict['Queue'])
if not result['OK']:
return result
queueDict = result['Value']
gridEnv = getGridEnv()
queueDict['GridEnv'] = gridEnv
queueDict['WorkingDirectory'] = mkdtemp()
result = ceFactory.getCE(gridType, pilotDict['DestinationSite'],
queueDict)
if not result['OK']:
shutil.rmtree(queueDict['WorkingDirectory'])
return result
ce = result['Value']
groupVOMS = getGroupOption(group, 'VOMSRole', group)
result = gProxyManager.getPilotProxyFromVOMSGroup(owner, groupVOMS)
if not result['OK']:
gLogger.error(result['Message'])
gLogger.error('Could not get proxy:',
'User "%s", Group "%s"' % (owner, groupVOMS))
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = result['Value']
ce.setProxy(proxy)
pilotStamp = pilotDict['PilotStamp']
pRef = pilotReference
if pilotStamp:
pRef = pRef + ':::' + pilotStamp
result = ce.getJobOutput(pRef)
if not result['OK']:
shutil.rmtree(queueDict['WorkingDirectory'])
return result
stdout, error = result['Value']
if stdout:
result = pilotDB.storePilotOutput(pilotReference, stdout,
error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',
result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
shutil.rmtree(queueDict['WorkingDirectory'])
return S_OK(resultDict)
def getQueuesResolved(
siteDict, queueCECache, gridEnv=None, setup=None, workingDir="", checkPlatform=False, instantiateCEs=False
):
"""Get the list of relevant CEs (what is in siteDict) and their descriptions.
The main goal of this method is to return a dictionary of queues
"""
queueDict = {}
ceFactory = ComputingElementFactory()
for site in siteDict:
for ce in siteDict[site]:
ceDict = siteDict[site][ce]
pilotRunDirectory = ceDict.get("PilotRunDirectory", "")
# ceMaxRAM = ceDict.get('MaxRAM', None)
qDict = ceDict.pop("Queues")
for queue in qDict:
queueName = "%s_%s" % (ce, queue)
queueDict[queueName] = {}
queueDict[queueName]["ParametersDict"] = qDict[queue]
queueDict[queueName]["ParametersDict"]["Queue"] = queue
queueDict[queueName]["ParametersDict"]["GridCE"] = ce
queueDict[queueName]["ParametersDict"]["Site"] = site
queueDict[queueName]["ParametersDict"]["GridEnv"] = gridEnv
queueDict[queueName]["ParametersDict"]["Setup"] = setup
# Evaluate the CPU limit of the queue according to the Glue convention
computeQueueCPULimit(queueDict[queueName]["ParametersDict"])
# Tags & RequiredTags defined on the Queue level and on the CE level are concatenated
# This also converts them from a string to a list if required.
resolveTags(ceDict, queueDict[queueName]["ParametersDict"])
# Some parameters can be defined on the CE level and are inherited by all Queues
setAdditionalParams(ceDict, queueDict[queueName]["ParametersDict"])
if pilotRunDirectory:
queueDict[queueName]["ParametersDict"]["JobExecDir"] = pilotRunDirectory
ceQueueDict = dict(ceDict)
ceQueueDict.update(queueDict[queueName]["ParametersDict"])
if instantiateCEs:
# Generate the CE object for the queue or pick the already existing one
# if the queue definition did not change
queueHash = generateQueueHash(ceQueueDict)
if queueName in queueCECache and queueCECache[queueName]["Hash"] == queueHash:
queueCE = queueCECache[queueName]["CE"]
else:
result = ceFactory.getCE(ceName=ce, ceType=ceDict["CEType"], ceParametersDict=ceQueueDict)
if not result["OK"]:
queueDict.pop(queueName)
continue
queueCECache.setdefault(queueName, {})
queueCECache[queueName]["Hash"] = queueHash
queueCECache[queueName]["CE"] = result["Value"]
queueCE = queueCECache[queueName]["CE"]
queueDict[queueName]["ParametersDict"].update(queueCE.ceParameters)
queueDict[queueName]["CE"] = queueCE
result = queueDict[queueName]["CE"].isValid()
if not result["OK"]:
queueDict.pop(queueName)
queueCECache.pop(queueName)
continue
queueDict[queueName]["CEName"] = ce
queueDict[queueName]["CEType"] = ceDict["CEType"]
queueDict[queueName]["Site"] = site
queueDict[queueName]["QueueName"] = queue
queueDict[queueName]["QueryCEFlag"] = ceDict.get("QueryCEFlag", "false")
if checkPlatform:
setPlatform(ceDict, queueDict[queueName]["ParametersDict"])
bundleProxy = queueDict[queueName]["ParametersDict"].get("BundleProxy", ceDict.get("BundleProxy"))
if bundleProxy and bundleProxy.lower() in ["true", "yes", "1"]:
queueDict[queueName]["BundleProxy"] = True
return S_OK(queueDict)
def getQueues( self, resourceDict ):
""" Get the list of relevant CEs and their descriptions
"""
self.queueDict = {}
ceFactory = ComputingElementFactory()
for site in resourceDict:
for ce in resourceDict[site]:
ceDict = resourceDict[site][ce]
ceTags = ceDict.get( 'Tag', [] )
pilotRunDirectory = ceDict.get( 'PilotRunDirectory', '' )
if isinstance( ceTags, basestring ):
ceTags = fromChar( ceTags )
ceMaxRAM = ceDict.get( 'MaxRAM', None )
qDict = ceDict.pop( 'Queues' )
for queue in qDict:
queueName = '%s_%s' % ( ce, queue )
self.queueDict[queueName] = {}
self.queueDict[queueName]['ParametersDict'] = qDict[queue]
self.queueDict[queueName]['ParametersDict']['Queue'] = queue
self.queueDict[queueName]['ParametersDict']['Site'] = site
self.queueDict[queueName]['ParametersDict']['GridEnv'] = self.gridEnv
self.queueDict[queueName]['ParametersDict']['Setup'] = gConfig.getValue( '/DIRAC/Setup', 'unknown' )
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if "maxCPUTime" in self.queueDict[queueName]['ParametersDict'] and \
"SI00" in self.queueDict[queueName]['ParametersDict']:
maxCPUTime = float( self.queueDict[queueName]['ParametersDict']['maxCPUTime'] )
# For some sites there are crazy values in the CS
maxCPUTime = max( maxCPUTime, 0 )
maxCPUTime = min( maxCPUTime, 86400 * 12.5 )
si00 = float( self.queueDict[queueName]['ParametersDict']['SI00'] )
queueCPUTime = 60. / 250. * maxCPUTime * si00
self.queueDict[queueName]['ParametersDict']['CPUTime'] = int( queueCPUTime )
queueTags = self.queueDict[queueName]['ParametersDict'].get( 'Tag' )
if queueTags and isinstance( queueTags, basestring ):
queueTags = fromChar( queueTags )
self.queueDict[queueName]['ParametersDict']['Tag'] = queueTags
if ceTags:
if queueTags:
allTags = list( set( ceTags + queueTags ) )
self.queueDict[queueName]['ParametersDict']['Tag'] = allTags
else:
self.queueDict[queueName]['ParametersDict']['Tag'] = ceTags
maxRAM = self.queueDict[queueName]['ParametersDict'].get( 'MaxRAM' )
maxRAM = ceMaxRAM if not maxRAM else maxRAM
if maxRAM:
self.queueDict[queueName]['ParametersDict']['MaxRAM'] = maxRAM
if pilotRunDirectory:
self.queueDict[queueName]['ParametersDict']['JobExecDir'] = pilotRunDirectory
qwDir = os.path.join( self.workingDirectory, queue )
mkDir(qwDir)
self.queueDict[queueName]['ParametersDict']['WorkingDirectory'] = qwDir
platform = ''
if "Platform" in self.queueDict[queueName]['ParametersDict']:
platform = self.queueDict[queueName]['ParametersDict']['Platform']
elif "Platform" in ceDict:
platform = ceDict['Platform']
elif "OS" in ceDict:
architecture = ceDict.get( 'architecture', 'x86_64' )
OS = ceDict['OS']
platform = '_'.join( [architecture, OS] )
if platform and not platform in self.platforms:
self.platforms.append( platform )
if not "Platform" in self.queueDict[queueName]['ParametersDict'] and platform:
result = Resources.getDIRACPlatform( platform )
if result['OK']:
self.queueDict[queueName]['ParametersDict']['Platform'] = result['Value'][0]
ceQueueDict = dict( ceDict )
ceQueueDict.update( self.queueDict[queueName]['ParametersDict'] )
# Generate the CE object for the queue or pick the already existing one
# if the queue definition did not change
queueHash = self.__generateQueueHash( ceQueueDict )
if queueName in self.queueCECache and self.queueCECache[queueName]['Hash'] == queueHash:
queueCE = self.queueCECache[queueName]['CE']
else:
result = ceFactory.getCE( ceName = ce,
ceType = ceDict['CEType'],
ceParametersDict = ceQueueDict )
if not result['OK']:
return result
self.queueCECache.setdefault( queueName, {} )
self.queueCECache[queueName]['Hash'] = queueHash
self.queueCECache[queueName]['CE'] = result['Value']
queueCE = self.queueCECache[queueName]['CE']
self.queueDict[queueName]['CE'] = queueCE
self.queueDict[queueName]['CEName'] = ce
self.queueDict[queueName]['CEType'] = ceDict['CEType']
self.queueDict[queueName]['Site'] = site
self.queueDict[queueName]['QueueName'] = queue
self.queueDict[queueName]['Platform'] = platform
result = self.queueDict[queueName]['CE'].isValid()
if not result['OK']:
self.log.fatal( result['Message'] )
return result
if 'BundleProxy' in self.queueDict[queueName]['ParametersDict']:
if self.queueDict[queueName]['ParametersDict']['BundleProxy'].lower() in ['true','yes','1']:
self.queueDict[queueName]['BundleProxy'] = True
elif 'BundleProxy' in ceDict:
if ceDict['BundleProxy'].lower() in ['true','yes','1']:
self.queueDict[queueName]['BundleProxy'] = True
if site not in self.sites:
self.sites.append( site )
return S_OK()
def __getGridJobOutput(self,pilotReference):
""" Get the pilot job standard output and standard error files for the Grid
job reference
"""
result = pilotDB.getPilotInfo(pilotReference)
if not result['OK'] or not result[ 'Value' ]:
return S_ERROR('Failed to get info for pilot ' + pilotReference)
pilotDict = result['Value'][pilotReference]
owner = pilotDict['OwnerDN']
group = pilotDict['OwnerGroup']
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
result = pilotDB.getPilotOutput(pilotReference)
if result['OK']:
stdout = result['Value']['StdOut']
error = result['Value']['StdErr']
if stdout or error:
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
return S_OK(resultDict)
else:
gLogger.warn( 'Empty pilot output found for %s' % pilotReference )
gridType = pilotDict['GridType']
if gridType in ["LCG","gLite","CREAM"]:
group = getGroupOption(group,'VOMSRole',group)
ret = gProxyManager.getPilotProxyFromVOMSGroup( owner, group )
if not ret['OK']:
gLogger.error( ret['Message'] )
gLogger.error( 'Could not get proxy:', 'User "%s", Group "%s"' % ( owner, group ) )
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = ret['Value']
pilotStamp = pilotDict['PilotStamp']
result = getPilotOutput( proxy, gridType, pilotReference, pilotStamp )
if not result['OK']:
return S_ERROR('Failed to get pilot output: '+result['Message'])
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
stdout = result['StdOut']
error = result['StdErr']
fileList = result['FileList']
if stdout:
result = pilotDB.storePilotOutput(pilotReference,stdout,error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = fileList
return S_OK(resultDict)
else:
# Instantiate the appropriate CE
ceFactory = ComputingElementFactory()
result = getQueue( pilotDict['GridSite'], pilotDict['DestinationSite'], pilotDict['Queue'] )
if not result['OK']:
return result
queueDict = result['Value']
result = ceFactory.getCE( gridType, pilotDict['DestinationSite'], queueDict )
if not result['OK']:
return result
ce = result['Value']
pilotStamp = pilotDict['PilotStamp']
pRef = pilotReference
if pilotStamp:
pRef = pRef + ':::' + pilotStamp
result = ce.getJobOutput( pRef )
if not result['OK']:
return result
stdout,error = result['Value']
if stdout:
result = pilotDB.storePilotOutput(pilotReference,stdout,error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
return S_OK( resultDict )
def export_killPilot(pilotRefList):
""" Kill the specified pilots
"""
# Make a list if it is not yet
pilotRefs = list(pilotRefList)
if type(pilotRefList) in StringTypes:
pilotRefs = [pilotRefList]
# Regroup pilots per site and per owner
pilotRefDict = {}
for pilotReference in pilotRefs:
result = pilotDB.getPilotInfo(pilotReference)
if not result['OK'] or not result['Value']:
return S_ERROR('Failed to get info for pilot ' +
pilotReference)
pilotDict = result['Value'][pilotReference]
owner = pilotDict['OwnerDN']
group = pilotDict['OwnerGroup']
queue = '@@@'.join([
owner, group, pilotDict['GridSite'],
pilotDict['DestinationSite'], pilotDict['Queue']
])
gridType = pilotDict['GridType']
pilotRefDict.setdefault(queue, {})
pilotRefDict[queue].setdefault('PilotList', [])
pilotRefDict[queue]['PilotList'].append(pilotReference)
pilotRefDict[queue]['GridType'] = gridType
# Do the work now queue by queue
ceFactory = ComputingElementFactory()
failed = []
for key, pilotDict in pilotRefDict.items():
owner, group, site, ce, queue = key.split('@@@')
result = getQueue(site, ce, queue)
if not result['OK']:
return result
queueDict = result['Value']
gridType = pilotDict['GridType']
result = ceFactory.getCE(gridType, ce, queueDict)
if not result['OK']:
return result
ce = result['Value']
# FIXME: quite hacky. Should be either removed, or based on some flag
if gridType in ["LCG", "gLite", "CREAM", "ARC", "Globus"]:
group = getGroupOption(group, 'VOMSRole', group)
ret = gProxyManager.getPilotProxyFromVOMSGroup(owner, group)
if not ret['OK']:
gLogger.error(ret['Message'])
gLogger.error('Could not get proxy:',
'User "%s", Group "%s"' % (owner, group))
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = ret['Value']
ce.setProxy(proxy)
pilotList = pilotDict['PilotList']
result = ce.killJob(pilotList)
if not result['OK']:
failed.extend(pilotList)
if failed:
return S_ERROR('Failed to kill at least some pilots')
return S_OK()
def __getGridJobOutput(self, pilotReference):
""" Get the pilot job standard output and standard error files for the Grid
job reference
"""
result = pilotDB.getPilotInfo(pilotReference)
if not result['OK'] or not result['Value']:
return S_ERROR('Failed to get info for pilot ' + pilotReference)
pilotDict = result['Value'][pilotReference]
owner = pilotDict['OwnerDN']
group = pilotDict['OwnerGroup']
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
result = pilotDB.getPilotOutput(pilotReference)
if result['OK']:
stdout = result['Value']['StdOut']
error = result['Value']['StdErr']
if stdout or error:
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
return S_OK(resultDict)
else:
gLogger.warn('Empty pilot output found for %s' %
pilotReference)
gridType = pilotDict['GridType']
if gridType in ["LCG", "gLite", "CREAM"]:
group = getGroupOption(group, 'VOMSRole', group)
ret = gProxyManager.getPilotProxyFromVOMSGroup(owner, group)
if not ret['OK']:
gLogger.error(ret['Message'])
gLogger.error('Could not get proxy:',
'User "%s", Group "%s"' % (owner, group))
return S_ERROR("Failed to get the pilot's owner proxy")
proxy = ret['Value']
pilotStamp = pilotDict['PilotStamp']
result = getPilotOutput(proxy, gridType, pilotReference,
pilotStamp)
if not result['OK']:
return S_ERROR('Failed to get pilot output: ' +
result['Message'])
# FIXME: What if the OutputSandBox is not StdOut and StdErr, what do we do with other files?
stdout = result['StdOut']
error = result['StdErr']
fileList = result['FileList']
if stdout:
result = pilotDB.storePilotOutput(pilotReference, stdout,
error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',
result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = fileList
return S_OK(resultDict)
else:
# Instantiate the appropriate CE
ceFactory = ComputingElementFactory()
result = Resources(group=group).getQueueDescription(
pilotDict['Queue'])
if not result['OK']:
return result
queueDict = result['Value']
result = ceFactory.getCE(gridType, pilotDict['DestinationSite'],
queueDict)
if not result['OK']:
return result
ce = result['Value']
pilotStamp = pilotDict['PilotStamp']
pRef = pilotReference
if pilotStamp:
pRef = pRef + ':::' + pilotStamp
result = ce.getJobOutput(pRef)
if not result['OK']:
return result
stdout, error = result['Value']
if stdout:
result = pilotDB.storePilotOutput(pilotReference, stdout,
error)
if not result['OK']:
gLogger.error('Failed to store pilot output:',
result['Message'])
resultDict = {}
resultDict['StdOut'] = stdout
resultDict['StdErr'] = error
resultDict['OwnerDN'] = owner
resultDict['OwnerGroup'] = group
resultDict['FileList'] = []
return S_OK(resultDict)
def getQueues(self):
""" Get the list of relevant CEs and their descriptions
"""
self.queueDict = {}
ceFactory = ComputingElementFactory()
ceTypes = self.am_getOption('CETypes', [])
ceConfList = self.am_getOption('CEs', [])
for siteName in self.siteNames:
# Look up CE definitions in the site CS description
ceList = []
gridType = siteName.split('.')[0]
result = gConfig.getSections('/Resources/Sites/%s/%s/CEs' %
(gridType, siteName))
if not result['OK']:
return S_ERROR('Failed to look up the CS for the site %s CEs' %
siteName)
if not result['Value']:
return S_ERROR('No CEs found for site %s' % siteName)
ceTotalList = result['Value']
for ce in ceTotalList:
if (ceConfList and ce in ceConfList) or not ceConfList:
ceType = gConfig.getValue(
'/Resources/Sites/%s/%s/CEs/%s/CEType' %
(gridType, siteName, ce), 'Unknown')
result = gConfig.getOptionsDict(
'/Resources/Sites/%s/%s/CEs/%s' %
(gridType, siteName, ce))
if not result['OK']:
return S_ERROR('Failed to look up the CS for ce %s' %
ce)
ceDict = result['Value']
if "SubmissionMode" in ceDict and ceDict[
'SubmissionMode'].lower() == "direct":
if ceType in ceTypes:
ceList.append((ce, ceType, ceDict))
for ce, ceType, ceDict in ceList:
section = '/Resources/Sites/%s/%s/CEs/%s/Queues' % (
gridType, siteName, ce)
result = gConfig.getSections(section)
if not result['OK']:
return S_ERROR('Failed to look up the CS for queues')
if not result['Value']:
return S_ERROR('No Queues found for site %s, ce %s' %
(siteName, ce))
queues = result['Value']
for queue in queues:
result = gConfig.getOptionsDict('%s/%s' % (section, queue))
if not result['OK']:
return S_ERROR(
'Failed to look up the CS for ce,queue %s,%s' %
(ce, queue))
queueName = '%s_%s' % (ce, queue)
self.queueDict[queueName] = {}
self.queueDict[queueName]['ParametersDict'] = result[
'Value']
self.queueDict[queueName]['ParametersDict'][
'Queue'] = queue
self.queueDict[queueName]['ParametersDict'][
'Site'] = siteName
self.queueDict[queueName]['ParametersDict'][
'GridEnv'] = self.gridEnv
self.queueDict[queueName]['ParametersDict'][
'Setup'] = gConfig.getValue('/DIRAC/Setup', 'unknown')
# Evaluate the CPU limit of the queue according to the Glue convention
# To Do: should be a utility
if "maxCPUTime" in self.queueDict[queueName]['ParametersDict'] and \
"SI00" in self.queueDict[queueName]['ParametersDict']:
maxCPUTime = float(self.queueDict[queueName]
['ParametersDict']['maxCPUTime'])
# For some sites there are crazy values in the CS
maxCPUTime = max(maxCPUTime, 0)
maxCPUTime = min(maxCPUTime, 86400 * 12.5)
si00 = float(self.queueDict[queueName]
['ParametersDict']['SI00'])
queueCPUTime = 60. / 250. * maxCPUTime * si00
self.queueDict[queueName]['ParametersDict'][
'CPUTime'] = int(queueCPUTime)
qwDir = os.path.join(self.workingDirectory, queue)
if not os.path.exists(qwDir):
os.makedirs(qwDir)
self.queueDict[queueName]['ParametersDict'][
'WorkingDirectory'] = qwDir
queueDict = dict(ceDict)
queueDict.update(
self.queueDict[queueName]['ParametersDict'])
result = ceFactory.getCE(ceName=ce,
ceType=ceType,
ceParametersDict=queueDict)
if not result['OK']:
return result
self.queueDict[queueName]['CE'] = result['Value']
self.queueDict[queueName]['CEName'] = ce
self.queueDict[queueName]['CEType'] = ceType
self.queueDict[queueName]['Site'] = siteName
self.queueDict[queueName]['QueueName'] = queue
result = self.queueDict[queueName]['CE'].isValid()
if not result['OK']:
self.log.fatal(result['Message'])
return result
if 'BundleProxy' in self.queueDict[queueName][
'ParametersDict']:
self.queueDict[queueName]['BundleProxy'] = True
return S_OK()
