def hasLHESource(self):
"""
Returns a tuple containing a bool to indicate usage of an
LHESource and an integer for the number of input files.
"""
if bootstrapDone():
self.logger.debug("Getting lhe info from bootstrap cachefile.")
info = self.getCfgInfo()
return info['lheinfo']
isLHE, numFiles = False, 0
if getattr(self.fullConfig.process, 'source'):
source = self.fullConfig.process.source
try:
isLHE = str(source.type_()) == 'LHESource'
except AttributeError as ex:
msg = "\nInvalid CMSSW configuration: Failed to check if 'process.source' is of type 'LHESource': %s" % (
ex)
raise ConfigurationException(msg)
if isLHE:
if hasattr(source, 'fileNames'):
numFiles = len(source.fileNames)
else:
msg = "\nInvalid CMSSW configuration: Object 'process.source', of type 'LHESource', is missing attribute 'fileNames'."
raise ConfigurationException(msg)
return isLHE, numFiles
def hasLHESource(self):
"""
Returns a tuple containing a bool to indicate usage of an
LHESource and an integer for the number of input files.
"""
if bootstrapDone():
self.logger.debug("Getting lhe info from bootstrap cachefile.")
info = self.getCfgInfo()
return info['lheinfo']
isLHE, numFiles = False, 0
if getattr(self.fullConfig.process, 'source'):
source = self.fullConfig.process.source
try:
isLHE = str(source.type_()) == 'LHESource'
except AttributeError as ex:
msg = "Invalid CMSSW configuration: Failed to check if 'process.source' is of type 'LHESource': %s" % (ex)
raise ConfigurationException(msg)
if isLHE:
if hasattr(source, 'fileNames'):
numFiles = len(source.fileNames)
else:
msg = "Invalid CMSSW configuration: Object 'process.source', of type 'LHESource', is missing attribute 'fileNames'."
raise ConfigurationException(msg)
return isLHE, numFiles
def outputFiles(self):
"""
Returns a tuple of lists of output files. First element is PoolOutput files,
second is TFileService files.
"""
if bootstrapDone():
self.logger.debug("Getting output files from bootstrap cachefile.")
info = self.getCfgInfo()
return info['outfiles']
edmfiles = []
process = self.fullConfig.process
#determine all paths/endpaths which will run
pathsToRun = set()
if process.schedule is not None:
for p in process.schedule:
pathsToRun.add(p.label())
#determine all modules on EndPaths
modulesOnEndPaths = set()
for m in process.endpaths_().itervalues():
if len(pathsToRun) == 0 or m.label() in pathsToRun:
for n in m.moduleNames():
modulesOnEndPaths.add(n)
outputModules = set()
for n, o in process.outputModules_().iteritems():
if n in modulesOnEndPaths and hasattr(o, 'fileName'):
edmfile = re.sub(r'^file:', '', o.fileName.value())
edmfile = os.path.basename(edmfile)
edmfiles.append(edmfile)
outputModules.add(o)
## If there are multiple output modules, make sure they have dataset.filterName set.
if len(outputModules) > 1:
for outputModule in outputModules:
try:
dataset = getattr(outputModule, 'dataset')
filterName = getattr(dataset, 'filterName')
except AttributeError:
raise RuntimeError(
'\nYour output module %s does not have a "dataset" PSet '
% outputModule.label() +
'or the PSet does not have a "filterName" member.')
## Find files written by TFileService.
tfiles = []
if 'TFileService' in process.services:
tFileService = process.services['TFileService']
if "fileName" in tFileService.parameterNames_():
tfile = re.sub(r'^file:', '',
getattr(tFileService, 'fileName').value())
tfile = os.path.basename(tfile)
tfiles.append(tfile)
return edmfiles, tfiles
def outputFiles(self):
"""
Returns a tuple of lists of output files. First element is PoolOutput files,
second is TFileService files.
"""
if bootstrapDone():
self.logger.debug("Getting output files from bootstrap cachefile.")
info = self.getCfgInfo()
return info["outfiles"]
edmfiles = []
process = self.fullConfig.process
# determine all paths/endpaths which will run
pathsToRun = set()
if process.schedule is not None:
for p in process.schedule:
pathsToRun.add(p.label())
# determine all modules on EndPaths
modulesOnEndPaths = set()
for m in process.endpaths_().itervalues():
if len(pathsToRun) == 0 or m.label() in pathsToRun:
for n in m.moduleNames():
modulesOnEndPaths.add(n)
outputModules = set()
for n, o in process.outputModules_().iteritems():
if n in modulesOnEndPaths and hasattr(o, "fileName"):
edmfile = re.sub(r"^file:", "", o.fileName.value())
edmfile = os.path.basename(edmfile)
edmfiles.append(edmfile)
outputModules.add(o)
## If there are multiple output modules, make sure they have dataset.filterName set.
if len(outputModules) > 1:
for outputModule in outputModules:
try:
dataset = getattr(outputModule, "dataset")
filterName = getattr(dataset, "filterName")
except AttributeError:
raise RuntimeError(
'Your output module %s does not have a "dataset" PSet ' % outputModule.label()
+ 'or the PSet does not have a "filterName" member.'
)
## Find files written by TFileService.
tfiles = []
if "TFileService" in process.services:
tFileService = process.services["TFileService"]
if "fileName" in tFileService.parameterNames_():
tfile = re.sub(r"^file:", "", getattr(tFileService, "fileName").value())
tfile = os.path.basename(tfile)
tfiles.append(tfile)
return edmfiles, tfiles
def __init__(self, logger=None):
self.logger = logger if logger else logging
if bootstrapDone():
self.logger.debug("Loading required information from the bootstrap environment file")
try:
self.initFromFile()
except EnvironmentException, ee:
self.logger.info(str(ee))
self.logger.info("Will try to find the necessary information from the environment")
self.initFromEnv()
def __init__(self, logger=None):
self.logger = logger if logger else logging
if bootstrapDone():
self.logger.debug("Loading required information from the bootstrap environment file")
try:
self.initFromFile()
except EnvironmentException as ee:
self.logger.info(str(ee))
self.logger.info("Will try to find the necessary information from the environment")
self.initFromEnv()
else:
self.logger.debug("Loading required information from the environment")
self.initFromEnv()
self.logger.debug("Found %s for %s with base %s" % (self.getCmsswVersion(), self.getScramArch(), self.getCmsswBase()))
def __init__(self, config, userConfig=None, logger=None):
global configurationCache
self.config = config
self.logger = logger if logger else logging
self.fullConfig = None
self.outputFile = None
if userConfig:
cfgBaseName = os.path.basename(userConfig).replace(".py", "")
cfgDirName = os.path.dirname(os.path.abspath(userConfig))
if not os.path.isfile(userConfig):
msg = "Cannot find CMSSW configuration file %s in %s" % (
userConfig, os.getcwd())
raise ConfigurationException(msg)
self.logger.info("Importing CMSSW configuration %s" % (userConfig))
pyCfgParams = getattr(self.config.JobType, 'pyCfgParams', [])
originalArgv = sys.argv
sys.argv = [userConfig]
if pyCfgParams:
sys.argv.extend(pyCfgParams)
msg = "Additional parameters for the CMSSW configuration are: %s" % (
pyCfgParams)
self.logger.debug(msg)
configFile, pathname, description = imp.find_module(
cfgBaseName, [cfgDirName])
cacheLine = (tuple(sys.path), tuple(pathname), tuple(sys.argv))
if cacheLine in configurationCache:
self.fullConfig = configurationCache[cacheLine]
configFile.close()
elif not bootstrapDone():
sys.path.append(os.getcwd())
try:
oldstdout = sys.stdout
sys.stdout = open(logger.logfile, 'a')
self.fullConfig = imp.load_module(cfgBaseName, configFile,
pathname, description)
finally:
sys.stdout.close()
sys.stdout = oldstdout
configFile.close()
configurationCache[cacheLine] = self.fullConfig
self.logger.info("Finished importing CMSSW configuration %s" %
(userConfig))
sys.argv = originalArgv
def __init__(self, config, userConfig=None, logger=None):
global configurationCache
self.config = config
self.logger = logger if logger else logging
self.fullConfig = None
self.outputFile = None
if userConfig:
cfgBaseName = os.path.basename(userConfig).replace(".py", "")
cfgDirName = os.path.dirname(os.path.abspath(userConfig))
if not os.path.isfile(userConfig):
msg = "Cannot find CMSSW configuration file %s in %s" % (userConfig, os.getcwd())
raise ConfigurationException(msg)
self.logger.info("Importing CMSSW configuration %s" % (userConfig))
pyCfgParams = getattr(self.config.JobType, 'pyCfgParams', [])
originalArgv = sys.argv
sys.argv = [userConfig]
if pyCfgParams:
sys.argv.extend(pyCfgParams)
msg = "Additional parameters for the CMSSW configuration are: %s" % (pyCfgParams)
self.logger.debug(msg)
configFile, pathname, description = imp.find_module(cfgBaseName, [cfgDirName])
cacheLine = (tuple(sys.path), tuple(pathname), tuple(sys.argv))
if cacheLine in configurationCache:
self.fullConfig = configurationCache[cacheLine]
configFile.close()
elif not bootstrapDone():
sys.path.append(os.getcwd())
try:
oldstdout = sys.stdout
sys.stdout = open(logger.logfile, 'a')
self.fullConfig = imp.load_module(cfgBaseName, configFile, pathname, description)
finally:
sys.stdout.close()
sys.stdout = oldstdout
configFile.close()
configurationCache[cacheLine] = self.fullConfig
self.logger.info("Finished importing CMSSW configuration %s" % (userConfig))
sys.argv = originalArgv
def hasLHESource(self):
"""
Returns a tuple containing a bool to indicate usage of an
LHESource and an integer for the number of input files.
"""
if bootstrapDone():
self.logger.debug("Getting lhe info from bootstrap cachefile.")
info = self.getCfgInfo()
return info['lheinfo']
lhe = False
files = 0
if hasattr(self.fullConfig.process, 'source'):
source = self.fullConfig.process.source
lhe = str(source.type_()) == 'LHESource'
files = len(source.fileNames) if lhe else 0
return lhe, files
def hasPoolSource(self):
"""
Returns if an PoolSource is present in the parameter set.
"""
if bootstrapDone():
self.logger.debug("Getting source info from bootstrap cachefile.")
info = self.getCfgInfo()
return info['poolinfo']
isPool = False
if getattr(self.fullConfig.process, 'source'):
source = self.fullConfig.process.source
try:
isPool = str(source.type_()) == 'PoolSource'
except AttributeError as ex:
msg = "Invalid CMSSW configuration: Failed to check if 'process.source' is of type 'PoolSource': %s" % (ex)
raise ConfigurationException(msg)
return isPool
def hasPoolSource(self):
"""
Returns if an PoolSource is present in the parameter set.
"""
if bootstrapDone():
self.logger.debug("Getting source info from bootstrap cachefile.")
info = self.getCfgInfo()
return info['poolinfo']
isPool = False
if getattr(self.fullConfig.process, 'source'):
source = self.fullConfig.process.source
try:
isPool = str(source.type_()) == 'PoolSource'
except AttributeError as ex:
msg = "Invalid CMSSW configuration: Failed to check if 'process.source' is of type 'PoolSource': %s" % (ex)
raise ConfigurationException(msg)
return isPool
def __init__(self, logger=None):
self.logger = logger if logger else logging
if bootstrapDone():
self.logger.debug(
"Loading required information from the bootstrap environment file"
)
try:
self.initFromFile()
except EnvironmentException as ee:
self.logger.info(str(ee))
self.logger.info(
"Will try to find the necessary information from the environment"
)
self.initFromEnv()
else:
self.logger.debug(
"Loading required information from the environment")
self.initFromEnv()
self.logger.debug(
"Found %s for %s with base %s" %
(self.getCmsswVersion(), self.getScramArch(), self.getCmsswBase()))
def run(self, filecacheurl = None):
"""
Override run() for JobType
"""
configArguments = {'addoutputfiles' : [],
'adduserfiles' : [],
'tfileoutfiles' : [],
'edmoutfiles' : [],
}
if getattr(self.config.Data, 'useParent', False) and getattr(self.config.Data, 'secondaryInputDataset', None):
msg = "Invalid CRAB configuration: Parameters Data.useParent and Data.secondaryInputDataset cannot be used together."
raise ConfigurationException(msg)
# Get SCRAM environment
scram = ScramEnvironment(logger=self.logger)
configArguments.update({'jobarch': scram.getScramArch(),
'jobsw': scram.getCmsswVersion()})
# Build tarball
if self.workdir:
tarUUID = PandaInterface.wrappedUuidGen()
self.logger.debug('UNIQUE NAME: tarUUID %s ' % tarUUID)
if len(tarUUID):
tarFilename = os.path.join(self.workdir, tarUUID + 'default.tgz')
debugTarFilename = os.path.join(self.workdir, 'debugFiles.tgz')
cfgOutputName = os.path.join(self.workdir, BOOTSTRAP_CFGFILE)
else:
raise EnvironmentException('Problem with uuidgen while preparing for Sandbox upload.')
else:
_, tarFilename = tempfile.mkstemp(suffix='.tgz')
_, cfgOutputName = tempfile.mkstemp(suffix='_cfg.py')
if getattr(self.config.Data, 'inputDataset', None):
configArguments['inputdata'] = self.config.Data.inputDataset
## Create CMSSW config.
self.logger.debug("self.config: %s" % (self.config))
self.logger.debug("self.config.JobType.psetName: %s" % (self.config.JobType.psetName))
## The loading of a CMSSW pset in the CMSSWConfig constructor is not idempotent
## in the sense that a second loading of the same pset may not produce the same
## result. Therefore there is a cache in CMSSWConfig to avoid loading any CMSSW
## pset twice. However, some "complicated" psets seem to evade the caching.
## Thus, to be safe, keep the CMSSWConfig instance in a class variable, so that
## it can be reused later if wanted (for example, in PrivateMC when checking if
## the pset has an LHE source) instead of having to load the pset again.
## As for what does "complicated" psets mean, Daniel Riley said that there are
## some psets where one module modifies the configuration from another module.
self.cmsswCfg = CMSSWConfig(config=self.config, logger=self.logger,
userConfig=self.config.JobType.psetName)
## If there is a CMSSW pset, do a basic validation of it.
if not bootstrapDone() and self.config.JobType.psetName:
valid, msg = self.cmsswCfg.validateConfig()
if not valid:
raise ConfigurationException(msg)
## We need to put the pickled CMSSW configuration in the right place.
## Here, we determine if the bootstrap script already run and prepared everything
## for us. In such case we move the file, otherwise we pickle.dump the pset
if not bootstrapDone():
# Write out CMSSW config
self.cmsswCfg.writeFile(cfgOutputName)
else:
# Move the pickled and the configuration files created by the bootstrap script
self.moveCfgFile(cfgOutputName)
## Interrogate the CMSSW pset for output files (only output files produced by
## PoolOutputModule or TFileService are identified automatically). Do this
## automatic detection even if JobType.disableAutomaticOutputCollection = True,
## so that we can still classify the output files in EDM, TFile and additional
## output files in the Task DB (and the job ad).
## TODO: Do we really need this classification at all? cmscp and PostJob read
## the FJR to know if an output file is EDM, TFile or other.
edmfiles, tfiles = self.cmsswCfg.outputFiles()
## If JobType.disableAutomaticOutputCollection = True, ignore the EDM and TFile
## output files that are not listed in JobType.outputFiles.
if getattr(self.config.JobType, 'disableAutomaticOutputCollection', getParamDefaultValue('JobType.disableAutomaticOutputCollection')):
outputFiles = [re.sub(r'^file:', '', file) for file in getattr(self.config.JobType, 'outputFiles', [])]
edmfiles = [file for file in edmfiles if file in outputFiles]
tfiles = [file for file in tfiles if file in outputFiles]
## Get the list of additional output files that have to be collected as given
## in JobType.outputFiles, but remove duplicates listed already as EDM files or
## TFiles.
addoutputFiles = [re.sub(r'^file:', '', file) for file in getattr(self.config.JobType, 'outputFiles', []) if re.sub(r'^file:', '', file) not in edmfiles+tfiles]
self.logger.debug("The following EDM output files will be collected: %s" % edmfiles)
self.logger.debug("The following TFile output files will be collected: %s" % tfiles)
self.logger.debug("The following user output files will be collected: %s" % addoutputFiles)
configArguments['edmoutfiles'] = edmfiles
configArguments['tfileoutfiles'] = tfiles
configArguments['addoutputfiles'].extend(addoutputFiles)
## Give warning message in case no output file was detected in the CMSSW pset
## nor was any specified in the CRAB configuration.
if not configArguments['edmoutfiles'] and not configArguments['tfileoutfiles'] and not configArguments['addoutputfiles']:
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
if getattr(self.config.JobType, 'disableAutomaticOutputCollection', getParamDefaultValue('JobType.disableAutomaticOutputCollection')):
msg += " Automatic detection of output files in the CMSSW configuration is disabled from the CRAB configuration"
msg += " and no output file was explicitly specified in the CRAB configuration."
else:
msg += " CRAB could not detect any output file in the CMSSW configuration"
msg += " nor was any explicitly specified in the CRAB configuration."
msg += " Hence CRAB will not collect any output file from this task."
self.logger.warning(msg)
## UserTarball calls ScramEnvironment which can raise EnvironmentException.
## Since ScramEnvironment is already called above and the exception is not
## handled, we are sure that if we reached this point it will not raise EnvironmentException.
## But otherwise we should take this into account.
with UserTarball(name=tarFilename, logger=self.logger, config=self.config) as tb:
inputFiles = [re.sub(r'^file:', '', file) for file in getattr(self.config.JobType, 'inputFiles', [])]
tb.addFiles(userFiles=inputFiles, cfgOutputName=cfgOutputName)
configArguments['adduserfiles'] = [os.path.basename(f) for f in inputFiles]
try:
uploadResult = tb.upload(filecacheurl = filecacheurl)
except HTTPException as hte:
if 'X-Error-Info' in hte.headers:
reason = hte.headers['X-Error-Info']
reason_re = re.compile(r'\AFile size is ([0-9]*)B\. This is bigger than the maximum allowed size of ([0-9]*)B\.$')
re_match = reason_re.match(reason)
if re_match:
ISBSize = int(re_match.group(1))
ISBSizeLimit = int(re_match.group(2))
reason = "%sError%s:" % (colors.RED, colors.NORMAL)
reason += " Input sanbox size is ~%sMB. This is bigger than the maximum allowed size of %sMB." % (ISBSize/1024/1024, ISBSizeLimit/1024/1024)
ISBContent = sorted(tb.content, reverse=True)
biggestFileSize = ISBContent[0][0]
ndigits = int(math.ceil(math.log(biggestFileSize+1, 10)))
reason += "\nInput sanbox content sorted by size[Bytes]:"
for (size, name) in ISBContent:
reason += ("\n%" + str(ndigits) + "s\t%s") % (size, name)
raise ClientException(reason)
raise hte
except Exception as e:
msg = ("Impossible to calculate the checksum of the sandbox tarball.\nError message: %s.\n"
"More details can be found in %s" % (e, self.logger.logfile))
LOGGERS['CRAB3'].exception(msg) #the traceback is only printed into the logfile
raise ClientException(msg)
debugFilesUploadResult = None
with UserTarball(name=debugTarFilename, logger=self.logger, config=self.config) as dtb:
dtb.addMonFiles()
try:
debugFilesUploadResult = dtb.upload(filecacheurl = filecacheurl)
except Exception as e:
msg = ("Problem uploading debug_files.tar.gz.\nError message: %s.\n"
"More details can be found in %s" % (e, self.logger.logfile))
LOGGERS['CRAB3'].exception(msg) #the traceback is only printed into the logfile
configArguments['cacheurl'] = filecacheurl
configArguments['cachefilename'] = "%s.tar.gz" % uploadResult
if debugFilesUploadResult is not None:
configArguments['debugfilename'] = "%s.tar.gz" % debugFilesUploadResult
self.logger.debug("Result uploading input files: %(cachefilename)s " % configArguments)
# Upload list of user-defined input files to process as the primary input
userFilesList = getattr(self.config.Data, 'userInputFiles', None)
if userFilesList:
self.logger.debug("Attaching list of user-specified primary input files.")
userFilesList = map(string.strip, userFilesList)
userFilesList = [file for file in userFilesList if file]
if len(userFilesList) != len(set(userFilesList)):
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
msg += " CRAB configuration parameter Data.userInputFiles contains duplicated entries."
msg += " Duplicated entries will be removed."
self.logger.warning(msg)
configArguments['userfiles'] = set(userFilesList)
configArguments['primarydataset'] = getattr(self.config.Data, 'outputPrimaryDataset', 'CRAB_UserFiles')
lumi_mask_name = getattr(self.config.Data, 'lumiMask', None)
lumi_list = None
if lumi_mask_name:
self.logger.debug("Attaching lumi mask %s to the request" % (lumi_mask_name))
try:
lumi_list = getLumiList(lumi_mask_name, logger = self.logger)
except ValueError as ex:
msg = "%sError%s:" % (colors.RED, colors.NORMAL)
msg += " Failed to load lumi mask %s : %s" % (lumi_mask_name, ex)
raise ConfigurationException(msg)
run_ranges = getattr(self.config.Data, 'runRange', None)
if run_ranges:
run_ranges_is_valid = re.match('^\d+((?!(-\d+-))(\,|\-)\d+)*$', run_ranges)
if run_ranges_is_valid:
run_list = getRunList(run_ranges)
if lumi_list:
lumi_list.selectRuns(run_list)
if not lumi_list:
msg = "Invalid CRAB configuration: The intersection between the lumi mask and the run range is null."
raise ConfigurationException(msg)
else:
if len(run_list) > 50000:
msg = "CRAB configuration parameter Data.runRange includes %s runs." % str(len(run_list))
msg += " When Data.lumiMask is not specified, Data.runRange can not include more than 50000 runs."
raise ConfigurationException(msg)
lumi_list = LumiList(runs = run_list)
else:
msg = "Invalid CRAB configuration: Parameter Data.runRange should be a comma separated list of integers or (inclusive) ranges. Example: '12345,99900-99910'"
raise ConfigurationException(msg)
if lumi_list:
configArguments['runs'] = lumi_list.getRuns()
## For each run we encode the lumis as a string representing a list of integers: [[1,2],[5,5]] ==> '1,2,5,5'
lumi_mask = lumi_list.getCompactList()
configArguments['lumis'] = [str(reduce(lambda x,y: x+y, lumi_mask[run]))[1:-1].replace(' ','') for run in configArguments['runs']]
configArguments['jobtype'] = 'Analysis'
return tarFilename, configArguments
def run(self, filecacheurl=None):
"""
Override run() for JobType
"""
configArguments = {
'addoutputfiles': [],
'adduserfiles': [],
'tfileoutfiles': [],
'edmoutfiles': [],
}
if getattr(self.config.Data, 'useParent', False) and getattr(
self.config.Data, 'secondaryInputDataset', None):
msg = "Invalid CRAB configuration: Parameters Data.useParent and Data.secondaryInputDataset cannot be used together."
raise ConfigurationException(msg)
# Get SCRAM environment
scram = ScramEnvironment(logger=self.logger)
configArguments.update({
'jobarch': scram.getScramArch(),
'jobsw': scram.getCmsswVersion()
})
# Build tarball
if self.workdir:
tarUUID = PandaInterface.wrappedUuidGen()
self.logger.debug('UNIQUE NAME: tarUUID %s ' % tarUUID)
if len(tarUUID):
tarFilename = os.path.join(self.workdir,
tarUUID + 'default.tgz')
cfgOutputName = os.path.join(self.workdir, BOOTSTRAP_CFGFILE)
else:
raise EnvironmentException(
'Problem with uuidgen while preparing for Sandbox upload.')
else:
_dummy, tarFilename = tempfile.mkstemp(suffix='.tgz')
_dummy, cfgOutputName = tempfile.mkstemp(suffix='_cfg.py')
if getattr(self.config.Data, 'inputDataset', None):
configArguments['inputdata'] = self.config.Data.inputDataset
## Create CMSSW config.
self.logger.debug("self.config: %s" % (self.config))
self.logger.debug("self.config.JobType.psetName: %s" %
(self.config.JobType.psetName))
## The loading of a CMSSW pset in the CMSSWConfig constructor is not idempotent
## in the sense that a second loading of the same pset may not produce the same
## result. Therefore there is a cache in CMSSWConfig to avoid loading any CMSSW
## pset twice. However, some "complicated" psets seem to evade the caching.
## Thus, to be safe, keep the CMSSWConfig instance in a class variable, so that
## it can be reused later if wanted (for example, in PrivateMC when checking if
## the pset has an LHE source) instead of having to load the pset again.
## As for what does "complicated" psets mean, Daniel Riley said that there are
## some psets where one module modifies the configuration from another module.
self.cmsswCfg = CMSSWConfig(config=self.config,
logger=self.logger,
userConfig=self.config.JobType.psetName)
## If there is a CMSSW pset, do a basic validation of it.
if not bootstrapDone() and self.config.JobType.psetName:
valid, msg = self.cmsswCfg.validateConfig()
if not valid:
raise ConfigurationException(msg)
## We need to put the pickled CMSSW configuration in the right place.
## Here, we determine if the bootstrap script already run and prepared everything
## for us. In such case we move the file, otherwise we pickle.dump the pset
if not bootstrapDone():
# Write out CMSSW config
self.cmsswCfg.writeFile(cfgOutputName)
else:
# Move the pickled and the configuration files created by the bootstrap script
self.moveCfgFile(cfgOutputName)
## Interrogate the CMSSW pset for output files (only output files produced by
## PoolOutputModule or TFileService are identified automatically). Do this
## automatic detection even if JobType.disableAutomaticOutputCollection = True,
## so that we can still classify the output files in EDM, TFile and additional
## output files in the Task DB (and the job ad).
## TODO: Do we really need this classification at all? cmscp and PostJob read
## the FJR to know if an output file is EDM, TFile or other.
edmfiles, tfiles = self.cmsswCfg.outputFiles()
## If JobType.disableAutomaticOutputCollection = True, ignore the EDM and TFile
## output files that are not listed in JobType.outputFiles.
if getattr(
self.config.JobType, 'disableAutomaticOutputCollection',
getParamDefaultValue(
'JobType.disableAutomaticOutputCollection')):
outputFiles = [
re.sub(r'^file:', '', file)
for file in getattr(self.config.JobType, 'outputFiles', [])
]
edmfiles = [file for file in edmfiles if file in outputFiles]
tfiles = [file for file in tfiles if file in outputFiles]
## Get the list of additional output files that have to be collected as given
## in JobType.outputFiles, but remove duplicates listed already as EDM files or
## TFiles.
addoutputFiles = [
re.sub(r'^file:', '', file)
for file in getattr(self.config.JobType, 'outputFiles', [])
if re.sub(r'^file:', '', file) not in edmfiles + tfiles
]
self.logger.debug(
"The following EDM output files will be collected: %s" % edmfiles)
self.logger.debug(
"The following TFile output files will be collected: %s" % tfiles)
self.logger.debug(
"The following user output files will be collected: %s" %
addoutputFiles)
configArguments['edmoutfiles'] = edmfiles
configArguments['tfileoutfiles'] = tfiles
configArguments['addoutputfiles'].extend(addoutputFiles)
## Give warning message in case no output file was detected in the CMSSW pset
## nor was any specified in the CRAB configuration.
if not configArguments['edmoutfiles'] and not configArguments[
'tfileoutfiles'] and not configArguments['addoutputfiles']:
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
if getattr(
self.config.JobType, 'disableAutomaticOutputCollection',
getParamDefaultValue(
'JobType.disableAutomaticOutputCollection')):
msg += " Automatic detection of output files in the CMSSW configuration is disabled from the CRAB configuration"
msg += " and no output file was explicitly specified in the CRAB configuration."
else:
msg += " CRAB could not detect any output file in the CMSSW configuration"
msg += " nor was any explicitly specified in the CRAB configuration."
msg += " Hence CRAB will not collect any output file from this task."
self.logger.warning(msg)
## UserTarball calls ScramEnvironment which can raise EnvironmentException.
## Since ScramEnvironment is already called above and the exception is not
## handled, we are sure that if we reached this point it will not raise EnvironmentException.
## But otherwise we should take this into account.
with UserTarball(name=tarFilename,
logger=self.logger,
config=self.config) as tb:
inputFiles = [
re.sub(r'^file:', '', file)
for file in getattr(self.config.JobType, 'inputFiles', [])
]
tb.addFiles(userFiles=inputFiles, cfgOutputName=cfgOutputName)
configArguments['adduserfiles'] = [
os.path.basename(f) for f in inputFiles
]
try:
uploadResult = tb.upload(filecacheurl=filecacheurl)
except HTTPException as hte:
if 'X-Error-Info' in hte.headers:
reason = hte.headers['X-Error-Info']
reason_re = re.compile(
r'\AFile size is ([0-9]*)B\. This is bigger than the maximum allowed size of ([0-9]*)B\.$'
)
re_match = reason_re.match(reason)
if re_match:
ISBSize = int(re_match.group(1))
ISBSizeLimit = int(re_match.group(2))
reason = "%sError%s:" % (colors.RED, colors.NORMAL)
reason += " Input sanbox size is ~%sMB. This is bigger than the maximum allowed size of %sMB." % (
ISBSize / 1024 / 1024, ISBSizeLimit / 1024 / 1024)
ISBContent = sorted(tb.content, reverse=True)
biggestFileSize = ISBContent[0][0]
ndigits = int(
math.ceil(math.log(biggestFileSize + 1, 10)))
reason += "\nInput sanbox content sorted by size[Bytes]:"
for (size, name) in ISBContent:
reason += ("\n%" + str(ndigits) + "s\t%s") % (size,
name)
raise ClientException(reason)
raise hte
except Exception as e:
msg = (
"Impossible to calculate the checksum of the sandbox tarball.\nError message: %s.\n"
"More details can be found in %s" %
(e, self.logger.logfile))
LOGGERS['CRAB3'].exception(
msg) #the traceback is only printed into the logfile
raise ClientException(msg)
configArguments['cacheurl'] = filecacheurl
configArguments['cachefilename'] = "%s.tar.gz" % uploadResult
self.logger.debug("Result uploading input files: %(cachefilename)s " %
configArguments)
# Upload list of user-defined input files to process as the primary input
userFilesList = getattr(self.config.Data, 'userInputFiles', None)
if userFilesList:
self.logger.debug(
"Attaching list of user-specified primary input files.")
userFilesList = map(string.strip, userFilesList)
userFilesList = [file for file in userFilesList if file]
if len(userFilesList) != len(set(userFilesList)):
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
msg += " CRAB configuration parameter Data.userInputFiles contains duplicated entries."
msg += " Duplicated entries will be removed."
self.logger.warning(msg)
configArguments['userfiles'] = set(userFilesList)
configArguments['primarydataset'] = getattr(
self.config.Data, 'outputPrimaryDataset', 'CRAB_UserFiles')
lumi_mask_name = getattr(self.config.Data, 'lumiMask', None)
lumi_list = None
if lumi_mask_name:
self.logger.debug("Attaching lumi mask %s to the request" %
(lumi_mask_name))
try:
lumi_list = getLumiList(lumi_mask_name, logger=self.logger)
except ValueError as ex:
msg = "%sError%s:" % (colors.RED, colors.NORMAL)
msg += " Failed to load lumi mask %s : %s" % (lumi_mask_name,
ex)
raise ConfigurationException(msg)
run_ranges = getattr(self.config.Data, 'runRange', None)
if run_ranges:
run_ranges_is_valid = re.match('^\d+((?!(-\d+-))(\,|\-)\d+)*$',
run_ranges)
if run_ranges_is_valid:
run_list = getRunList(run_ranges)
if lumi_list:
lumi_list.selectRuns(run_list)
if not lumi_list:
msg = "Invalid CRAB configuration: The intersection between the lumi mask and the run range is null."
raise ConfigurationException(msg)
else:
if len(run_list) > 50000:
msg = "CRAB configuration parameter Data.runRange includes %s runs." % str(
len(run_list))
msg += " When Data.lumiMask is not specified, Data.runRange can not include more than 50000 runs."
raise ConfigurationException(msg)
lumi_list = LumiList(runs=run_list)
else:
msg = "Invalid CRAB configuration: Parameter Data.runRange should be a comma separated list of integers or (inclusive) ranges. Example: '12345,99900-99910'"
raise ConfigurationException(msg)
if lumi_list:
configArguments['runs'] = lumi_list.getRuns()
## For each run we encode the lumis as a string representing a list of integers: [[1,2],[5,5]] ==> '1,2,5,5'
lumi_mask = lumi_list.getCompactList()
configArguments['lumis'] = [
str(reduce(lambda x, y: x + y,
lumi_mask[run]))[1:-1].replace(' ', '')
for run in configArguments['runs']
]
configArguments['jobtype'] = 'Analysis'
return tarFilename, configArguments
def run(self, filecacheurl=None):
"""
Override run() for JobType
"""
configArguments = {"addoutputfiles": [], "adduserfiles": [], "tfileoutfiles": [], "edmoutfiles": []}
# Get SCRAM environment
scram = ScramEnvironment(logger=self.logger)
configArguments.update({"jobarch": scram.getScramArch(), "jobsw": scram.getCmsswVersion()})
# Build tarball
if self.workdir:
tarUUID = PandaInterface.wrappedUuidGen()
self.logger.debug("UNIQUE NAME: tarUUID %s " % tarUUID)
if len(tarUUID):
tarFilename = os.path.join(self.workdir, tarUUID + "default.tgz")
cfgOutputName = os.path.join(self.workdir, BOOTSTRAP_CFGFILE)
else:
raise EnvironmentException("Problem with uuidgen while preparing for Sandbox upload.")
else:
_dummy, tarFilename = tempfile.mkstemp(suffix=".tgz")
_dummy, cfgOutputName = tempfile.mkstemp(suffix="_cfg.py")
if getattr(self.config.Data, "inputDataset", None):
configArguments["inputdata"] = self.config.Data.inputDataset
## Create CMSSW config.
self.logger.debug("self.config: %s" % (self.config))
self.logger.debug("self.config.JobType.psetName: %s" % (self.config.JobType.psetName))
## The loading of a CMSSW pset in the CMSSWConfig constructor is not idempotent
## in the sense that a second loading of the same pset may not produce the same
## result. Therefore there is a cache in CMSSWConfig to avoid loading any CMSSW
## pset twice. However, some "complicated" psets seem to evade the caching.
## Thus, to be safe, keep the CMSSWConfig instance in a class variable, so that
## it can be reused later if wanted (for example, in PrivateMC when checking if
## the pset has an LHE source) instead of having to load the pset again.
## As for what does "complicated" psets mean, Daniel Riley said that there are
## some psets where one module modifies the configuration from another module.
self.cmsswCfg = CMSSWConfig(config=self.config, logger=self.logger, userConfig=self.config.JobType.psetName)
## If there is a CMSSW pset, do a basic validation of it.
if not bootstrapDone() and self.config.JobType.psetName:
valid, msg = self.cmsswCfg.validateConfig()
if not valid:
raise ConfigurationException(msg)
## We need to put the pickled CMSSW configuration in the right place.
## Here, we determine if the bootstrap script already run and prepared everything
## for us. In such case we move the file, otherwise we pickle.dump the pset
if not bootstrapDone():
# Write out CMSSW config
self.cmsswCfg.writeFile(cfgOutputName)
else:
# Move the pickled configuration file created by the bootstrap script
self.moveCfgFile(cfgOutputName)
## Interrogate the CMSSW pset for output files (only output files produced by
## PoolOutputModule or TFileService are identified automatically). Do this
## automatic detection even if JobType.disableAutomaticOutputCollection = True,
## so that we can still classify the output files in EDM, TFile and additional
## output files in the Task DB (and the job ad).
## TODO: Do we really need this classification at all? cmscp and PostJob read
## the FJR to know if an output file is EDM, TFile or other.
edmfiles, tfiles = self.cmsswCfg.outputFiles()
## If JobType.disableAutomaticOutputCollection = True, ignore the EDM and TFile
## output files that are not listed in JobType.outputFiles.
if getattr(
self.config.JobType,
"disableAutomaticOutputCollection",
getParamDefaultValue("JobType.disableAutomaticOutputCollection"),
):
outputFiles = [re.sub(r"^file:", "", file) for file in getattr(self.config.JobType, "outputFiles", [])]
edmfiles = [file for file in edmfiles if file in outputFiles]
tfiles = [file for file in tfiles if file in outputFiles]
## Get the list of additional output files that have to be collected as given
## in JobType.outputFiles, but remove duplicates listed already as EDM files or
## TFiles.
addoutputFiles = [
re.sub(r"^file:", "", file)
for file in getattr(self.config.JobType, "outputFiles", [])
if re.sub(r"^file:", "", file) not in edmfiles + tfiles
]
self.logger.debug("The following EDM output files will be collected: %s" % edmfiles)
self.logger.debug("The following TFile output files will be collected: %s" % tfiles)
self.logger.debug("The following user output files will be collected: %s" % addoutputFiles)
configArguments["edmoutfiles"] = edmfiles
configArguments["tfileoutfiles"] = tfiles
configArguments["addoutputfiles"].extend(addoutputFiles)
## Give warning message in case no output file was detected in the CMSSW pset
## nor was any specified in the CRAB configuration.
if (
not configArguments["edmoutfiles"]
and not configArguments["tfileoutfiles"]
and not configArguments["addoutputfiles"]
):
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
if getattr(
self.config.JobType,
"disableAutomaticOutputCollection",
getParamDefaultValue("JobType.disableAutomaticOutputCollection"),
):
msg += " Automatic detection of output files in the CMSSW configuration is disabled from the CRAB configuration"
msg += " and no output file was explicitly specified in the CRAB configuration."
else:
msg += " CRAB could not detect any output file in the CMSSW configuration"
msg += " nor was any explicitly specified in the CRAB configuration."
msg += " Hence CRAB will not collect any output file from this task."
self.logger.warning(msg)
## UserTarball calls ScramEnvironment which can raise EnvironmentException.
## Since ScramEnvironment is already called above and the exception is not
## handled, we are sure that if we reached this point it will not raise EnvironmentException.
## But otherwise we should take this into account.
with UserTarball(name=tarFilename, logger=self.logger, config=self.config) as tb:
inputFiles = [re.sub(r"^file:", "", file) for file in getattr(self.config.JobType, "inputFiles", [])]
tb.addFiles(userFiles=inputFiles, cfgOutputName=cfgOutputName)
configArguments["adduserfiles"] = [os.path.basename(f) for f in inputFiles]
uploadResults = tb.upload(filecacheurl=filecacheurl)
self.logger.debug("Result uploading input files: %s " % str(uploadResults))
configArguments["cacheurl"] = filecacheurl
configArguments["cachefilename"] = uploadResults[0]
isbchecksum = uploadResults[1]
# Upload list of user-defined input files to process as the primary input
userFilesList = getattr(self.config.Data, "userInputFiles", None)
if userFilesList:
self.logger.debug("Attaching list of user-specified primary input files.")
userFilesList = map(string.strip, userFilesList)
userFilesList = [file for file in userFilesList if file]
if len(userFilesList) != len(set(userFilesList)):
msg = "%sWarning%s:" % (colors.RED, colors.NORMAL)
msg += " CRAB configuration parameter Data.userInputFiles contains duplicated entries."
msg += " Duplicated entries will be removed."
self.logger.warning(msg)
configArguments["userfiles"] = set(userFilesList)
## Get the user-specified primary dataset name.
primaryDataset = getattr(self.config.Data, "primaryDataset", "CRAB_UserFiles")
# Normalizes "foo/bar" and "/foo/bar" to "/foo/bar"
primaryDataset = "/" + os.path.join(*primaryDataset.split("/"))
if not re.match("/%(primDS)s.*" % (lfnParts), primaryDataset):
self.logger.warning("Invalid primary dataset name %s; publication may fail." % (primaryDataset))
configArguments["inputdata"] = primaryDataset
lumi_mask_name = getattr(self.config.Data, "lumiMask", None)
lumi_list = None
if lumi_mask_name:
self.logger.debug("Attaching lumi mask %s to the request" % (lumi_mask_name))
try:
lumi_list = getLumiList(lumi_mask_name, logger=self.logger)
except ValueError as ex:
msg = "%sError%s:" % (colors.RED, colors.NORMAL)
msg += " Failed to load lumi mask %s : %s" % (lumi_mask_name, ex)
raise ConfigurationException(msg)
run_ranges = getattr(self.config.Data, "runRange", None)
if run_ranges:
run_ranges_is_valid = re.match("^\d+((?!(-\d+-))(\,|\-)\d+)*$", run_ranges)
if run_ranges_is_valid:
run_list = getRunList(run_ranges)
if lumi_list:
lumi_list.selectRuns(run_list)
if not lumi_list:
msg = "Invalid CRAB configuration: The intersection between the lumi mask and the run range is null."
raise ConfigurationException(msg)
else:
if len(run_list) > 50000:
msg = "CRAB configuration parameter Data.runRange includes %s runs." % str(len(run_list))
msg += (
" When Data.lumiMask is not specified, Data.runRange can not include more than 50000 runs."
)
raise ConfigurationException(msg)
lumi_list = LumiList(runs=run_list)
else:
msg = "Invalid CRAB configuration: Parameter Data.runRange should be a comma separated list of integers or (inclusive) ranges. Example: '12345,99900-99910'"
raise ConfigurationException(msg)
if lumi_list:
configArguments["runs"] = lumi_list.getRuns()
## For each run we encode the lumis as a string representing a list of integers: [[1,2],[5,5]] ==> '1,2,5,5'
lumi_mask = lumi_list.getCompactList()
configArguments["lumis"] = [
str(reduce(lambda x, y: x + y, lumi_mask[run]))[1:-1].replace(" ", "")
for run in configArguments["runs"]
]
configArguments["jobtype"] = "Analysis"
return tarFilename, configArguments, isbchecksum
