ViewVC Help

View File | Revision Log | Show Annotations | Root Listing

root/cvsroot/COMP/CRAB/python/Splitter.py

Comparing COMP/CRAB/python/Splitter.py (file contents):
Revision 1.29 by ewv, Thu Oct 1 22:00:40 2009 UTC vs.
Revision 1.46 by spiga, Mon Sep 27 08:56:08 2010 UTC

#	Line 3 | Line 3 | __revision__ = "$Id$"
3		__version__ = "$Revision$"
4
5		import common
6	–	from sets import Set
6		from crab_exceptions import *
7		from crab_util import *
8
#	Line 14 | Line 13 | from WMCore.DataStructs.Subscription imp
13		from WMCore.DataStructs.Workflow import Workflow
14		from WMCore.JobSplitting.SplitterFactory import SplitterFactory
15		from WMCore.SiteScreening.BlackWhiteListParser import SEBlackWhiteListParser
16	+	try: # Can remove when CMSSW 3.7 and earlier are dropped
17	+	from FWCore.PythonUtilities.LumiList import LumiList
18	+	except ImportError:
19	+	from LumiList import LumiList
20
21		class JobSplitter:
22		def __init__( self, cfg_params,  args ):
#	Line 33 | Line 36 | class JobSplitter:
36		seBlackList = cfg_params.get('GRID.se_black_list',[])
37		self.blackWhiteListParser = SEBlackWhiteListParser(self.seWhiteList, seBlackList, common.logger())
38
39	+	## check if has been asked for a non default file to store/read analyzed fileBlocks
40	+	defaultName = common.work_space.shareDir()+'AnalyzedBlocks.txt'
41	+	self.fileBlocks_FileName = os.path.abspath(self.cfg_params.get('CMSSW.fileblocks_file',defaultName))
42	+
43
44		def checkUserSettings(self):
45		## Events per job
#	Line 62 | Line 69 | class JobSplitter:
69		self.total_number_of_events = 0
70		self.selectTotalNumberEvents = 0
71
72	+	return
73
74		def checkLumiSettings(self):
75		"""
#	Line 85 | Line 93 | class JobSplitter:
93		settings += 1
94
95		if settings != 2:
96	<	msg = 'When running on analysis datasets you must specify two and only two of:\n'
96	>	msg = 'When splitting by lumi section you must specify two and only two of:\n'
97		msg += '  number_of_jobs, lumis_per_job, total_number_of_lumis'
98		raise CrabException(msg)
99		if self.limitNJobs and self.limitJobLumis:
100		self.limitTotalLumis = True
101		self.totalNLumis = self.lumisPerJob * self.theNumberOfJobs
102
103	+	# Has the user specified runselection?
104	+	if (self.cfg_params.has_key('CMSSW.runselection')):
105	+	common.logger.info('You have specified runselection and split by lumi.')
106	+	common.logger.info('Good lumi list will be the intersection of runselection and lumimask or ADS (if any).')
107	+	return
108
109		def ComputeSubBlockSites( self, blockSites ):
110		"""
#	Line 373 | Line 386 | class JobSplitter:
386		allBlock = []
387
388		blockCounter = 0
389	+	saveFblocks =''
390		for block in blocks:
391		if block in jobsOfBlock.keys() :
392		blockCounter += 1
#	Line 383 | Line 397 | class JobSplitter:
397		if len(sites) == 0:
398		noSiteBlock.append( spanRanges(jobsOfBlock[block]) )
399		bloskNoSite.append( blockCounter )
400	+	else:
401	+	saveFblocks += str(block)+'\n'
402	+	writeTXTfile(self, self.fileBlocks_FileName , saveFblocks)
403
404		common.logger.info(screenOutput)
405		if len(noSiteBlock) > 0 and len(bloskNoSite) > 0:
#	Line 411 | Line 428 | class JobSplitter:
428		common.logger.info(msg)
429
430		if bloskNoSite == allBlock:
431	<	raise CrabException('No jobs created')
431	>	msg += 'Requested jobs cannot be Created! \n'
432	>	if self.cfg_params.has_key('GRID.se_white_list'):
433	>	msg += '\tWARNING: SE White List: '+self.cfg_params['GRID.se_white_list']+'\n'
434	>	msg += '\t(Hint: By whitelisting you force the job to run at this particular site(s).\n'
435	>	msg += '\tPlease check if the dataset is available at this site!)'
436	>	if self.cfg_params.has_key('GRID.ce_white_list'):
437	>	msg += '\tWARNING: CE White List: '+self.cfg_params['GRID.ce_white_list']+'\n'
438	>	msg += '\t(Hint: By whitelisting you force the job to run at this particular site(s).\n'
439	>	msg += '\tPlease check if the dataset is available at this site!)\n'
440	>	raise CrabException(msg)
441
442		return
443
#	Line 458 | Line 484 | class JobSplitter:
484		jobfactory = splitter(subs)
485
486		#loop over all runs
461	–	set = Set(runList)
487		list_of_lists = []
488		jobDestination = []
489	+	list_of_blocks = []
490		count = 0
491		for jobGroup in  jobfactory():
492		if count <  self.theNumberOfJobs:
#	Line 472 | Line 498 | class JobSplitter:
498		list_of_lists.append([fullString,str(-1),str(0)])
499		#need to check single file location
500		jobDestination.append(res['locations'])
501	+	list_of_blocks.append(res['block'])
502		count +=1
503	<	# prepare dict output
503	>	# prepare dict output
504		dictOut = {}
505		dictOut['params']= ['InputFiles','MaxEvents','SkipEvents']
506		dictOut['args'] = list_of_lists
507		dictOut['jobDestination'] = jobDestination
508		dictOut['njobs']=count
509
510	+	self.cacheBlocks(list_of_blocks,jobDestination)
511	+
512		return dictOut
513
514		def getJobInfo( self,jobGroup ):
#	Line 493 | Line 522 | class JobSplitter:
522		for loc in file['locations']:
523		if tmp_check < 1 :
524		locations.append(loc)
525	+	res['block']= file['block']
526		tmp_check = tmp_check + 1
497	–	### qui va messo il check per la locations
527		res['lfns'] = lfns
528		res['locations'] = locations
529		return res
#	Line 539 | Line 568 | class JobSplitter:
568
569		managedGenerators =self.args['managedGenerators']
570		generator = self.args['generator']
571	<	firstRun = self.cfg_params.get('CMSSW.first_run', 1)
571	>	firstLumi = self.cfg_params.get('CMSSW.first_lumi', 1)
572
573		self.prepareSplittingNoInput()
574
#	Line 560 | Line 589 | class JobSplitter:
589		## Since there is no input, any site is good
590		jobDestination.append([""]) # must be empty to correctly write the XML
591		args=[]
592	<	if (firstRun): # Pythia first run
593	<	args.append(str(int(firstRun)+i))
592	>	if (firstLumi): # Pythia first lumi
593	>	args.append(str(int(firstLumi)+i))
594		if (generator in managedGenerators):
595		args.append(generator)
596		if (generator == 'comphep' and i == 0):
#	Line 575 | Line 604 | class JobSplitter:
604
605		dictOut = {}
606		dictOut['params'] = ['MaxEvents']
607	<	if (firstRun):
608	<	dictOut['params'] = ['FirstRun','MaxEvents']
609	<	if ( generator in managedGenerators ) :
610	<	dictOut['params'] = ['FirstRun', 'Generator', 'FirstEvent', 'MaxEvents']
607	>	if (firstLumi):
608	>	dictOut['params'] = ['FirstLumi','MaxEvents']
609	>	if (generator in managedGenerators):
610	>	dictOut['params'] = ['FirstLumi', 'Generator', 'FirstEvent', 'MaxEvents']
611		else:
612		if (generator in managedGenerators) :
613		dictOut['params'] = ['Generator', 'FirstEvent', 'MaxEvents']
#	Line 635 | Line 664 | class JobSplitter:
664		so the job will have AT LEAST as many lumis as requested, perhaps
665		more
666		"""
667	<
667	>	self.useParent = int(self.cfg_params.get('CMSSW.use_parent',0))
668		common.logger.debug('Splitting by Lumi')
669		self.checkLumiSettings()
670
#	Line 643 | Line 672 | class JobSplitter:
672		pubdata = self.args['pubdata']
673
674		lumisPerFile  = pubdata.getLumis()
675	<
675	>	self.parentFiles=pubdata.getParent()
676		# Make the list of WMBS files for job splitter
677		fileList = pubdata.getListFiles()
678	<	thefiles = Fileset(name='FilesToSplit')
678	>	wmFileList = []
679		for jobFile in fileList:
680		block = jobFile['Block']['Name']
681		try:
#	Line 654 | Line 683 | class JobSplitter:
683		except:
684		continue
685		wmbsFile = File(jobFile['LogicalFileName'])
686	+	if not  blockSites[block]:
687	+	wmbsFile['locations'].add('Nowhere')
688		[ wmbsFile['locations'].add(x) for x in blockSites[block] ]
689		wmbsFile['block'] = block
690		for lumi in lumisPerFile[jobFile['LogicalFileName']]:
691		wmbsFile.addRun(Run(lumi[0], lumi[1]))
692	<	thefiles.addFile(wmbsFile)
692	>	wmFileList.append(wmbsFile)
693	>
694	>	fileSet = set(wmFileList)
695	>	thefiles = Fileset(name='FilesToSplit', files = fileSet)
696
697		# Create the factory and workflow
698		work = Workflow()
#	Line 671 | Line 705 | class JobSplitter:
705		jobDestination = []
706		jobCount = 0
707		lumisCreated = 0
708	<
708	>	list_of_blocks = []
709		if not self.limitJobLumis:
710		self.lumisPerJob = pubdata.getMaxLumis() // self.theNumberOfJobs + 1
711		common.logger.info('Each job will process about %s lumis.' %
712		self.lumisPerJob)
713
714	<	for jobGroup in  jobFactory(lumis_per_job = self.lumisPerJob):
714	>	for jobGroup in jobFactory(lumis_per_job = self.lumisPerJob):
715		for job in jobGroup.jobs:
716		if (self.limitNJobs and jobCount >= self.theNumberOfJobs):
717	<	common.logger.info('Limit on number of jobs reached.')
717	>	common.logger.info('Requested number of jobs reached.')
718		break
719		if (self.limitTotalLumis and lumisCreated >= self.totalNLumis):
720	<	common.logger.info('Limit on number of lumis reached.')
720	>	common.logger.info('Requested number of lumis reached.')
721		break
722		lumis = []
723		lfns  = []
724	+	if self.useParent==1:
725	+	parentlfns  = []
726	+	pString =""
727	+
728		locations = []
729	+	blocks = []
730		firstFile = True
731		# Collect information from all the files
732		for jobFile in job.getFiles():
733	+	doFile = False
734		if firstFile:  # Get locations from first file in the job
735		for loc in jobFile['locations']:
736		locations.append(loc)
737	+	blocks.append(jobFile['block'])
738		firstFile = False
739		# Accumulate Lumis from all files
740		for lumiList in jobFile['runs']:
741		theRun = lumiList.run
742		for theLumi in list(lumiList):
743	<	lumis.append( (theRun, theLumi) )
744	<
745	<	lfns.append(jobFile['lfn'])
743	>	if (not self.limitTotalLumis) or \
744	>	(lumisCreated < self.totalNLumis):
745	>	doFile = True
746	>	lumisCreated += 1
747	>	lumis.append( (theRun, theLumi) )
748	>	if doFile:
749	>	lfns.append(jobFile['lfn'])
750	>	if self.useParent==1:
751	>	parent = self.parentFiles[jobFile['lfn']]
752	>	for p in parent :
753	>	pString += p  + ','
754		fileString = ','.join(lfns)
755	<	lumiString = compressLumiString(lumis)
756	<	list_of_lists.append([fileString, str(-1), str(0), lumiString])
757	<
755	>	lumiLister = LumiList(lumis = lumis)
756	>	lumiString = lumiLister.getCMSSWString()
757	>	if self.useParent==1:
758	>	common.logger.debug("Files: "+fileString+" with the following parents: "+pString[:-1])
759	>	pfileString = pString[:-1]
760	>	list_of_lists.append([fileString, pfileString, str(-1), str(0), lumiString])
761	>	else:
762	>	list_of_lists.append([fileString, str(-1), str(0), lumiString])
763	>	list_of_blocks.append(blocks)
764		jobDestination.append(locations)
765		jobCount += 1
711	–	lumisCreated += len(lumis)
766		common.logger.debug('Job %s will run on %s files and %s lumis '
767		% (jobCount, len(lfns), len(lumis) ))
768
#	Line 718 | Line 772 | class JobSplitter:
772		# Prepare dict output matching back to non-WMBS job creation
773		dictOut = {}
774		dictOut['params'] = ['InputFiles', 'MaxEvents', 'SkipEvents', 'Lumis']
775	+	if self.useParent==1:
776	+	dictOut['params']= ['InputFiles','ParentFiles','MaxEvents','SkipEvents','Lumis']
777		dictOut['args'] = list_of_lists
778		dictOut['jobDestination'] = jobDestination
779		dictOut['njobs'] = jobCount
780
781	+	self.cacheBlocks(list_of_blocks,jobDestination)
782	+
783		return dictOut
784
785	+	def cacheBlocks(self, blocks,destinations):
786	+
787	+	saveFblocks=''
788	+	for i in range(len(blocks)):
789	+	sites=self.blackWhiteListParser.checkWhiteList(self.blackWhiteListParser.checkBlackList(destinations[i]))
790	+	if len(sites) != 0:
791	+	for block in blocks[i]:
792	+	saveFblocks += str(block)+'\n'
793	+	writeTXTfile(self, self.fileBlocks_FileName , saveFblocks)
794
795		def Algos(self):
796		"""
#	Line 738 | Line 805 | class JobSplitter:
805		}
806		return SplitAlogs
807
741	–
742	–
743	–	def compressLumiString(lumis):
744	–	"""
745	–	Turn a list of 2-tuples of run/lumi numbers into a list of the format
746	–	R1:L1,R2:L2-R3:L3 which is acceptable to CMSSW LumiBlockRange variable
747	–	"""
748	–
749	–	lumis.sort()
750	–	parts = []
751	–	startRange = None
752	–	endRange = None
753	–
754	–	for lumiBlock in lumis:
755	–	if not startRange: # This is the first one
756	–	startRange = lumiBlock
757	–	endRange = lumiBlock
758	–	elif lumiBlock == endRange: # Same Lumi (different files?)
759	–	pass
760	–	elif lumiBlock[0] == endRange[0] and lumiBlock[1] == endRange[1] + 1: # This is a continuation
761	–	endRange = lumiBlock
762	–	else: # This is the start of a new range
763	–	part = ':'.join(map(str, startRange))
764	–	if startRange != endRange:
765	–	part += '-' + ':'.join(map(str, endRange))
766	–	parts.append(part)
767	–	startRange = lumiBlock
768	–	endRange = lumiBlock
769	–
770	–	# Put out what's left
771	–	if startRange:
772	–	part = ':'.join(map(str, startRange))
773	–	if startRange != endRange:
774	–	part += '-' + ':'.join(map(str, endRange))
775	–	parts.append(part)
776	–
777	–	output = ','.join(parts)
778	–	return output
779	–
780	–

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines