def validateCommon(self): """Common validation stuff""" try: Lexicon.requestName(self.wmspec.name()) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Workflow name validation error: %s" % str(ex)) raise error if self.initialTask.siteWhitelist(): if isinstance(self.initialTask.siteWhitelist(), basestring): error = WorkQueueWMSpecError(self.wmspec, 'Invalid site whitelist: Must be tuple/list but is %s' % type( self.initialTask.siteWhitelist())) raise error try: [Lexicon.cmsname(site) for site in self.initialTask.siteWhitelist()] except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Site whitelist validation error: %s" % str(ex)) raise error else: error = WorkQueueWMSpecError(self.wmspec, "Site whitelist validation error: Empty site whitelist") raise error if self.initialTask.siteBlacklist(): if isinstance(self.initialTask.siteBlacklist(), basestring): error = WorkQueueWMSpecError(self.wmspec, 'Invalid site blacklist: Must be tuple/list but is %s' % type( self.initialTask.siteBlacklist())) raise error try: [Lexicon.cmsname(site) for site in self.initialTask.siteBlacklist()] except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Site blacklist validation error: %s" % str(ex)) raise error # splitter settings if self.args.get('SliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SliceSize parameter') raise error if self.args.get('SubSliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SubSliceSize parameter') raise error # check input dataset is valid try: if self.initialTask.getInputDatasetPath(): Lexicon.dataset(self.initialTask.getInputDatasetPath()) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Dataset validation error: %s" % str(ex)) raise error # if pileup is found, check that they are valid datasets try: pileupDatasets = self.wmspec.listPileupDatasets() for dbsUrl in pileupDatasets: for dataset in pileupDatasets[dbsUrl]: Lexicon.dataset(dataset) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Pileup dataset validation error: %s" % str(ex)) raise error
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() Lexicon.dataset(datasetPath) # check dataset name validBlocks = [] locations = None blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() for blockName in dbs.listFileBlocks(datasetPath): block = dbs.getDBSSummaryInfo(datasetPath, block = blockName) # check block restrictions if blockWhiteList and block['block'] not in blockWhiteList: continue if block['block'] in blockBlackList: continue # check run restrictions if runWhiteList or runBlackList: # listRuns returns a run number per lumi section full_lumi_list = dbs.listRuns(block = block['block']) runs = set(full_lumi_list) # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) accepted_lumis = [x for x in full_lumi_list if x in runs] ratio_accepted = 1. * len(accepted_lumis) / len(full_lumi_list) block[self.lumiType] = len(accepted_lumis) block['NumberOfFiles'] = float(block['NumberOfFiles']) * ratio_accepted block['NumberOfEvents'] = float(block['NumberOfEvents']) * ratio_accepted validBlocks.append(block) if locations is None: locations = set(sitesFromStorageEelements(dbs.listFileBlockLocation(block['block']))) else: locations = locations.intersection(set(sitesFromStorageEelements(dbs.listFileBlockLocation(block['block'])))) # all needed blocks present at these sites if locations: self.data[datasetPath] = list(locations) return validBlocks
def validateCommon(self): """Common validation stuff""" try: Lexicon.requestName(self.wmspec.name()) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Workflow name validation error: %s" % str(ex)) raise error if self.initialTask.siteWhitelist(): if isinstance(self.initialTask.siteWhitelist(), basestring): error = WorkQueueWMSpecError(self.wmspec, 'Invalid site whitelist: Must be tuple/list but is %s' % type( self.initialTask.siteWhitelist())) raise error try: [Lexicon.cmsname(site) for site in self.initialTask.siteWhitelist()] except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Site whitelist validation error: %s" % str(ex)) raise error if self.initialTask.siteBlacklist(): if isinstance(self.initialTask.siteBlacklist(), basestring): error = WorkQueueWMSpecError(self.wmspec, 'Invalid site blacklist: Must be tuple/list but is %s' % type( self.initialTask.siteBlacklist())) raise error try: [Lexicon.cmsname(site) for site in self.initialTask.siteBlacklist()] except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Site blacklist validation error: %s" % str(ex)) raise error # splitter settings if self.args.get('SliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SliceSize parameter') raise error if self.args.get('SubSliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SubSliceSize parameter') raise error # check input dataset is valid try: if self.initialTask.getInputDatasetPath(): Lexicon.dataset(self.initialTask.getInputDatasetPath()) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Dataset validation error: %s" % str(ex)) raise error # if pileup is found, check that they are valid datasets try: pileupDatasets = self.wmspec.listPileupDatasets() for dbsUrl in pileupDatasets: for dataset in pileupDatasets[dbsUrl]: Lexicon.dataset(dataset) except Exception as ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Pileup dataset validation error: %s" % str(ex)) raise error
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() Lexicon.dataset(datasetPath) # check dataset name validBlocks = [] locations = None blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() lumiMask = task.getLumiMask() if lumiMask: maskedBlocks = self.getMaskedBlocks(task, dbs, datasetPath) for blockName in dbs.listFileBlocks(datasetPath): # check block restrictions if blockWhiteList and blockName not in blockWhiteList: continue if blockName in blockBlackList: continue blockSummary = dbs.getDBSSummaryInfo(block=blockName) if int(blockSummary.get('NumberOfFiles', 0)) == 0: logging.warning("Block %s being rejected for lack of valid files to process", blockName) self.badWork.append(blockName) continue if self.args['SliceType'] == 'NumberOfRuns': blockSummary['NumberOfRuns'] = dbs.listRuns(block=blockName) # check lumi restrictions if lumiMask: if blockName not in maskedBlocks: logging.warning("Block %s doesn't pass the lumi mask constraints", blockName) self.rejectedWork.append(blockName) continue acceptedLumiCount = sum([len(maskedBlocks[blockName][lfn].getLumis()) for lfn in maskedBlocks[blockName]]) ratioAccepted = 1. * acceptedLumiCount / float(blockSummary['NumberOfLumis']) maskedRuns = [maskedBlocks[blockName][lfn].getRuns() for lfn in maskedBlocks[blockName]] acceptedRuns = set(lumiMask.getRuns()).intersection(set().union(*maskedRuns)) blockSummary['NumberOfFiles'] = len(maskedBlocks[blockName]) blockSummary['NumberOfEvents'] = float(blockSummary['NumberOfEvents']) * ratioAccepted blockSummary[self.lumiType] = acceptedLumiCount blockSummary['NumberOfRuns'] = acceptedRuns # check run restrictions elif runWhiteList or runBlackList: runs = set(dbs.listRuns(block=blockName)) # multi run blocks need special account, requires more DBS calls recalculateLumiCounts = True if len(runs) > 1 else False # apply blacklist and whitelist runs = runs.difference(runBlackList) if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: logging.warning("Block %s doesn't pass the runs constraints", blockName) self.rejectedWork.append(blockName) continue if recalculateLumiCounts: # Recalculate the number of files, lumis and ~events accepted acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock(fileBlockName=blockName) for fileEntry in fileInfo: acceptedFile = False for lumiInfo in fileEntry['LumiList']: if lumiInfo['RunNumber'] in runs: acceptedFile = True acceptedLumiCount += len(lumiInfo['LumiSectionNumber']) if acceptedFile: acceptedFileCount += 1 acceptedEventCount += fileEntry['NumberOfEvents'] else: acceptedLumiCount = blockSummary["NumberOfLumis"] acceptedFileCount = blockSummary['NumberOfFiles'] acceptedEventCount = blockSummary['NumberOfEvents'] blockSummary[self.lumiType] = acceptedLumiCount blockSummary['NumberOfFiles'] = acceptedFileCount blockSummary['NumberOfEvents'] = acceptedEventCount blockSummary['NumberOfRuns'] = runs validBlocks.append(blockSummary) if locations is None: locations = set(dbs.listFileBlockLocation(blockName)) else: locations = locations.intersection(dbs.listFileBlockLocation(blockName)) # all needed blocks present at these sites if task.getTrustSitelists().get('trustlists'): siteWhitelist = task.siteWhitelist() siteBlacklist = task.siteBlacklist() self.sites = makeLocationsList(siteWhitelist, siteBlacklist) self.data[datasetPath] = self.sites elif locations: self.data[datasetPath] = list(set(self.cric.PNNstoPSNs(locations))) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() validBlocks = [] blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() if task.getLumiMask(): # if we have a lumi mask get only the relevant blocks maskedBlocks = self.getMaskedBlocks(task, dbs, datasetPath) if task.getTrustSitelists().get('trustlists'): siteWhitelist = task.siteWhitelist() siteBlacklist = task.siteBlacklist() self.sites = makeLocationsList(siteWhitelist, siteBlacklist) blocks = [] # Take data inputs or from spec if not self.data: if blockWhiteList: self.data = dict((block, []) for block in blockWhiteList) else: self.data = {datasetPath: []} # same structure as in WorkQueueElement for data in self.data: if data.find('#') > -1: Lexicon.block(data) # check block name datasetPath = str(data.split('#')[0]) blocks.append(str(data)) else: Lexicon.dataset(data) # check dataset name for block in dbs.listFileBlocks(data, onlyClosedBlocks=True): blocks.append(str(block)) for blockName in blocks: # check block restrictions if blockWhiteList and blockName not in blockWhiteList: continue if blockName in blockBlackList: continue if blockName in self.blockBlackListModifier: # Don't duplicate blocks rejected before or blocks that were included and therefore are now in the blacklist continue if task.getLumiMask() and blockName not in maskedBlocks: self.rejectedWork.append(blockName) continue block = dbs.getDBSSummaryInfo(datasetPath, block=blockName) # blocks with 0 valid files should be ignored # - ideally they would be deleted but dbs can't delete blocks if not block['NumberOfFiles'] or block['NumberOfFiles'] == '0': self.rejectedWork.append(blockName) continue # check lumi restrictions if task.getLumiMask(): accepted_lumis = sum([len(maskedBlocks[blockName][lfn].getLumis()) for lfn in maskedBlocks[blockName]]) # use the information given from getMaskedBlocks to compute che size of the block block['NumberOfFiles'] = len(maskedBlocks[blockName]) # ratio = lumis which are ok in the block / total num lumis ratioAccepted = accepted_lumis / block['NumberOfLumis'] block['NumberOfEvents'] = block['NumberOfEvents'] * ratioAccepted block[self.lumiType] = accepted_lumis # check run restrictions elif runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block=block['block']) runs = set(runLumis.keys()) recalculateLumiCounts = False if len(runs) > 1: # If more than one run in the block # Then we must calculate the lumi counts after filtering the run list # This has to be done rarely and requires calling DBS file information recalculateLumiCounts = True # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: self.rejectedWork.append(blockName) continue if len(runs) == len(runLumis): # If there is no change in the runs, then we can skip recalculating lumi counts recalculateLumiCounts = False if recalculateLumiCounts: # Recalculate effective size of block # We pull out file info, since we don't do this often acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock(fileBlockName=block['block']) for fileEntry in fileInfo: acceptedFile = False acceptedFileLumiCount = 0 for lumiInfo in fileEntry['LumiList']: runNumber = lumiInfo['RunNumber'] if runNumber in runs: acceptedFile = True acceptedFileLumiCount += 1 acceptedLumiCount += len(lumiInfo['LumiSectionNumber']) if acceptedFile: acceptedFileCount += 1 if len(fileEntry['LumiList']) != acceptedFileLumiCount: acceptedEventCount += acceptedFileLumiCount * fileEntry['NumberOfEvents'] / len(fileEntry['LumiList']) else: acceptedEventCount += fileEntry['NumberOfEvents'] block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = acceptedFileCount block['NumberOfEvents'] = acceptedEventCount # save locations if task.getTrustSitelists().get('trustlists'): self.data[block['block']] = self.sites else: self.data[block['block']] = self.siteDB.PNNstoPSNs(dbs.listFileBlockLocation(block['block'])) # TODO: need to decide what to do when location is no find. # There could be case for network problem (no connection to dbs, phedex) # or DBS se is not recorded (This will be retried anyway by location mapper) if not self.data[block['block']]: self.data[block['block']] = ["NoInitialSite"] # # No sites for this block, move it to rejected # self.rejectedWork.append(blockName) # continue validBlocks.append(block) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() validBlocks = [] blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() if task.getLumiMask( ): # if we have a lumi mask get only the relevant blocks maskedBlocks = self.getMaskedBlocks(task, dbs, datasetPath) if task.getTrustSitelists().get('trustlists'): siteWhitelist = task.siteWhitelist() siteBlacklist = task.siteBlacklist() self.sites = makeLocationsList(siteWhitelist, siteBlacklist) blocks = [] # Take data inputs or from spec if not self.data: if blockWhiteList: self.data = dict((block, []) for block in blockWhiteList) else: self.data = { datasetPath: [] } # same structure as in WorkQueueElement for data in self.data: if data.find('#') > -1: Lexicon.block(data) # check block name datasetPath = str(data.split('#')[0]) blocks.append(str(data)) else: Lexicon.dataset(data) # check dataset name for block in dbs.listFileBlocks(data, onlyClosedBlocks=True): blocks.append(str(block)) for blockName in blocks: # check block restrictions if blockWhiteList and blockName not in blockWhiteList: continue if blockName in blockBlackList: continue if blockName in self.blockBlackListModifier: # Don't duplicate blocks rejected before or blocks that were included and therefore are now in the blacklist continue if task.getLumiMask() and blockName not in maskedBlocks: self.rejectedWork.append(blockName) continue block = dbs.getDBSSummaryInfo(datasetPath, block=blockName) # blocks with 0 valid files should be ignored # - ideally they would be deleted but dbs can't delete blocks if not block['NumberOfFiles'] or block['NumberOfFiles'] == '0': self.rejectedWork.append(blockName) continue # check lumi restrictions if task.getLumiMask(): accepted_lumis = sum([ len(maskedBlocks[blockName][lfn].getLumis()) for lfn in maskedBlocks[blockName] ]) # use the information given from getMaskedBlocks to compute che size of the block block['NumberOfFiles'] = len(maskedBlocks[blockName]) # ratio = lumis which are ok in the block / total num lumis ratioAccepted = 1. * accepted_lumis / float( block['NumberOfLumis']) block['NumberOfEvents'] = float( block['NumberOfEvents']) * ratioAccepted block[self.lumiType] = accepted_lumis # check run restrictions elif runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block=block['block']) runs = set(runLumis.keys()) recalculateLumiCounts = False if len(runs) > 1: # If more than one run in the block # Then we must calculate the lumi counts after filtering the run list # This has to be done rarely and requires calling DBS file information recalculateLumiCounts = True # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: self.rejectedWork.append(blockName) continue if len(runs) == len(runLumis): # If there is no change in the runs, then we can skip recalculating lumi counts recalculateLumiCounts = False if recalculateLumiCounts: # Recalculate effective size of block # We pull out file info, since we don't do this often acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock( fileBlockName=block['block']) for fileEntry in fileInfo: acceptedFile = False acceptedFileLumiCount = 0 for lumiInfo in fileEntry['LumiList']: runNumber = lumiInfo['RunNumber'] if runNumber in runs: acceptedFile = True acceptedFileLumiCount += 1 acceptedLumiCount += len( lumiInfo['LumiSectionNumber']) if acceptedFile: acceptedFileCount += 1 if len(fileEntry['LumiList'] ) != acceptedFileLumiCount: acceptedEventCount += float(acceptedFileLumiCount) * fileEntry['NumberOfEvents'] \ / len(fileEntry['LumiList']) else: acceptedEventCount += fileEntry[ 'NumberOfEvents'] block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = acceptedFileCount block['NumberOfEvents'] = acceptedEventCount # save locations if task.getTrustSitelists().get('trustlists'): self.data[block['block']] = self.sites else: self.data[block['block']] = self.siteDB.PNNstoPSNs( dbs.listFileBlockLocation(block['block'])) # TODO: need to decide what to do when location is no find. # There could be case for network problem (no connection to dbs, phedex) # or DBS se is not recorded (This will be retried anyway by location mapper) if not self.data[block['block']]: self.data[block['block']] = ["NoInitialSite"] # # No sites for this block, move it to rejected # self.rejectedWork.append(blockName) # continue validBlocks.append(block) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() Lexicon.dataset(datasetPath) # check dataset name validBlocks = [] locations = None blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() if task.getTrustSitelists(): siteWhitelist = task.siteWhitelist() siteBlacklist = task.siteBlacklist() self.sites = makeLocationsList(siteWhitelist, siteBlacklist) for blockName in dbs.listFileBlocks(datasetPath): block = dbs.getDBSSummaryInfo(datasetPath, block=blockName) # check block restrictions if blockWhiteList and block['block'] not in blockWhiteList: continue if block['block'] in blockBlackList: continue # check run restrictions if runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block=block['block']) runs = set(runLumis.keys()) recalculateLumiCounts = False if len(runs) > 1: # If more than one run in the block # Then we must calculate the lumi counts after filtering the run list # This has to be done rarely and requires calling DBS file information recalculateLumiCounts = True # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue if recalculateLumiCounts: # get correct lumi count # Recalculate effective size of block # We pull out file info, since we don't do this often acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock( fileBlockName=block['block']) for fileEntry in fileInfo: acceptedFile = False acceptedFileLumiCount = 0 for lumiInfo in fileEntry['LumiList']: runNumber = lumiInfo['RunNumber'] if runNumber in runs: acceptedFile = True acceptedFileLumiCount += 1 if acceptedFile: acceptedFileCount += 1 acceptedLumiCount += acceptedFileLumiCount if len(fileEntry['LumiList'] ) != acceptedFileLumiCount: acceptedEventCount += float( acceptedFileLumiCount ) * fileEntry['NumberOfEvents'] / len( fileEntry['LumiList']) else: acceptedEventCount += fileEntry[ 'NumberOfEvents'] else: acceptedLumiCount = block["NumberOfLumis"] acceptedFileCount = block['NumberOfFiles'] acceptedEventCount = block['NumberOfEvents'] # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = acceptedFileCount block['NumberOfEvents'] = acceptedEventCount validBlocks.append(block) if locations is None: locations = set(dbs.listFileBlockLocation(block['block'])) else: locations = locations.intersection( dbs.listFileBlockLocation(block['block'])) # all needed blocks present at these sites if self.wmspec.getTrustLocationFlag(): self.data[datasetPath] = self.sites elif locations: self.data[datasetPath] = list( set(self.siteDB.PNNstoPSNs(locations))) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() validBlocks = [] blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() if task.getLumiMask(): #if we have a lumi mask get only the relevant blocks maskedBlocks = self.getMaskedBlocks(task, dbs, datasetPath) if task.inputLocationFlag(): # Then get the locations from the site whitelist/blacklist + SiteDB siteWhitelist = task.siteWhitelist() siteBlacklist = task.siteBlacklist() if siteWhitelist: # Just get the ses matching the whitelists self.sites = siteWhitelist elif siteBlacklist: # Get all CMS sites less the blacklist allSites = cmsSiteNames() self.sites = list(set(allSites) - set (siteBlacklist)) else: # Run at any CMS site self.sites = cmsSiteNames() blocks = [] # Take data inputs or from spec if not self.data: if blockWhiteList: self.data = dict((block, []) for block in blockWhiteList) else: self.data = {datasetPath : []} # same structure as in WorkQueueElement for data in self.data: if data.find('#') > -1: Lexicon.block(data) # check block name datasetPath = str(data.split('#')[0]) blocks.append(str(data)) else: Lexicon.dataset(data) # check dataset name for block in dbs.listFileBlocks(data): blocks.append(str(block)) for blockName in blocks: # check block restrictions if blockWhiteList and blockName not in blockWhiteList: continue if blockName in blockBlackList: continue if task.getLumiMask() and blockName not in maskedBlocks: continue block = dbs.getDBSSummaryInfo(datasetPath, block = blockName) # blocks with 0 valid files should be ignored # - ideally they would be deleted but dbs can't delete blocks if not block['NumberOfFiles'] or block['NumberOfFiles'] == '0': continue #check lumi restrictions if task.getLumiMask(): accepted_lumis = sum( [ len(maskedBlocks[blockName][file]) for file in maskedBlocks[blockName] ] ) #use the information given from getMaskedBlocks to compute che size of the block block['NumberOfFiles'] = len(maskedBlocks[blockName]) #ratio = lumis which are ok in the block / total num lumis ratioAccepted = 1. * accepted_lumis / float(block['NumberOfLumis']) block['NumberOfEvents'] = float(block['NumberOfEvents']) * ratioAccepted block[self.lumiType] = accepted_lumis # check run restrictions elif runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block = block['block']) runs = set(runLumis.keys()) recalculateLumiCounts = False if len(runs) > 1: # If more than one run in the block # Then we must calculate the lumi counts after filtering the run list # This has to be done rarely and requires calling DBS file information recalculateLumiCounts = True # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue if len(runs) == len(runLumis): # If there is no change in the runs, then we can skip recalculating lumi counts recalculateLumiCounts = False if recalculateLumiCounts: # Recalculate effective size of block # We pull out file info, since we don't do this often acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock(fileBlockName = block['block']) for fileEntry in fileInfo: acceptedFile = False acceptedFileLumiCount = 0 for lumiInfo in fileEntry['LumiList']: runNumber = lumiInfo['RunNumber'] if runNumber in runs: acceptedFile = True acceptedFileLumiCount += 1 if acceptedFile: acceptedFileCount += 1 acceptedLumiCount += acceptedFileLumiCount if len(fileEntry['LumiList']) != acceptedFileLumiCount: acceptedEventCount += float(acceptedFileLumiCount) * fileEntry['NumberOfEvents']/len(fileEntry['LumiList']) else: acceptedEventCount += fileEntry['NumberOfEvents'] block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = acceptedFileCount block['NumberOfEvents'] = acceptedEventCount # save locations if task.inputLocationFlag(): self.data[block['block']] = self.sites else: self.data[block['block']] = sitesFromStorageEelements(dbs.listFileBlockLocation(block['block'])) validBlocks.append(block) return validBlocks
except Exception, ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Site blacklist validation error: %s" % str(ex)) raise error # splitter settings if self.args.get('SliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SliceSize parameter') raise error if self.args.get('SubSliceSize', 1) <= 0: error = WorkQueueWMSpecError(self.wmspec, 'Zero or negative SubSliceSize parameter') raise error # check input dataset is valid try: if self.initialTask.getInputDatasetPath(): Lexicon.dataset(self.initialTask.getInputDatasetPath()) except Exception, ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Dataset validation error: %s" % str(ex)) raise error # if pileup is found, check that they are valid datasets try: pileupDatasets = self.wmspec.listPileupDatasets() for dbsUrl in pileupDatasets: for dataset in pileupDatasets[dbsUrl]: Lexicon.dataset(dataset) except Exception, ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError(self.wmspec, "Pileup dataset validation error: %s" % str(ex)) raise error def newQueueElement(self, **args):
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() Lexicon.dataset(datasetPath) # check dataset name validBlocks = [] locations = None blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() for blockName in dbs.listFileBlocks(datasetPath): block = dbs.getDBSSummaryInfo(datasetPath, block=blockName) # check block restrictions if blockWhiteList and block['block'] not in blockWhiteList: continue if block['block'] in blockBlackList: continue # check run restrictions if runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block=block['block']) runs = set(runLumis.keys()) # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) acceptedLumiCount = 0 fullLumiCount = 0 acceptedLumiCount = 0 fullLumiCount = 0 for run in runLumis: if run in runs: acceptedLumiCount += runLumis[run] fullLumiCount += runLumis[run] ratioAccepted = float(acceptedLumiCount) / fullLumiCount block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = int( float(block['NumberOfFiles']) * ratioAccepted) block['NumberOfEvents'] = int( float(block['NumberOfEvents']) * ratioAccepted) validBlocks.append(block) if locations is None: locations = set( sitesFromStorageEelements( dbs.listFileBlockLocation(block['block']))) else: locations = locations.intersection( set( sitesFromStorageEelements( dbs.listFileBlockLocation(block['block'])))) # all needed blocks present at these sites if locations: self.data[datasetPath] = list(locations) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() Lexicon.dataset(datasetPath) # check dataset name validBlocks = [] locations = None blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() siteWhiteList = task.siteWhitelist() for blockName in dbs.listFileBlocks(datasetPath): block = dbs.getDBSSummaryInfo(datasetPath, block = blockName) # check block restrictions if blockWhiteList and block['block'] not in blockWhiteList: continue if block['block'] in blockBlackList: continue # check run restrictions if runWhiteList or runBlackList: # listRunLumis returns a dictionary with the lumi sections per run runLumis = dbs.listRunLumis(block = block['block']) runs = set(runLumis.keys()) recalculateLumiCounts = False if len(runs) > 1: # If more than one run in the block # Then we must calculate the lumi counts after filtering the run list # This has to be done rarely and requires calling DBS file information recalculateLumiCounts = True # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue if recalculateLumiCounts: # get correct lumi count # Recalculate effective size of block # We pull out file info, since we don't do this often acceptedLumiCount = 0 acceptedEventCount = 0 acceptedFileCount = 0 fileInfo = dbs.listFilesInBlock(fileBlockName = block['block']) for fileEntry in fileInfo: acceptedFile = False acceptedFileLumiCount = 0 for lumiInfo in fileEntry['LumiList']: runNumber = lumiInfo['RunNumber'] if runNumber in runs: acceptedFile = True acceptedFileLumiCount += 1 if acceptedFile: acceptedFileCount += 1 acceptedLumiCount += acceptedFileLumiCount if len(fileEntry['LumiList']) != acceptedFileLumiCount: acceptedEventCount += float(acceptedFileLumiCount) * fileEntry['NumberOfEvents']/len(fileEntry['LumiList']) else: acceptedEventCount += fileEntry['NumberOfEvents'] else: acceptedLumiCount = block["NumberOfLumis"] acceptedFileCount = block['NumberOfFiles'] acceptedEventCount = block['NumberOfEvents'] # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) fullLumiCount = block["NumberOfLumis"] block[self.lumiType] = acceptedLumiCount block['NumberOfFiles'] = acceptedFileCount block['NumberOfEvents'] = acceptedEventCount validBlocks.append(block) if locations is None: locations = set(sitesFromStorageEelements(dbs.listFileBlockLocation(block['block']))) else: locations = locations.intersection(set(sitesFromStorageEelements(dbs.listFileBlockLocation(block['block'])))) if self.wmspec.locationDataSourceFlag(): locations = locations.union(siteWhiteList) # all needed blocks present at these sites if locations: self.data[datasetPath] = list(locations) return validBlocks
raise error # splitter settings if self.args.get('SliceSize', 1) <= 0: error = WorkQueueWMSpecError( self.wmspec, 'Zero or negative SliceSize parameter') raise error if self.args.get('SubSliceSize', 1) <= 0: error = WorkQueueWMSpecError( self.wmspec, 'Zero or negative SubSliceSize parameter') raise error # check input dataset is valid try: if self.initialTask.getInputDatasetPath(): Lexicon.dataset(self.initialTask.getInputDatasetPath()) except Exception, ex: # can throw many errors e.g. AttributeError, AssertionError etc. error = WorkQueueWMSpecError( self.wmspec, "Dataset validation error: %s" % str(ex)) raise error def newQueueElement(self, **args): args.setdefault('Status', 'Available') args.setdefault('WMSpec', self.wmspec) args.setdefault('Task', self.initialTask) args.setdefault('RequestName', self.wmspec.name()) args.setdefault('TaskName', self.initialTask.name()) args.setdefault('Dbs', self.initialTask.dbsUrl()) args.setdefault('SiteWhitelist', self.initialTask.siteWhitelist()) args.setdefault('SiteBlacklist', self.initialTask.siteBlacklist()) args.setdefault('EndPolicy', self.wmspec.endPolicyParameters())
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() validBlocks = [] blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() if task.getLumiMask(): #if we have a lumi mask get only the relevant blocks maskedBlocks = self.getMaskedBlocks(task, dbs, datasetPath) blocks = [] # Take data inputs or from spec if not self.data: if blockWhiteList: self.data = dict((block, []) for block in blockWhiteList) else: self.data = {datasetPath : []} # same structure as in WorkQueueElement for data in self.data: if data.find('#') > -1: Lexicon.block(data) # check block name datasetPath = str(data.split('#')[0]) blocks.append(str(data)) else: Lexicon.dataset(data) # check dataset name for block in dbs.listFileBlocks(data): blocks.append(str(block)) for blockName in blocks: # check block restrictions if blockWhiteList and blockName not in blockWhiteList: continue if blockName in blockBlackList: continue if task.getLumiMask() and blockName not in maskedBlocks: continue block = dbs.getDBSSummaryInfo(datasetPath, block = blockName) # blocks with 0 valid files should be ignored # - ideally they would be deleted but dbs can't delete blocks if not block['NumberOfFiles'] or block['NumberOfFiles'] == '0': continue #check lumi restrictions if task.getLumiMask(): accepted_lumis = sum( [ len(maskedBlocks[blockName][file]) for file in maskedBlocks[blockName] ] ) #use the information given from getMaskedBlocks to compute che size of the block block['NumberOfFiles'] = len(maskedBlocks[blockName]) #ratio = lumis which are ok in the block / total num lumis ratio_accepted = 1. * accepted_lumis / float(block['NumberOfLumis']) block['NumberOfEvents'] = float(block['NumberOfEvents']) * ratio_accepted block[self.lumiType] = accepted_lumis # check run restrictions elif runWhiteList or runBlackList: # listRuns returns a run number per lumi section full_lumi_list = dbs.listRuns(block = block['block']) runs = set(full_lumi_list) # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) accepted_lumis = [x for x in full_lumi_list if x in runs] ratio_accepted = 1. * len(accepted_lumis) / len(full_lumi_list) block[self.lumiType] = len(accepted_lumis) block['NumberOfFiles'] = float(block['NumberOfFiles']) * ratio_accepted block['NumberOfEvents'] = float(block['NumberOfEvents']) * ratio_accepted # save locations self.data[block['block']] = sitesFromStorageEelements(dbs.listFileBlockLocation(block['block'])) validBlocks.append(block) return validBlocks
def validBlocks(self, task, dbs): """Return blocks that pass the input data restriction""" datasetPath = task.getInputDatasetPath() validBlocks = [] blockWhiteList = task.inputBlockWhitelist() blockBlackList = task.inputBlockBlacklist() runWhiteList = task.inputRunWhitelist() runBlackList = task.inputRunBlacklist() blocks = [] # Take data inputs or from spec if not self.data: self.data = {datasetPath : []} # same structure as in WorkQueueElement #blocks = dbs.getFileBlocksInfo(datasetPath, locations = False) #else: #dataItems = self.data.keys() for data in self.data: if data.find('#') > -1: Lexicon.block(data) # check dataset name datasetPath = str(data.split('#')[0]) blocks.extend(dbs.getFileBlocksInfo(datasetPath, blockName = str(data), locations = True)) else: Lexicon.dataset(data) # check dataset name blocks.extend(dbs.getFileBlocksInfo(datasetPath, locations = True)) for block in blocks: # blocks with 0 valid files should be ignored # - ideally they would be deleted but dbs can't delete blocks if not block['NumberOfFiles']: continue # check block restrictions if blockWhiteList and block['Name'] not in blockWhiteList: continue if block['Name'] in blockBlackList: continue # check run restrictions if runWhiteList or runBlackList: # listRuns returns a run number per lumi section full_lumi_list = dbs.listRuns(block = block['Name']) runs = set(full_lumi_list) # apply blacklist runs = runs.difference(runBlackList) # if whitelist only accept listed runs if runWhiteList: runs = runs.intersection(runWhiteList) # any runs left are ones we will run on, if none ignore block if not runs: continue # recalculate effective size of block # make a guess for new event/file numbers from ratio # of accepted lumi sections (otherwise have to pull file info) accepted_lumis = [x for x in full_lumi_list if x in runs] ratio_accepted = 1. * len(accepted_lumis) / len(full_lumi_list) block[self.lumiType] = len(accepted_lumis) block['NumberOfFiles'] *= ratio_accepted block['NumberOfEvents'] *= ratio_accepted # get lumi info if needed and not already available if self.args['SliceType'] == self.lumiType and not block.get(self.lumiType): blockSummary = dbs.getDBSSummaryInfo(block = block["Name"]) block[self.lumiType] = blockSummary[self.lumiType] # save locations self.data[block['Name']] = sitesFromStorageEelements([x['Name'] for x in block['StorageElementList']]) validBlocks.append(block) return validBlocks