class GridPilotDirector( PilotDirector ):
"""
Base Grid PilotDirector class
Derived classes must declare:
self.Middleware: It must correspond to the string before "PilotDirector".
(For proper naming of the logger)
self.ResourceBrokers: list of Brokers used by the Director.
(For proper error reporting)
"""
def __init__( self, submitPool ):
"""
Define some defaults and call parent __init__
"""
self.gridEnv = GRIDENV
self.cpuPowerRef = CPU_POWER_REF
self.requirements = REQUIREMENTS
self.rank = RANK
self.fuzzyRank = FUZZY_RANK
self.__failingWMSCache = DictCache()
self.__ticketsWMSCache = DictCache()
self.__listMatchWMSCache = DictCache()
PilotDirector.__init__( self, submitPool )
def configure( self, csSection, submitPool ):
"""
Here goes common configuration for all Grid PilotDirectors
"""
PilotDirector.configure( self, csSection, submitPool )
self.reloadConfiguration( csSection, submitPool )
self.__failingWMSCache.purgeExpired()
self.__ticketsWMSCache.purgeExpired()
for rb in self.__failingWMSCache.getKeys():
if rb in self.resourceBrokers:
try:
self.resourceBrokers.remove( rb )
except:
pass
self.resourceBrokers = List.randomize( self.resourceBrokers )
if self.gridEnv:
self.log.info( ' GridEnv: ', self.gridEnv )
if self.resourceBrokers:
self.log.info( ' ResourceBrokers:', ', '.join( self.resourceBrokers ) )
def configureFromSection( self, mySection ):
"""
reload from CS
"""
PilotDirector.configureFromSection( self, mySection )
self.gridEnv = gConfig.getValue( mySection + '/GridEnv', self.gridEnv )
if not self.gridEnv:
# No specific option found, try a general one
setup = gConfig.getValue( '/DIRAC/Setup', '' )
if setup:
instance = gConfig.getValue( '/DIRAC/Setups/%s/WorkloadManagement' % setup, '' )
if instance:
self.gridEnv = gConfig.getValue( '/Systems/WorkloadManagement/%s/GridEnv' % instance, '' )
self.resourceBrokers = gConfig.getValue( mySection + '/ResourceBrokers' , self.resourceBrokers )
self.cpuPowerRef = gConfig.getValue( mySection + '/CPUPowerRef' , self.cpuPowerRef )
self.requirements = gConfig.getValue( mySection + '/Requirements' , self.requirements )
self.rank = gConfig.getValue( mySection + '/Rank' , self.rank )
self.fuzzyRank = gConfig.getValue( mySection + '/FuzzyRank' , self.fuzzyRank )
def _submitPilots( self, workDir, taskQueueDict, pilotOptions, pilotsToSubmit,
ceMask, submitPrivatePilot, privateTQ, proxy, pilotsPerJob ):
"""
This method does the actual pilot submission to the Grid RB
The logic is as follows:
- If there are no available RB it return error
- If there is no VOMS extension in the proxy, return error
- It creates a temp directory
- Prepare a JDL
it has some part common to gLite and LCG (the payload description)
it has some part specific to each middleware
"""
taskQueueID = taskQueueDict['TaskQueueID']
# ownerDN = taskQueueDict['OwnerDN']
credDict = proxy.getCredentials()['Value']
ownerDN = credDict['identity']
ownerGroup = credDict[ 'group' ]
if not self.resourceBrokers:
# Since we can exclude RBs from the list, it may become empty
return S_ERROR( ERROR_RB )
# Need to get VOMS extension for the later interactions with WMS
ret = gProxyManager.getVOMSAttributes( proxy )
if not ret['OK']:
self.log.error( ERROR_VOMS, ret['Message'] )
return S_ERROR( ERROR_VOMS )
if not ret['Value']:
return S_ERROR( ERROR_VOMS )
vomsGroup = ret['Value'][0]
workingDirectory = tempfile.mkdtemp( prefix = 'TQ_%s_' % taskQueueID, dir = workDir )
self.log.verbose( 'Using working Directory:', workingDirectory )
# Write JDL
retDict = self._prepareJDL( taskQueueDict, workingDirectory, pilotOptions, pilotsPerJob,
ceMask, submitPrivatePilot, privateTQ )
jdl = retDict['JDL']
pilotRequirements = retDict['Requirements']
rb = retDict['RB']
if not jdl:
try:
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR( ERROR_JDL )
# Check that there are available queues for the Job:
if self.enableListMatch:
availableCEs = []
now = Time.dateTime()
availableCEs = self.listMatchCache.get( pilotRequirements )
if availableCEs == False:
availableCEs = self._listMatch( proxy, jdl, taskQueueID, rb )
if availableCEs != False:
self.log.verbose( 'LastListMatch', now )
self.log.verbose( 'AvailableCEs ', availableCEs )
self.listMatchCache.add( pilotRequirements, self.listMatchDelay * 60,
value = availableCEs ) # it is given in minutes
if not availableCEs:
try:
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR( ERROR_CE + ' TQ: %d' % taskQueueID )
# Now we are ready for the actual submission, so
self.log.verbose( 'Submitting Pilots for TaskQueue', taskQueueID )
submitRet = self._submitPilot( proxy, pilotsPerJob, jdl, taskQueueID, rb )
try:
shutil.rmtree( workingDirectory )
except:
pass
if not submitRet:
return S_ERROR( 'Pilot Submission Failed for TQ %d ' % taskQueueID )
# pilotReference, resourceBroker = submitRet
submittedPilots = 0
if pilotsPerJob != 1 and len( submitRet ) != pilotsPerJob:
# Parametric jobs are used
for pilotReference, resourceBroker in submitRet:
pilotReference = self._getChildrenReferences( proxy, pilotReference, taskQueueID )
submittedPilots += len( pilotReference )
pilotAgentsDB.addPilotTQReference( pilotReference, taskQueueID, ownerDN,
ownerGroup, resourceBroker, self.gridMiddleware,
pilotRequirements )
else:
for pilotReference, resourceBroker in submitRet:
pilotReference = [pilotReference]
submittedPilots += len( pilotReference )
pilotAgentsDB.addPilotTQReference( pilotReference, taskQueueID, ownerDN,
ownerGroup, resourceBroker, self.gridMiddleware, pilotRequirements )
# add some sleep here
time.sleep( 0.1 * submittedPilots )
if pilotsToSubmit > pilotsPerJob:
# Additional submissions are necessary, need to get a new token and iterate.
pilotsToSubmit -= pilotsPerJob
result = gProxyManager.requestToken( ownerDN, ownerGroup, max( pilotsToSubmit, self.maxJobsInFillMode ) )
if not result[ 'OK' ]:
self.log.error( ERROR_TOKEN, result['Message'] )
result = S_ERROR( ERROR_TOKEN )
result['Value'] = submittedPilots
return result
( token, numberOfUses ) = result[ 'Value' ]
for option in pilotOptions:
if option.find( '-o /Security/ProxyToken=' ) == 0:
pilotOptions.remove( option )
pilotOptions.append( '-o /Security/ProxyToken=%s' % token )
pilotsPerJob = max( 1, min( pilotsPerJob, int( numberOfUses / self.maxJobsInFillMode ) ) )
result = self._submitPilots( workDir, taskQueueDict, pilotOptions,
pilotsToSubmit, ceMask,
submitPrivatePilot, privateTQ,
proxy, pilotsPerJob )
if not result['OK']:
if 'Value' not in result:
result['Value'] = 0
result['Value'] += submittedPilots
return result
submittedPilots += result['Value']
return S_OK( submittedPilots )
def _prepareJDL( self, taskQueueDict, workingDirectory, pilotOptions, pilotsToSubmit, ceMask, submitPrivatePilot, privateTQ ):
"""
This method should be overridden in a subclass
"""
self.log.error( '_prepareJDL() method should be implemented in a subclass' )
sys.exit()
def _JobJDL( self, taskQueueDict, pilotOptions, ceMask ):
"""
The Job JDL is the same for LCG and GLite
"""
pilotJDL = 'Executable = "%s";\n' % os.path.basename( self.pilot )
executable = self.pilot
pilotJDL += 'Arguments = "%s";\n' % ' '.join( pilotOptions )
pilotJDL += 'CPUTimeRef = %s;\n' % taskQueueDict['CPUTime']
pilotJDL += 'CPUPowerRef = %s;\n' % self.cpuPowerRef
pilotJDL += """CPUWorkRef = real( CPUTimeRef * CPUPowerRef );
Lookup = "CPUScalingReferenceSI00=*";
cap = isList( other.GlueCECapability ) ? other.GlueCECapability : { "dummy" };
i0 = regexp( Lookup, cap[0] ) ? 0 : undefined;
i1 = isString( cap[1] ) && regexp( Lookup, cap[1] ) ? 1 : i0;
i2 = isString( cap[2] ) && regexp( Lookup, cap[2] ) ? 2 : i1;
i3 = isString( cap[3] ) && regexp( Lookup, cap[3] ) ? 3 : i2;
i4 = isString( cap[4] ) && regexp( Lookup, cap[4] ) ? 4 : i3;
i5 = isString( cap[5] ) && regexp( Lookup, cap[5] ) ? 5 : i4;
index = isString( cap[6] ) && regexp( Lookup, cap[6] ) ? 6 : i5;
i = isUndefined( index ) ? 0 : index;
QueuePowerRef = real( ! isUndefined( index ) ? int( substr( cap[i], size( Lookup ) - 1 ) ) : other.GlueHostBenchmarkSI00 );
QueueTimeRef = real( other.GlueCEPolicyMaxCPUTime * 60 );
QueueWorkRef = QueuePowerRef * QueueTimeRef;
"""
requirements = list( self.requirements )
if 'GridCEs' in taskQueueDict and taskQueueDict['GridCEs']:
# if there an explicit Grig CE requested by the TQ, remove the Ranking requirement
for req in self.requirements:
if req.strip().lower()[:6] == 'rank >':
requirements.remove( req )
requirements.append( 'QueueWorkRef > CPUWorkRef' )
siteRequirements = '\n || '.join( [ 'other.GlueCEInfoHostName == "%s"' % s for s in ceMask ] )
requirements.append( "( %s\n )" % siteRequirements )
pilotRequirements = '\n && '.join( requirements )
pilotJDL += 'pilotRequirements = %s;\n' % pilotRequirements
pilotJDL += 'Rank = %s;\n' % self.rank
pilotJDL += 'FuzzyRank = %s;\n' % self.fuzzyRank
pilotJDL += 'StdOutput = "%s";\n' % outputSandboxFiles[0]
pilotJDL += 'StdError = "%s";\n' % outputSandboxFiles[1]
pilotJDL += 'InputSandbox = { "%s" };\n' % '", "'.join( [ self.install, executable ] )
pilotJDL += 'OutputSandbox = { %s };\n' % ', '.join( [ '"%s"' % f for f in outputSandboxFiles ] )
self.log.verbose( pilotJDL )
return ( pilotJDL, pilotRequirements )
def parseListMatchStdout( self, proxy, cmd, taskQueueID, rb ):
"""
Parse List Match stdout to return list of matched CE's
"""
self.log.verbose( 'Executing List Match for TaskQueue', taskQueueID )
start = time.time()
ret = executeGridCommand( proxy, cmd, self.gridEnv )
if not ret['OK']:
self.log.error( 'Failed to execute List Match:', ret['Message'] )
self.__sendErrorMail( rb, 'List Match', cmd, ret, proxy )
return False
if ret['Value'][0] != 0:
self.log.error( 'Error executing List Match:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.__sendErrorMail( rb, 'List Match', cmd, ret, proxy )
return False
self.log.info( 'List Match Execution Time: %.2f for TaskQueue %d' % ( ( time.time() - start ), taskQueueID ) )
stdout = ret['Value'][1]
stderr = ret['Value'][2]
availableCEs = []
# Parse std.out
for line in List.fromChar( stdout, '\n' ):
if re.search( '/jobmanager-', line ) or re.search( '/cream-', line ):
# TODO: the line has to be stripped from extra info
availableCEs.append( line )
if not availableCEs:
self.log.info( 'List-Match failed to find CEs for TaskQueue', taskQueueID )
self.log.info( stdout )
self.log.info( stderr )
else:
self.log.debug( 'List-Match returns:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.log.info( 'List-Match found %s CEs for TaskQueue' % len( availableCEs ), taskQueueID )
self.log.verbose( ', '.join( availableCEs ) )
return availableCEs
def parseJobSubmitStdout( self, proxy, cmd, taskQueueID, rb ):
"""
Parse Job Submit stdout to return pilot reference
"""
start = time.time()
self.log.verbose( 'Executing Job Submit for TaskQueue', taskQueueID )
ret = executeGridCommand( proxy, cmd, self.gridEnv )
if not ret['OK']:
self.log.error( 'Failed to execute Job Submit:', ret['Message'] )
self.__sendErrorMail( rb, 'Job Submit', cmd, ret, proxy )
return False
if ret['Value'][0] != 0:
self.log.error( 'Error executing Job Submit:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.__sendErrorMail( rb, 'Job Submit', cmd, ret, proxy )
return False
self.log.info( 'Job Submit Execution Time: %.2f for TaskQueue %d' % ( ( time.time() - start ), taskQueueID ) )
stdout = ret['Value'][1]
stderr = ret['Value'][2]
submittedPilot = None
failed = 1
rb = ''
for line in List.fromChar( stdout, '\n' ):
m = re.search( "(https:\S+)", line )
if ( m ):
glite_id = m.group( 1 )
submittedPilot = glite_id
if not rb:
m = re.search( "https://(.+):.+", glite_id )
rb = m.group( 1 )
failed = 0
if failed:
self.log.error( 'Job Submit returns no Reference:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
return False
self.log.info( 'Reference %s for TaskQueue %s' % ( glite_id, taskQueueID ) )
return glite_id, rb
def _writeJDL( self, filename, jdlList ):
try:
f = open( filename, 'w' )
f.write( '\n'.join( jdlList ) )
f.close()
except Exception, x:
self.log.exception()
return ''
return filename
class DIRACPilotDirector(PilotDirector):
"""
DIRAC PilotDirector class
"""
def __init__( self, submitPool ):
"""
Define some defaults and call parent __init__
"""
self.gridMiddleware = 'DIRAC'
PilotDirector.__init__( self, submitPool )
self.computingElementList = COMPUTING_ELEMENTS
self.computingElementDict = {}
self.addComputingElement( self.computingElementList )
self.siteName = gConfig.getValue('/LocalSite/Site','')
if not self.siteName:
self.log.error( 'Can not run a Director if Site Name is not defined' )
sys.exit()
self.__failingCECache = DictCache()
self.__ticketsCECache = DictCache()
def configure(self, csSection, submitPool ):
"""
Here goes common configuration for DIRAC PilotDirector
"""
PilotDirector.configure( self, csSection, submitPool )
self.reloadConfiguration( csSection, submitPool )
self.__failingCECache.purgeExpired()
self.__ticketsCECache.purgeExpired()
for ce in self.__failingCECache.getKeys():
if ce in self.computingElementDict.keys():
try:
del self.computingElementDict[ce]
except:
pass
if self.computingElementDict:
self.log.info( ' ComputingElements:', ', '.join(self.computingElementDict.keys()) )
else:
return
# FIXME: this is to start testing
ceName, computingElementDict = self.computingElementDict.items()[0]
self.computingElement = computingElementDict['CE']
self.log.debug( self.computingElement.getDynamicInfo() )
self.log.info( ' SiteName:', self.siteName )
def configureFromSection( self, mySection ):
"""
reload from CS
"""
PilotDirector.configureFromSection( self, mySection )
self.computingElementList = gConfig.getValue( mySection+'/ComputingElements' , self.computingElementList )
self.addComputingElement( self.computingElementList )
self.siteName = gConfig.getValue( mySection+'/SiteName' , self.siteName )
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 _submitPilots( self, workDir, taskQueueDict, pilotOptions, pilotsToSubmit,
ceMask, submitPrivatePilot, privateTQ, proxy, pilotsPerJob ):
"""
This method does the actual pilot submission to the DIRAC CE
The logic is as follows:
- If there are no available CE it return error
- If there is no queue available in the CE's, it returns error
- It creates a temp directory
- It prepare a PilotScript
"""
taskQueueID = taskQueueDict['TaskQueueID']
ownerDN = taskQueueDict['OwnerDN']
submittedPilots = 0
# if self.computingElement not in self.computingElementDict:
# # Since we can exclude CEs from the list, it may become empty
# return S_ERROR( ERROR_CE )
pilotRequirements = []
pilotRequirements.append( ( 'CPUTime', taskQueueDict['CPUTime'] ) )
# do we need to care about anything else?
pilotRequirementsString = str( pilotRequirements )
# Check that there are available queues for the Jobs:
if self.enableListMatch:
availableQueues = []
# now = Time.dateTime()
cachedAvailableQueues = self.listMatchCache.get( pilotRequirementsString )
if cachedAvailableQueues == False:
availableQueues = self._listQueues( pilotRequirements )
if availableQueues != False:
self.listMatchCache.add( pilotRequirementsString, self.listMatchDelay, availableQueues )
self.log.verbose( 'Available Queues for TaskQueue ', "%s: %s" % ( taskQueueID, str(availableQueues) ) )
else:
availableQueues = cachedAvailableQueues
if not availableQueues:
return S_ERROR( ERROR_CE + ' TQ: %d' % taskQueueID )
baseDir = os.getcwd()
workingDirectory = tempfile.mkdtemp( prefix= 'TQ_%s_' % taskQueueID, dir = workDir )
self.log.verbose( 'Using working Directory:', workingDirectory )
os.chdir( workingDirectory )
# set the Site Name
pilotOptions.append( "-n '%s'" % self.siteName)
# submit pilots for every CE available
for CE in self.computingElementDict.keys():
ceName = CE
computingElement = self.computingElementDict[CE]['CE']
# add possible requirements from Site and CE
for req, val in getResourceDict( ceName ).items():
pilotOptions.append( "-o '/AgentJobRequirements/%s=%s'" % ( req, val ) )
ceConfigDict = self.computingElementDict[CE]
if 'ClientPlatform' in ceConfigDict:
pilotOptions.append( "-p '%s'" % ceConfigDict['ClientPlatform'])
if 'SharedArea' in ceConfigDict:
pilotOptions.append( "-o '/LocalSite/SharedArea=%s'" % ceConfigDict['SharedArea'] )
if 'CPUScalingFactor' in ceConfigDict:
pilotOptions.append( "-o '/LocalSite/CPUScalingFactor=%s'" % ceConfigDict['CPUScalingFactor'] )
if 'CPUNormalizationFactor' in ceConfigDict:
pilotOptions.append( "-o '/LocalSite/CPUNormalizationFactor=%s'" % ceConfigDict['CPUNormalizationFactor'] )
self.log.info( "pilotOptions: ", ' '.join(pilotOptions))
httpProxy = ''
if 'HttpProxy' in ceConfigDict:
httpProxy = ceConfigDict['HttpProxy']
pilotDir = ''
if 'JobExecDir' in ceConfigDict:
pilotExecDir = ceConfigDict['JobExecDir']
try:
pilotScript = self._writePilotScript( workingDirectory, pilotOptions, proxy, httpProxy, pilotExecDir )
except:
self.log.exception( ERROR_SCRIPT )
try:
os.chdir( baseDir )
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR( ERROR_SCRIPT )
self.log.info("Pilots to submit: ", pilotsToSubmit)
while submittedPilots < pilotsToSubmit:
# Find out how many pilots can be submitted
ret = computingElement.available( )
if not ret['OK']:
self.log.error('Can not determine if pilot should be submitted: ', ret['Message'])
break
maxPilotsToSubmit = ret['Value']
self.log.info("Submit Pilots: ", maxPilotsToSubmit)
if not maxPilotsToSubmit:
break
# submit the pilots and then check again
for i in range( min(maxPilotsToSubmit,pilotsToSubmit-submittedPilots) ):
submission = computingElement.submitJob(pilotScript, '', '')
if not submission['OK']:
self.log.error('Pilot submission failed: ', submission['Message'])
# cleanup
try:
os.chdir( baseDir )
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR('Pilot submission failed after ' + str(submittedPilots) + ' pilots submitted successful')
submittedPilots += 1
# let the batch system some time to digest the submitted job
time.sleep(1)
#next CE
try:
os.chdir( baseDir )
shutil.rmtree( workingDirectory )
except:
pass
return S_OK(submittedPilots)
def _listQueues( self, pilotRequirements ):
"""
For each defined CE return the list of Queues with available, running and waiting slots,
matching the requirements of the pilots.
Currently only CPU time is considered
"""
availableQueues = []
result = self.computingElement.available( pilotRequirements )
if not result['OK']:
self.log.error( 'Can not determine available queues', result['Message'] )
return False
return result['Value']
def _writePilotScript( self, workingDirectory, pilotOptions, proxy, httpProxy, pilotExecDir ):
"""
Prepare the script to execute the pilot
For the moment it will do like Grid Pilots, a full DIRAC installation
It assumes that the pilot script will have access to the submit working directory
"""
try:
compressedAndEncodedProxy = base64.encodestring( bz2.compress( proxy.dumpAllToString()['Value'] ) ).replace('\n','')
compressedAndEncodedPilot = base64.encodestring( bz2.compress( open( self.pilot, "rb" ).read(), 9 ) ).replace('\n','')
compressedAndEncodedInstall = base64.encodestring( bz2.compress( open( self.install, "rb" ).read(), 9 ) ).replace('\n','')
except:
self.log.exception('Exception during file compression of proxy, dirac-pilot or dirac-install')
return S_ERROR('Exception during file compression of proxy, dirac-pilot or dirac-install')
localPilot = """#!/bin/bash
/usr/bin/env python << EOF
#
import os, tempfile, sys, shutil, base64, bz2
try:
pilotExecDir = '%(pilotExecDir)s'
if not pilotExecDir:
pilotExecDir = None
pilotWorkingDirectory = tempfile.mkdtemp( suffix = 'pilot', prefix = 'DIRAC_', dir = pilotExecDir )
os.chdir( pilotWorkingDirectory )
open( 'proxy', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedProxy)s" ) ) )
open( '%(pilotScript)s', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedPilot)s" ) ) )
open( '%(installScript)s', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedInstall)s" ) ) )
os.chmod("proxy",0600)
os.chmod("%(pilotScript)s",0700)
os.chmod("%(installScript)s",0700)
if "LD_LIBRARY_PATH" not in os.environ:
os.environ["LD_LIBRARY_PATH"]=""
os.environ["X509_USER_PROXY"]=os.path.join(pilotWorkingDirectory, 'proxy')
if "%(httpProxy)s":
os.environ["HTTP_PROXY"]="%(httpProxy)s"
os.environ["X509_CERT_DIR"]=os.path.join(pilotWorkingDirectory, 'etc/grid-security/certificates')
# TODO: structure the output
print '==========================================================='
print 'Environment of execution host'
for key in os.environ.keys():
print key + '=' + os.environ[key]
print '==========================================================='
except Exception, x:
print >> sys.stderr, x
sys.exit(-1)
cmd = "python %(pilotScript)s %(pilotOptions)s"
print 'Executing: ', cmd
sys.stdout.flush()
os.system( cmd )
shutil.rmtree( pilotWorkingDirectory )
EOF
""" % { 'compressedAndEncodedProxy': compressedAndEncodedProxy,
'compressedAndEncodedPilot': compressedAndEncodedPilot,
'compressedAndEncodedInstall': compressedAndEncodedInstall,
'httpProxy': httpProxy,
'pilotScript': os.path.basename(self.pilot),
'installScript': os.path.basename(self.install),
'pilotOptions': ' '.join( pilotOptions ),
'pilotExecDir': pilotExecDir }
fd, name = tempfile.mkstemp( suffix = '_pilotwrapper.py', prefix = 'DIRAC_', dir=workingDirectory)
pilotWrapper = os.fdopen(fd, 'w')
pilotWrapper.write( localPilot )
pilotWrapper.close()
return name
def _getPilotProxyFromDIRACGroup( self, ownerDN, ownerGroup, requiredTimeLeft ):
"""
Download a limited pilot proxy with VOMS extensions depending on the group
"""
#Assign VOMS attribute
vomsAttr = CS.getVOMSAttributeForGroup( ownerGroup )
if not vomsAttr:
self.log.info( "Downloading a proxy without VOMS extensions" )
return gProxyManager.downloadProxy( ownerDN, ownerGroup, limited = True,
requiredTimeLeft = requiredTimeLeft )
else:
self.log.info( "Downloading a proxy with '%s' VOMS extension" % vomsAttr )
return gProxyManager.downloadVOMSProxy( ownerDN,
ownerGroup,
limited = True,
requiredTimeLeft = requiredTimeLeft,
requiredVOMSAttribute = vomsAttr )
class DIRACPilotDirector(PilotDirector):
"""
DIRAC PilotDirector class
"""
def __init__(self, submitPool):
"""
Define some defaults and call parent __init__
"""
self.gridMiddleware = 'DIRAC'
PilotDirector.__init__(self, submitPool)
self.computingElementList = COMPUTING_ELEMENTS
self.computingElementDict = {}
self.addComputingElement(self.computingElementList)
self.siteName = gConfig.getValue('/LocalSite/Site', '')
if not self.siteName:
self.log.error(
'Can not run a Director if Site Name is not defined')
sys.exit()
self.__failingCECache = DictCache()
self.__ticketsCECache = DictCache()
def configure(self, csSection, submitPool):
"""
Here goes common configuration for DIRAC PilotDirector
"""
PilotDirector.configure(self, csSection, submitPool)
self.reloadConfiguration(csSection, submitPool)
self.__failingCECache.purgeExpired()
self.__ticketsCECache.purgeExpired()
for ce in self.__failingCECache.getKeys():
if ce in self.computingElementDict.keys():
try:
del self.computingElementDict[ce]
except:
pass
if self.computingElementDict:
self.log.info(' ComputingElements:',
', '.join(self.computingElementDict.keys()))
else:
return
# FIXME: this is to start testing
ceName, computingElementDict = self.computingElementDict.items()[0]
self.computingElement = computingElementDict['CE']
self.log.debug(self.computingElement.getDynamicInfo())
self.log.info(' SiteName:', self.siteName)
def configureFromSection(self, mySection):
"""
reload from CS
"""
PilotDirector.configureFromSection(self, mySection)
self.computingElementList = gConfig.getValue(
mySection + '/ComputingElements', self.computingElementList)
self.addComputingElement(self.computingElementList)
self.siteName = gConfig.getValue(mySection + '/SiteName',
self.siteName)
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 _submitPilots(self, workDir, taskQueueDict, pilotOptions,
pilotsToSubmit, ceMask, submitPrivatePilot, privateTQ,
proxy, pilotsPerJob):
"""
This method does the actual pilot submission to the DIRAC CE
The logic is as follows:
- If there are no available CE it return error
- If there is no queue available in the CE's, it returns error
- It creates a temp directory
- It prepare a PilotScript
"""
taskQueueID = taskQueueDict['TaskQueueID']
ownerDN = taskQueueDict['OwnerDN']
submittedPilots = 0
# if self.computingElement not in self.computingElementDict:
# # Since we can exclude CEs from the list, it may become empty
# return S_ERROR( ERROR_CE )
pilotRequirements = []
pilotRequirements.append(('CPUTime', taskQueueDict['CPUTime']))
# do we need to care about anything else?
pilotRequirementsString = str(pilotRequirements)
# Check that there are available queues for the Jobs:
if self.enableListMatch:
availableQueues = []
# now = Time.dateTime()
cachedAvailableQueues = self.listMatchCache.get(
pilotRequirementsString)
if cachedAvailableQueues == False:
availableQueues = self._listQueues(pilotRequirements)
if availableQueues != False:
self.listMatchCache.add(pilotRequirementsString,
self.listMatchDelay,
availableQueues)
self.log.verbose(
'Available Queues for TaskQueue ',
"%s: %s" % (taskQueueID, str(availableQueues)))
else:
availableQueues = cachedAvailableQueues
if not availableQueues:
return S_ERROR(ERROR_CE + ' TQ: %d' % taskQueueID)
baseDir = os.getcwd()
workingDirectory = tempfile.mkdtemp(prefix='TQ_%s_' % taskQueueID,
dir=workDir)
self.log.verbose('Using working Directory:', workingDirectory)
os.chdir(workingDirectory)
# set the Site Name
pilotOptions.append("-n '%s'" % self.siteName)
# submit pilots for every CE available
for CE in self.computingElementDict.keys():
ceName = CE
computingElement = self.computingElementDict[CE]['CE']
# add possible requirements from Site and CE
for req, val in getResourceDict(ceName).items():
pilotOptions.append("-o '/AgentJobRequirements/%s=%s'" %
(req, val))
ceConfigDict = self.computingElementDict[CE]
if 'ClientPlatform' in ceConfigDict:
pilotOptions.append("-p '%s'" % ceConfigDict['ClientPlatform'])
if 'SharedArea' in ceConfigDict:
pilotOptions.append("-o '/LocalSite/SharedArea=%s'" %
ceConfigDict['SharedArea'])
if 'CPUScalingFactor' in ceConfigDict:
pilotOptions.append("-o '/LocalSite/CPUScalingFactor=%s'" %
ceConfigDict['CPUScalingFactor'])
if 'CPUNormalizationFactor' in ceConfigDict:
pilotOptions.append(
"-o '/LocalSite/CPUNormalizationFactor=%s'" %
ceConfigDict['CPUNormalizationFactor'])
self.log.info("pilotOptions: ", ' '.join(pilotOptions))
httpProxy = ''
if 'HttpProxy' in ceConfigDict:
httpProxy = ceConfigDict['HttpProxy']
pilotDir = ''
if 'JobExecDir' in ceConfigDict:
pilotExecDir = ceConfigDict['JobExecDir']
try:
pilotScript = self._writePilotScript(workingDirectory,
pilotOptions, proxy,
httpProxy, pilotExecDir)
except:
self.log.exception(ERROR_SCRIPT)
try:
os.chdir(baseDir)
shutil.rmtree(workingDirectory)
except:
pass
return S_ERROR(ERROR_SCRIPT)
self.log.info("Pilots to submit: ", pilotsToSubmit)
while submittedPilots < pilotsToSubmit:
# Find out how many pilots can be submitted
ret = computingElement.available()
if not ret['OK']:
self.log.error(
'Can not determine if pilot should be submitted: ',
ret['Message'])
break
maxPilotsToSubmit = ret['Value']
self.log.info("Submit Pilots: ", maxPilotsToSubmit)
if not maxPilotsToSubmit:
break
# submit the pilots and then check again
for i in range(
min(maxPilotsToSubmit,
pilotsToSubmit - submittedPilots)):
submission = computingElement.submitJob(
pilotScript, '', '')
if not submission['OK']:
self.log.error('Pilot submission failed: ',
submission['Message'])
# cleanup
try:
os.chdir(baseDir)
shutil.rmtree(workingDirectory)
except:
pass
return S_ERROR('Pilot submission failed after ' +
str(submittedPilots) +
' pilots submitted successful')
submittedPilots += 1
# let the batch system some time to digest the submitted job
time.sleep(1)
#next CE
try:
os.chdir(baseDir)
shutil.rmtree(workingDirectory)
except:
pass
return S_OK(submittedPilots)
def _listQueues(self, pilotRequirements):
"""
For each defined CE return the list of Queues with available, running and waiting slots,
matching the requirements of the pilots.
Currently only CPU time is considered
"""
availableQueues = []
result = self.computingElement.available(pilotRequirements)
if not result['OK']:
self.log.error('Can not determine available queues',
result['Message'])
return False
return result['Value']
def _writePilotScript(self, workingDirectory, pilotOptions, proxy,
httpProxy, pilotExecDir):
"""
Prepare the script to execute the pilot
For the moment it will do like Grid Pilots, a full DIRAC installation
It assumes that the pilot script will have access to the submit working directory
"""
try:
compressedAndEncodedProxy = base64.encodestring(
bz2.compress(proxy.dumpAllToString()['Value'])).replace(
'\n', '')
compressedAndEncodedPilot = base64.encodestring(
bz2.compress(open(self.pilot, "rb").read(),
9)).replace('\n', '')
compressedAndEncodedInstall = base64.encodestring(
bz2.compress(open(self.install, "rb").read(),
9)).replace('\n', '')
except:
self.log.exception(
'Exception during file compression of proxy, dirac-pilot or dirac-install'
)
return S_ERROR(
'Exception during file compression of proxy, dirac-pilot or dirac-install'
)
localPilot = """#!/bin/bash
/usr/bin/env python << EOF
#
import os, tempfile, sys, shutil, base64, bz2
try:
pilotExecDir = '%(pilotExecDir)s'
if not pilotExecDir:
pilotExecDir = None
pilotWorkingDirectory = tempfile.mkdtemp( suffix = 'pilot', prefix = 'DIRAC_', dir = pilotExecDir )
os.chdir( pilotWorkingDirectory )
open( 'proxy', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedProxy)s" ) ) )
open( '%(pilotScript)s', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedPilot)s" ) ) )
open( '%(installScript)s', "w" ).write(bz2.decompress( base64.decodestring( "%(compressedAndEncodedInstall)s" ) ) )
os.chmod("proxy",0600)
os.chmod("%(pilotScript)s",0700)
os.chmod("%(installScript)s",0700)
if "LD_LIBRARY_PATH" not in os.environ:
os.environ["LD_LIBRARY_PATH"]=""
os.environ["X509_USER_PROXY"]=os.path.join(pilotWorkingDirectory, 'proxy')
if "%(httpProxy)s":
os.environ["HTTP_PROXY"]="%(httpProxy)s"
os.environ["X509_CERT_DIR"]=os.path.join(pilotWorkingDirectory, 'etc/grid-security/certificates')
# TODO: structure the output
print '==========================================================='
print 'Environment of execution host'
for key in os.environ.keys():
print key + '=' + os.environ[key]
print '==========================================================='
except Exception, x:
print >> sys.stderr, x
sys.exit(-1)
cmd = "python %(pilotScript)s %(pilotOptions)s"
print 'Executing: ', cmd
sys.stdout.flush()
os.system( cmd )
shutil.rmtree( pilotWorkingDirectory )
EOF
""" % {
'compressedAndEncodedProxy': compressedAndEncodedProxy,
'compressedAndEncodedPilot': compressedAndEncodedPilot,
'compressedAndEncodedInstall': compressedAndEncodedInstall,
'httpProxy': httpProxy,
'pilotScript': os.path.basename(self.pilot),
'installScript': os.path.basename(self.install),
'pilotOptions': ' '.join(pilotOptions),
'pilotExecDir': pilotExecDir
}
fd, name = tempfile.mkstemp(suffix='_pilotwrapper.py',
prefix='DIRAC_',
dir=workingDirectory)
pilotWrapper = os.fdopen(fd, 'w')
pilotWrapper.write(localPilot)
pilotWrapper.close()
return name
def _getPilotProxyFromDIRACGroup(self, ownerDN, ownerGroup,
requiredTimeLeft):
"""
Download a limited pilot proxy with VOMS extensions depending on the group
"""
#Assign VOMS attribute
vomsAttr = CS.getVOMSAttributeForGroup(ownerGroup)
if not vomsAttr:
self.log.info("Downloading a proxy without VOMS extensions")
return gProxyManager.downloadProxy(
ownerDN,
ownerGroup,
limited=True,
requiredTimeLeft=requiredTimeLeft)
else:
self.log.info("Downloading a proxy with '%s' VOMS extension" %
vomsAttr)
return gProxyManager.downloadVOMSProxy(
ownerDN,
ownerGroup,
limited=True,
requiredTimeLeft=requiredTimeLeft,
requiredVOMSAttribute=vomsAttr)
class GridPilotDirector( PilotDirector ):
"""
Base Grid PilotDirector class
Derived classes must declare:
self.Middleware: It must correspond to the string before "PilotDirector".
(For proper naming of the logger)
self.ResourceBrokers: list of Brokers used by the Director.
(For proper error reporting)
"""
def __init__( self, submitPool ):
"""
Define some defaults and call parent __init__
"""
self.gridEnv = GRIDENV
self.cpuPowerRef = CPU_POWER_REF
self.requirements = REQUIREMENTS
self.rank = RANK
self.fuzzyRank = FUZZY_RANK
self.__failingWMSCache = DictCache()
self.__ticketsWMSCache = DictCache()
self.__listMatchWMSCache = DictCache()
PilotDirector.__init__( self, submitPool )
def configure( self, csSection, submitPool ):
"""
Here goes common configuration for all Grid PilotDirectors
"""
PilotDirector.configure( self, csSection, submitPool )
self.reloadConfiguration( csSection, submitPool )
self.__failingWMSCache.purgeExpired()
self.__ticketsWMSCache.purgeExpired()
for rb in self.__failingWMSCache.getKeys():
if rb in self.resourceBrokers:
try:
self.resourceBrokers.remove( rb )
except:
pass
self.resourceBrokers = List.randomize( self.resourceBrokers )
if self.gridEnv:
self.log.info( ' GridEnv: ', self.gridEnv )
if self.resourceBrokers:
self.log.info( ' ResourceBrokers:', ', '.join( self.resourceBrokers ) )
def configureFromSection( self, mySection ):
"""
reload from CS
"""
PilotDirector.configureFromSection( self, mySection )
self.gridEnv = gConfig.getValue( mySection + '/GridEnv', self.gridEnv )
if not self.gridEnv:
# No specific option found, try a general one
setup = gConfig.getValue( '/DIRAC/Setup', '' )
if setup:
instance = gConfig.getValue( '/DIRAC/Setups/%s/WorkloadManagement' % setup, '' )
if instance:
self.gridEnv = gConfig.getValue( '/Systems/WorkloadManagement/%s/GridEnv' % instance, '' )
self.resourceBrokers = gConfig.getValue( mySection + '/ResourceBrokers' , self.resourceBrokers )
self.cpuPowerRef = gConfig.getValue( mySection + '/CPUPowerRef' , self.cpuPowerRef )
self.requirements = gConfig.getValue( mySection + '/Requirements' , self.requirements )
self.rank = gConfig.getValue( mySection + '/Rank' , self.rank )
self.fuzzyRank = gConfig.getValue( mySection + '/FuzzyRank' , self.fuzzyRank )
def _submitPilots( self, workDir, taskQueueDict, pilotOptions, pilotsToSubmit,
ceMask, submitPrivatePilot, privateTQ, proxy, pilotsPerJob ):
"""
This method does the actual pilot submission to the Grid RB
The logic is as follows:
- If there are no available RB it return error
- If there is no VOMS extension in the proxy, return error
- It creates a temp directory
- Prepare a JDL
it has some part common to gLite and LCG (the payload description)
it has some part specific to each middleware
"""
taskQueueID = taskQueueDict['TaskQueueID']
# ownerDN = taskQueueDict['OwnerDN']
ownerDN = proxy.getCredentials()['Value']['identity']
if not self.resourceBrokers:
# Since we can exclude RBs from the list, it may become empty
return S_ERROR( ERROR_RB )
# Need to get VOMS extension for the later interactions with WMS
ret = gProxyManager.getVOMSAttributes( proxy )
if not ret['OK']:
self.log.error( ERROR_VOMS, ret['Message'] )
return S_ERROR( ERROR_VOMS )
if not ret['Value']:
return S_ERROR( ERROR_VOMS )
vomsGroup = ret['Value'][0]
workingDirectory = tempfile.mkdtemp( prefix = 'TQ_%s_' % taskQueueID, dir = workDir )
self.log.verbose( 'Using working Directory:', workingDirectory )
# Write JDL
retDict = self._prepareJDL( taskQueueDict, workingDirectory, pilotOptions, pilotsPerJob,
ceMask, submitPrivatePilot, privateTQ )
jdl = retDict['JDL']
pilotRequirements = retDict['Requirements']
rb = retDict['RB']
if not jdl:
try:
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR( ERROR_JDL )
# Check that there are available queues for the Job:
if self.enableListMatch:
availableCEs = []
now = Time.dateTime()
availableCEs = self.listMatchCache.get( pilotRequirements )
if availableCEs == False:
availableCEs = self._listMatch( proxy, jdl, taskQueueID, rb )
if availableCEs != False:
self.log.verbose( 'LastListMatch', now )
self.log.verbose( 'AvailableCEs ', availableCEs )
self.listMatchCache.add( pilotRequirements, self.listMatchDelay * 60,
value = availableCEs ) # it is given in minutes
if not availableCEs:
try:
shutil.rmtree( workingDirectory )
except:
pass
return S_ERROR( ERROR_CE + ' TQ: %d' % taskQueueID )
# Now we are ready for the actual submission, so
self.log.verbose( 'Submitting Pilots for TaskQueue', taskQueueID )
submitRet = self._submitPilot( proxy, pilotsPerJob, jdl, taskQueueID, rb )
try:
shutil.rmtree( workingDirectory )
except:
pass
if not submitRet:
return S_ERROR( 'Pilot Submission Failed for TQ %d ' % taskQueueID )
# pilotReference, resourceBroker = submitRet
submittedPilots = 0
if pilotsPerJob != 1 and len( submitRet ) != pilotsPerJob:
# Parametric jobs are used
for pilotReference, resourceBroker in submitRet:
pilotReference = self._getChildrenReferences( proxy, pilotReference, taskQueueID )
submittedPilots += len( pilotReference )
pilotAgentsDB.addPilotTQReference( pilotReference, taskQueueID, ownerDN,
vomsGroup, resourceBroker, self.gridMiddleware,
pilotRequirements )
else:
for pilotReference, resourceBroker in submitRet:
pilotReference = [pilotReference]
submittedPilots += len( pilotReference )
pilotAgentsDB.addPilotTQReference( pilotReference, taskQueueID, ownerDN,
vomsGroup, broker = resourceBroker, gridType = self.gridMiddleware,
requirements = pilotRequirements )
# add some sleep here
time.sleep( 0.1 * submittedPilots )
if pilotsToSubmit > pilotsPerJob:
# Additional submissions are necessary, need to get a new token and iterate.
pilotsToSubmit -= pilotsPerJob
ownerDN = self.genericPilotDN
ownerGroup = self.genericPilotGroup
result = gProxyManager.requestToken( ownerDN, ownerGroup, max( pilotsToSubmit, self.maxJobsInFillMode ) )
if not result[ 'OK' ]:
self.log.error( ERROR_TOKEN, result['Message'] )
return S_ERROR( ERROR_TOKEN )
( token, numberOfUses ) = result[ 'Value' ]
for option in pilotOptions:
if option.find( '-o /Security/ProxyToken=' ) == 0:
pilotOptions.remove( option )
pilotOptions.append( '-o /Security/ProxyToken=%s' % token )
pilotsPerJob = min( pilotsPerJob, int( numberOfUses / self.maxJobsInFillMode ) )
result = self._submitPilots( workDir, taskQueueDict, pilotOptions,
pilotsToSubmit, ceMask,
submitPrivatePilot, privateTQ,
proxy, pilotsPerJob )
if not result['OK']:
result['Value'] = submittedPilots
return result
submittedPilots += result['Value']
return S_OK( submittedPilots )
def _prepareJDL( self, taskQueueDict, workingDirectory, pilotOptions, pilotsToSubmit, ceMask, submitPrivatePilot, privateTQ ):
"""
This method should be overridden in a subclass
"""
self.log.error( '_prepareJDL() method should be implemented in a subclass' )
sys.exit()
def _JobJDL( self, taskQueueDict, pilotOptions, ceMask ):
"""
The Job JDL is the same for LCG and GLite
"""
pilotJDL = 'Executable = "%s";\n' % os.path.basename( self.pilot )
executable = self.pilot
pilotJDL += 'Arguments = "%s";\n' % ' '.join( pilotOptions )
pilotJDL += 'CPUTimeRef = %s;\n' % taskQueueDict['CPUTime']
pilotJDL += 'CPUPowerRef = %s;\n' % self.cpuPowerRef
pilotJDL += """CPUWorkRef = real( CPUTimeRef * CPUPowerRef );
Lookup = "CPUScalingReferenceSI00=*";
cap = isList( other.GlueCECapability ) ? other.GlueCECapability : { "dummy" };
i0 = regexp( Lookup, cap[0] ) ? 0 : undefined;
i1 = isString( cap[1] ) && regexp( Lookup, cap[1] ) ? 1 : i0;
i2 = isString( cap[2] ) && regexp( Lookup, cap[2] ) ? 2 : i1;
i3 = isString( cap[3] ) && regexp( Lookup, cap[3] ) ? 3 : i2;
i4 = isString( cap[4] ) && regexp( Lookup, cap[4] ) ? 4 : i3;
i5 = isString( cap[5] ) && regexp( Lookup, cap[5] ) ? 5 : i4;
index = isString( cap[6] ) && regexp( Lookup, cap[6] ) ? 6 : i5;
i = isUndefined( index ) ? 0 : index;
QueuePowerRef = real( ! isUndefined( index ) ? int( substr( cap[i], size( Lookup ) - 1 ) ) : other.GlueHostBenchmarkSI00 );
QueueTimeRef = real( other.GlueCEPolicyMaxCPUTime * 60 );
QueueWorkRef = QueuePowerRef * QueueTimeRef;
"""
requirements = list( self.requirements )
if 'GridCEs' in taskQueueDict and taskQueueDict['GridCEs']:
# if there an explicit Grig CE requested by the TQ, remove the Ranking requirement
for req in self.requirements:
if req.strip().lower()[:6] == 'rank >':
requirements.remove( req )
requirements.append( 'QueueWorkRef > CPUWorkRef' )
siteRequirements = '\n || '.join( [ 'other.GlueCEInfoHostName == "%s"' % s for s in ceMask ] )
requirements.append( "( %s\n )" % siteRequirements )
pilotRequirements = '\n && '.join( requirements )
pilotJDL += 'pilotRequirements = %s;\n' % pilotRequirements
pilotJDL += 'Rank = %s;\n' % self.rank
pilotJDL += 'FuzzyRank = %s;\n' % self.fuzzyRank
pilotJDL += 'StdOutput = "%s";\n' % outputSandboxFiles[0]
pilotJDL += 'StdError = "%s";\n' % outputSandboxFiles[1]
pilotJDL += 'InputSandbox = { "%s" };\n' % '", "'.join( [ self.install, executable ] )
pilotJDL += 'OutputSandbox = { %s };\n' % ', '.join( [ '"%s"' % f for f in outputSandboxFiles ] )
self.log.verbose( pilotJDL )
return ( pilotJDL, pilotRequirements )
def parseListMatchStdout( self, proxy, cmd, taskQueueID, rb ):
"""
Parse List Match stdout to return list of matched CE's
"""
self.log.verbose( 'Executing List Match for TaskQueue', taskQueueID )
start = time.time()
ret = executeGridCommand( proxy, cmd, self.gridEnv )
if not ret['OK']:
self.log.error( 'Failed to execute List Match:', ret['Message'] )
self.__sendErrorMail( rb, 'List Match', cmd, ret, proxy )
return False
if ret['Value'][0] != 0:
self.log.error( 'Error executing List Match:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.__sendErrorMail( rb, 'List Match', cmd, ret, proxy )
return False
self.log.info( 'List Match Execution Time: %.2f for TaskQueue %d' % ( ( time.time() - start ), taskQueueID ) )
stdout = ret['Value'][1]
stderr = ret['Value'][2]
availableCEs = []
# Parse std.out
for line in List.fromChar( stdout, '\n' ):
if re.search( '/jobmanager-', line ) or re.search( '/cream-', line ):
# TODO: the line has to be stripped from extra info
availableCEs.append( line )
if not availableCEs:
self.log.info( 'List-Match failed to find CEs for TaskQueue', taskQueueID )
self.log.info( stdout )
self.log.info( stderr )
else:
self.log.debug( 'List-Match returns:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.log.info( 'List-Match found %s CEs for TaskQueue' % len( availableCEs ), taskQueueID )
self.log.verbose( ', '.join( availableCEs ) )
return availableCEs
def parseJobSubmitStdout( self, proxy, cmd, taskQueueID, rb ):
"""
Parse Job Submit stdout to return pilot reference
"""
start = time.time()
self.log.verbose( 'Executing Job Submit for TaskQueue', taskQueueID )
ret = executeGridCommand( proxy, cmd, self.gridEnv )
if not ret['OK']:
self.log.error( 'Failed to execute Job Submit:', ret['Message'] )
self.__sendErrorMail( rb, 'Job Submit', cmd, ret, proxy )
return False
if ret['Value'][0] != 0:
self.log.error( 'Error executing Job Submit:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
self.__sendErrorMail( rb, 'Job Submit', cmd, ret, proxy )
return False
self.log.info( 'Job Submit Execution Time: %.2f for TaskQueue %d' % ( ( time.time() - start ), taskQueueID ) )
stdout = ret['Value'][1]
stderr = ret['Value'][2]
submittedPilot = None
failed = 1
rb = ''
for line in List.fromChar( stdout, '\n' ):
m = re.search( "(https:\S+)", line )
if ( m ):
glite_id = m.group( 1 )
submittedPilot = glite_id
if not rb:
m = re.search( "https://(.+):.+", glite_id )
rb = m.group( 1 )
failed = 0
if failed:
self.log.error( 'Job Submit returns no Reference:', str( ret['Value'][0] ) + '\n'.join( ret['Value'][1:3] ) )
return False
self.log.info( 'Reference %s for TaskQueue %s' % ( glite_id, taskQueueID ) )
return glite_id, rb
def _writeJDL( self, filename, jdlList ):
try:
f = open( filename, 'w' )
f.write( '\n'.join( jdlList ) )
f.close()
except Exception, x:
self.log.exception()
return ''
return filename
