def checkAutomaticAvail(self, allowedSplitAlgos):
scram = ScramEnvironment(logger=self.logger)
major, minor = [
int(v) for v in scram.getCmsswVersion().split('_', 3)[1:-1]
]
if major > 7 or (major == 7 and minor >= 2):
allowedSplitAlgos.append('Automatic')
def testInit(self):
"""
Test constructor
"""
scram = ScramEnvironment(logger=self.testLogger)
scram.getCmsswVersion()
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name, mode=mode, dereference=True)
self.checksum = None
def testAccessors(self):
"""
Test various accessors
"""
scram = ScramEnvironment(logger=self.testLogger)
self.assertEqual(scram.getCmsswVersion(), self.version)
self.assertEqual(scram.getScramArch(), self.arch)
self.assertEqual(scram.getCmsswBase(), self.base)
def testScram(self):
"""
Test Scram environment
"""
msg = "You must set up a CMSSW environment first"
scram = ScramEnvironment(logger=self.logger)
self.assertNotEqual(scram.getCmsswVersion(), None, msg)
self.assertNotEqual(scram.getScramArch(), None, msg)
self.assertNotEqual(scram.getCmsswBase(), None, msg)
def testScram(self):
"""
Test Scram environment
"""
msg = "You must set up a CMSSW environment first"
scram = ScramEnvironment(logger=self.logger)
self.assertNotEqual(scram.getCmsswVersion(), None, msg)
self.assertNotEqual(scram.getScramArch(), None, msg)
self.assertNotEqual(scram.getCmsswBase(), None, msg)
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
PandaInterface.LOGGER = logging.getLogger('CRAB3:traceback')
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ and interface/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name, mode=mode, dereference=True)
self.checksum = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'biglib', 'module']
if getattr(
self.config.JobType, 'sendPythonFolder',
configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
# /data/ subdirs contain data files needed by the code
# /interface/ subdirs contain C++ header files needed e.g. by ROOT6
dataDirs = ['data', 'interface']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug("Checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug("Adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath, 'src')
self.logger.debug("Adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException(
"The input file '%s' taken from parameter config.JobType.inputFiles cannot be found."
% globName)
for filename in fileNames:
self.logger.debug("Adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename,
os.path.basename(filename),
recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset files to the tarfile
if cfgOutputName:
basedir = os.path.dirname(cfgOutputName)
self.tarfile.add(cfgOutputName, arcname=BOOTSTRAP_CFGFILE)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_PKL),
arcname=BOOTSTRAP_CFGFILE_PKL)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_DUMP),
arcname=BOOTSTRAP_CFGFILE_DUMP)
#debug directory
configtmp = tempfile.NamedTemporaryFile(delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if not psetfilename == None:
self.tarfile.add(psetfilename, '/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to tarball')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
configtmp.close()
def writeContent(self):
"""Save the content of the tarball"""
self.content = [(int(x.size), x.name)
for x in self.tarfile.getmembers()]
def close(self):
"""
Calculate the checkum and close
"""
self.writeContent()
return self.tarfile.close()
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
self.logger.debug(
"Uploading archive %s to the CRAB cache. Using URI %s" %
(archiveName, filecacheurl))
ufc = CRABClient.Emulator.getEmulator('ufc')({
'endpoint': filecacheurl,
"pycurl": True
})
result = ufc.upload(archiveName, excludeList=USER_SANDBOX_EXCLUSIONS)
if 'hashkey' not in result:
self.logger.error("Failed to upload source files: %s" %
str(result))
raise CachefileNotFoundException
return str(result['hashkey'])
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root, _, files in os.walk(dir_, followlinks=True):
for file_ in files:
os.stat(os.path.join(root, file_))
except OSError as msg:
err = '%sError%s: Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_ , msg)
raise EnvironmentException(err)
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ and interface/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'biglib', 'module']
if getattr(self.config.JobType, 'sendPythonFolder', configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
# /data/ subdirs contain data files needed by the code
# /interface/ subdirs contain C++ header files needed e.g. by ROOT6
dataDirs = ['data','interface']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug(" checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug(" adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath,'src')
self.logger.debug(" adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException('The input file "%s" taken from parameter config.JobType.inputFiles cannot be found' % globName)
for filename in fileNames:
self.logger.debug(" adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset and crabconfig file to the tarfile
if cfgOutputName:
self.tarfile.add(cfgOutputName, arcname='PSet.py')
self.tarfile.add(os.path.splitext(cfgOutputName)[0]+'.pkl', arcname='PSet.pkl')
configtmp = tempfile.NamedTemporaryFile(delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if not psetfilename == None:
self.tarfile.add(psetfilename,'/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to tarball')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
configtmp.close()
def close(self):
"""
Calculate the checkum and clos
"""
self.calculateChecksum()
return self.tarfile.close()
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
self.logger.debug(" uploading archive to cache %s " % archiveName)
ufc = CRABClient.Emulator.getEmulator('ufc')({'endpoint' : filecacheurl})
result = ufc.upload(archiveName)
if 'hashkey' not in result:
self.logger.error("Failed to upload source files: %s" % str(result))
raise CachefileNotFoundException
return str(result['hashkey']) + '.tar.gz', self.checksum
def calculateChecksum(self):
"""
Calculate a checksum that doesn't depend on the tgz
creation data
"""
lsl = [(x.name, int(x.size), int(x.mtime), x.uname) for x in self.tarfile.getmembers()]
hasher = hashlib.md5(str(lsl))
self.logger.debug('tgz contents: %s' % lsl)
self.checksum = hasher.hexdigest()
self.logger.debug('MD5 checksum: %s' % self.checksum)
#Old way reads in the file again. May use for for non-tar files if needed.
#sha256sum = hashlib.sha256()
#with open(self.tarfile.name, 'rb') as f:
#while True:
#chunkdata = f.read(8192)
#if not chunkdata:
#break
#sha256sum.update(chunkdata)
#sha256sum.hexdigest()
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root , _ , files in os.walk(dir_, followlinks = True):
for file_ in files:
os.stat(os.path.join(root, file_ ))
except OSError as msg:
err = '%sError%s: Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_ , msg)
raise EnvironmentException(err)
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
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, requestConfig):
"""
Override run() for JobType
"""
configArguments = {
'addoutputfiles': [],
'adduserfiles': [],
'tfileoutfiles': [],
'edmoutfiles': [],
}
# Get SCRAM environment
scram = ScramEnvironment(logger=self.logger)
configArguments.update({
'jobarch': scram.scramArch,
'jobsw': scram.cmsswVersion,
})
# 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, 'CMSSW_cfg.py')
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')
#configArguments['userisburl'] = 'https://'+ self.config.General.ufccacheUrl + '/crabcache/file?hashkey=' + uploadResults['hashkey']#XXX hardcoded
#configArguments['userisburl'] = 'INSERTuserisburl'#XXX hardcoded
if getattr(self.config.Data, 'inputDataset', None):
configArguments['inputdata'] = self.config.Data.inputDataset
# configArguments['ProcessingVersion'] = getattr(self.config.Data, 'processingVersion', None)
# Create CMSSW config
self.logger.debug("self.config: %s" % self.config)
self.logger.debug("self.config.JobType.psetName: %s" %
self.config.JobType.psetName)
cmsswCfg = CMSSWConfig(config=self.config,
logger=self.logger,
userConfig=self.config.JobType.psetName)
## Interogate CMSSW config and user config for output file names. For now no use for EDM files or TFiles here.
edmfiles, tfiles = cmsswCfg.outputFiles()
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)
# Write out CMSSW config
cmsswCfg.writeFile(cfgOutputName)
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()
self.logger.debug("Result uploading input files: %s " %
str(uploadResults))
configArguments['cachefilename'] = uploadResults[1]
configArguments['cacheurl'] = uploadResults[0]
isbchecksum = uploadResults[2]
# Upload list of user-defined input files to process as the primary input
userFileName = getattr(self.config.Data, 'userInputFile', None)
if userFileName:
self.logger.debug(
"Attaching a list of user-specified primary input files from %s."
% userFileName)
fnames = []
for fname in open(userFileName).readlines():
fnames.append(fname.strip())
configArguments['userfiles'] = filter(
lambda x: x, fnames) #removing whitelines and empty objects
primDS = getattr(self.config.Data, 'primaryDataset', None)
if primDS:
# Normalizes "foo/bar" and "/foo/bar" to "/foo/bar"
primDS = "/" + os.path.join(*primDS.split("/"))
if not re.match("/%(primDS)s.*" % lfnParts, primDS):
self.logger.warning(
"Invalid primary dataset name %s for private MC; publishing may fail"
% primDS)
configArguments['inputdata'] = primDS
else:
configArguments['inputdata'] = getattr(self.config.Data,
'inputDataset',
'/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)
lumi_list = getLumiList(lumi_mask_name, logger=self.logger)
run_ranges = getattr(self.config.Data, 'runRange', None)
run_ranges_is_valid = run_ranges is not None and isinstance(
run_ranges, str) and 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)
else:
if len(run_list) > 50000:
msg = "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)
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
def run(self, requestConfig):
"""
Override run() for JobType
"""
configArguments = {
'addoutputfiles': [],
'adduserfiles': [],
'tfileoutfiles': [],
'edmoutfiles': [],
}
# Get SCRAM environment
scram = ScramEnvironment(logger=self.logger)
configArguments.update({
'jobarch': scram.scramArch,
'jobsw': scram.cmsswVersion,
})
# Build tarball
if self.workdir:
tarFilename = os.path.join(
self.workdir,
PandaInterface.wrappedUuidGen() + 'default.tgz')
cfgOutputName = os.path.join(self.workdir, 'CMSSW_cfg.py')
else:
_dummy, tarFilename = tempfile.mkstemp(suffix='.tgz')
_dummy, cfgOutputName = tempfile.mkstemp(suffix='_cfg.py')
#configArguments['userisburl'] = 'https://'+ self.config.General.ufccacheUrl + '/crabcache/file?hashkey=' + uploadResults['hashkey']#XXX hardcoded
#configArguments['userisburl'] = 'INSERTuserisburl'#XXX hardcoded
if getattr(self.config.Data, 'inputDataset', None):
configArguments['inputdata'] = self.config.Data.inputDataset
# configArguments['ProcessingVersion'] = getattr(self.config.Data, 'processingVersion', None)
# Create CMSSW config
self.logger.debug("self.config: %s" % self.config)
self.logger.debug("self.config.JobType.psetName: %s" %
self.config.JobType.psetName)
cmsswCfg = CMSSWConfig(config=self.config,
logger=self.logger,
userConfig=self.config.JobType.psetName)
# Interogate CMSSW config and user config for output file names, for now no use for edmFiles or TFiles here.
analysisFiles, edmFiles = cmsswCfg.outputFiles()
self.logger.debug("TFiles %s and EDM Files %s will be collected" %
(analysisFiles, edmFiles))
configArguments['tfileoutfiles'] = analysisFiles
configArguments['edmoutfiles'] = edmFiles
outputFiles = getattr(self.config.JobType, 'outputFiles', [])
self.logger.debug("User files %s will be collected" % outputFiles)
configArguments['addoutputfiles'].extend(outputFiles)
# Write out CMSSW config
cmsswCfg.writeFile(cfgOutputName)
with UserTarball(name=tarFilename,
logger=self.logger,
config=self.config) as tb:
inputFiles = getattr(self.config.JobType, 'inputFiles', [])
tb.addFiles(userFiles=inputFiles, cfgOutputName=cfgOutputName)
configArguments['adduserfiles'] = [
os.path.basename(f) for f in inputFiles
]
uploadResults = tb.upload()
self.logger.debug("Result uploading input files: %s " %
str(uploadResults))
configArguments['cachefilename'] = uploadResults[1]
configArguments['cacheurl'] = uploadResults[0]
isbchecksum = uploadResults[2]
# Upload lumi mask if it exists
lumiMaskName = getattr(self.config.Data, 'lumiMask', None)
if lumiMaskName:
self.logger.debug("Attaching lumi mask %s to the request" %
lumiMaskName)
lumiDict = getLumiMask(config=self.config, logger=self.logger)
configArguments['runs'] = lumiDict.keys()
#for each run we'll encode the lumis as a string representing a list of integers
#[[1,2],[5,5]] ==> '1,2,5,5'
configArguments['lumis'] = [ str(reduce(lambda x,y: x+y, \
lumiDict[run]))[1:-1].replace(' ','') \
for run in configArguments['runs'] ]
configArguments['jobtype'] = 'Analysis'
return tarFilename, configArguments, isbchecksum
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ and interface/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:bz2', config=None, logger=None, crabserver=None, s3tester=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name, mode=mode, dereference=True)
self.checksum = None
self.content = None
self.crabserver = crabserver
self.s3tester = s3tester
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'biglib', 'module']
if getattr(self.config.JobType, 'sendPythonFolder', configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
directories.append('cfipython')
if getattr(self.config.JobType, 'sendExternalFolder', configParametersInfo['JobType.sendExternalFolder']['default']):
externalDirPath = os.path.join(self.scram.getCmsswBase(), 'external')
if os.path.exists(externalDirPath) and os.listdir(externalDirPath) != []:
directories.append('external')
else:
self.logger.info("The config.JobType.sendExternalFolder parameter is set to True but the external directory "\
"doesn't exist or is empty, not adding to tarball. Path: %s" % externalDirPath)
# Note that dataDirs are only looked-for and added under the src/ folder.
# /data/ subdirs contain data files needed by the code
# /interface/ subdirs contain C++ header files needed e.g. by ROOT6
dataDirs = ['data', 'interface']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug("Checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug("Adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _, _ in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath, 'src')
self.logger.debug("Adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException("The input file '%s' taken from parameter config.JobType.inputFiles cannot be found." % globName)
for filename in fileNames:
self.logger.debug("Adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset files to the tarfile
if cfgOutputName:
basedir = os.path.dirname(cfgOutputName)
self.tarfile.add(cfgOutputName, arcname=BOOTSTRAP_CFGFILE)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_PKL), arcname=BOOTSTRAP_CFGFILE_PKL)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_DUMP), arcname=BOOTSTRAP_CFGFILE_DUMP)
def addMonFiles(self):
"""
Add monitoring files the debug tarball.
"""
configtmp = tempfile.NamedTemporaryFile(mode='w', delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if psetfilename:
self.tarfile.add(psetfilename, '/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to debug_files.tar.gz')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
configtmp.close()
def writeContent(self):
"""Save the content of the tarball"""
self.content = [(int(x.size), x.name) for x in self.tarfile.getmembers()]
def close(self):
"""
Calculate the checkum and close
"""
self.writeContent()
return self.tarfile.close()
def printSortedContent(self):
"""
To be used for diagnostic printouts
returns a string containing tarball content as a list of files sorted by size
already formatted for use in a print statement
"""
sortedContent = sorted(self.content, reverse=True)
biggestFileSize = sortedContent[0][0]
ndigits = int(math.ceil(math.log(biggestFileSize+1, 10)))
contentList = "\nsandbox content sorted by size[Bytes]:"
for (size, name) in sortedContent:
contentList += ("\n%" + str(ndigits) + "s\t%s") % (size, name)
return contentList
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
archiveSizeBytes = os.path.getsize(archiveName)
# in python3 and python2 with __future__ division, double / means integer division
archiveSizeKB = archiveSizeBytes//1024
if archiveSizeKB <= 512:
archiveSize = "%d KB" % archiveSizeKB
elif archiveSizeKB < 1024*10:
# in python3 and python2 with __future__ division, single / means floating point division
archiveSize = "%3f.1 MB" % (archiveSizeKB/1024)
else:
archiveSize = "%d MB" % (archiveSizeKB//1024)
if archiveSizeBytes > FILE_SIZE_LIMIT:
msg = ("%sError%s: input tarball size %s exceeds maximum allowed limit of %d MB" %
(colors.RED, colors.NORMAL, archiveSize, FILE_SIZE_LIMIT//1024//1024))
msg += self.printSortedContent()
raise SandboxTooBigException(msg)
msg = ("Uploading archive %s (%s) to the CRAB cache. Using URI %s" %
(archiveName, archiveSize, filecacheurl))
self.logger.debug(msg)
if 'S3' in filecacheurl.upper():
# use S3
# generate a 32char hash like UserFileCache used to do
hashkey = calculateChecksum(archiveName, exclude=NEW_USER_SANDBOX_EXCLUSIONS)
# the ".tar.gz" suffix here is forced by other places in the client which add it when
# storing tarball name in task table. Not very elegant to need to hardcode in several places.
cachename = "%s.tar.gz" % hashkey
# current code requires a taskname to extract username. Any dummy one will do
# next version of RESTCache will get username from cmsweb FE headers
uploadToS3(crabserver=self.crabserver, objecttype='sandbox', filepath=archiveName,
tarballname=cachename, logger=self.logger)
else:
# old way using UFC
ufc = CRABClient.Emulator.getEmulator('ufc')({'endpoint' : filecacheurl, "pycurl": True})
t1 = time.time()
result = ufc.upload(archiveName, excludeList=NEW_USER_SANDBOX_EXCLUSIONS)
ufcSeconds = int(time.time()-t1)
if 'hashkey' not in result:
self.logger.error("Failed to upload archive: %s" % str(result))
raise CachefileNotFoundException
hashkey = str(result['hashkey'])
# upload a copy to S3 dev as well, just to stress it a bit, this never raises
s3report = testS3upload(self.s3tester, archiveName, hashkey, self.logger)
# report also how long it took uploading to UFC (which surely worked if we are here)
s3report['ufcseconds'] = ufcSeconds
# upload S3 test report to crabcache
reportFile = '/tmp/crabs3report.' + uuid.uuid4().hex
with open(reportFile, 'w') as fp:
json.dump(s3report, fp)
reportName = 'S3-' + s3report['timestamp'] + ':s3report.json'
try:
ufc.uploadLog(reportFile, reportName)
self.logger.debug('Report of S3 upload stored on CrabCache as %s', reportName)
except Exception as e:
self.logger.debug(str(e))
os.remove(reportFile)
return hashkey
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root, _, files in os.walk(dir_, followlinks=True):
for file_ in files:
os.stat(os.path.join(root, file_))
except OSError as msg:
err = '%sError%s: Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_, msg)
raise EnvironmentException(err)
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ and interface/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'biglib', 'module']
if getattr(self.config.JobType, 'sendPythonFolder', configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
# /data/ subdirs contain data files needed by the code
# /interface/ subdirs contain C++ header files needed e.g. by ROOT6
dataDirs = ['data','interface']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug("Checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug("Adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath,'src')
self.logger.debug("Adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException("The input file '%s' taken from parameter config.JobType.inputFiles cannot be found." % globName)
for filename in fileNames:
self.logger.debug("Adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset files to the tarfile
if cfgOutputName:
basedir = os.path.dirname(cfgOutputName)
self.tarfile.add(cfgOutputName, arcname=BOOTSTRAP_CFGFILE)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_PKL), arcname=BOOTSTRAP_CFGFILE_PKL)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_DUMP), arcname=BOOTSTRAP_CFGFILE_DUMP)
#debug directory
configtmp = tempfile.NamedTemporaryFile(delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if not psetfilename == None:
self.tarfile.add(psetfilename,'/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to tarball')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
configtmp.close()
def writeContent(self):
"""Save the content of the tarball"""
self.content = [(int(x.size), x.name) for x in self.tarfile.getmembers()]
def close(self):
"""
Calculate the checkum and close
"""
self.writeContent()
return self.tarfile.close()
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
self.logger.debug("Uploading archive %s to the CRAB cache. Using URI %s" % (archiveName, filecacheurl))
ufc = CRABClient.Emulator.getEmulator('ufc')({'endpoint' : filecacheurl, "pycurl": True})
result = ufc.upload(archiveName, excludeList = USER_SANDBOX_EXCLUSIONS)
if 'hashkey' not in result:
self.logger.error("Failed to upload source files: %s" % str(result))
raise CachefileNotFoundException
return str(result['hashkey'])
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root , _ , files in os.walk(dir_, followlinks = True):
for file_ in files:
os.stat(os.path.join(root, file_ ))
except OSError as msg:
err = '%sError%s: Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_ , msg)
raise EnvironmentException(err)
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ and interface/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:bz2', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
self.content = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'biglib', 'module']
if getattr(self.config.JobType, 'sendPythonFolder', configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
directories.append('cfipython')
if getattr(self.config.JobType, 'sendExternalFolder', configParametersInfo['JobType.sendExternalFolder']['default']):
externalDirPath = os.path.join(self.scram.getCmsswBase(), 'external')
if os.path.exists(externalDirPath) and os.listdir(externalDirPath) != []:
directories.append('external')
else:
self.logger.info("The config.JobType.sendExternalFolder parameter is set to True but the external directory "\
"doesn't exist or is empty, not adding to tarball. Path: %s" % externalDirPath)
# Note that dataDirs are only looked-for and added under the src/ folder.
# /data/ subdirs contain data files needed by the code
# /interface/ subdirs contain C++ header files needed e.g. by ROOT6
dataDirs = ['data','interface']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug("Checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug("Adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _, _ in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath,'src')
self.logger.debug("Adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException("The input file '%s' taken from parameter config.JobType.inputFiles cannot be found." % globName)
for filename in fileNames:
self.logger.debug("Adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset files to the tarfile
if cfgOutputName:
basedir = os.path.dirname(cfgOutputName)
self.tarfile.add(cfgOutputName, arcname=BOOTSTRAP_CFGFILE)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_PKL), arcname=BOOTSTRAP_CFGFILE_PKL)
self.tarfile.add(os.path.join(basedir, BOOTSTRAP_CFGFILE_DUMP), arcname=BOOTSTRAP_CFGFILE_DUMP)
def addMonFiles(self):
"""
Add monitoring files the debug tarball.
"""
configtmp = tempfile.NamedTemporaryFile(delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if not psetfilename == None:
self.tarfile.add(psetfilename,'/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to debug_files.tar.gz')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
configtmp.close()
def writeContent(self):
"""Save the content of the tarball"""
self.content = [(int(x.size), x.name) for x in self.tarfile.getmembers()]
def close(self):
"""
Calculate the checkum and close
"""
self.writeContent()
return self.tarfile.close()
def printSortedContent(self):
"""
To be used for diagnostic printouts
returns a string containing tarball content as a list of files sorted by size
already formatted for use in a print statement
"""
sortedContent = sorted(self.content, reverse=True)
biggestFileSize = sortedContent[0][0]
ndigits = int(math.ceil(math.log(biggestFileSize+1, 10)))
contentList = "\nsandbox content sorted by size[Bytes]:"
for (size, name) in sortedContent:
contentList += ("\n%" + str(ndigits) + "s\t%s") % (size, name)
return contentList
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
archiveSizeBytes = os.path.getsize(archiveName)
# in python3 and python2 with __future__ division, double / means integer division
archiveSizeKB = archiveSizeBytes//1024
if archiveSizeKB <= 512 :
archiveSize = "%d KB" % archiveSizeKB
elif archiveSizeKB < 1024*10 :
# in python3 and python2 with __future__ division, single / means floating point division
archiveSize = "%3f.1 MB" % (archiveSizeKB/1024)
else:
archiveSize = "%d MB" % (archiveSizeKB//1024)
if archiveSizeBytes > FILE_SIZE_LIMIT :
msg=("%sError%s: input tarball size %s exceeds maximum allowed limit of %d MB" % (colors.RED, colors.NORMAL, archiveSize, FILE_SIZE_LIMIT//1024//1024))
msg += self.printSortedContent()
raise SandboxTooBigException(msg)
msg=("Uploading archive %s (%s) to the CRAB cache. Using URI %s" % (archiveName, archiveSize, filecacheurl))
self.logger.debug(msg)
ufc = CRABClient.Emulator.getEmulator('ufc')({'endpoint' : filecacheurl, "pycurl": True})
result = ufc.upload(archiveName, excludeList = NEW_USER_SANDBOX_EXCLUSIONS)
if 'hashkey' not in result:
self.logger.error("Failed to upload archive: %s" % str(result))
raise CachefileNotFoundException
return str(result['hashkey'])
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root , _ , files in os.walk(dir_, followlinks = True):
for file_ in files:
os.stat(os.path.join(root, file_ ))
except OSError as msg:
err = '%sError%s: Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_ , msg)
raise EnvironmentException(err)
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
def checkAutomaticAvail(self, allowedSplitAlgos):
scram = ScramEnvironment(logger=self.logger)
major, minor = [int(v) for v in scram.getCmsswVersion().split('_', 3)[1:-1]]
if major > 7 or (major == 7 and minor >= 2):
allowedSplitAlgos.append('Automatic')
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.scramArch,
'jobsw' : scram.cmsswVersion, })
# 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, 'CMSSW_cfg.py')
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
# configArguments['ProcessingVersion'] = getattr(self.config.Data, 'processingVersion', None)
# Create CMSSW config
self.logger.debug("self.config: %s" % self.config)
self.logger.debug("self.config.JobType.psetName: %s" % self.config.JobType.psetName)
cmsswCfg = CMSSWConfig(config=self.config, logger=self.logger,
userConfig=self.config.JobType.psetName)
## Interogate CMSSW config and user config for output file names. For now no use for EDM files or TFiles here.
edmfiles, tfiles = cmsswCfg.outputFiles()
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)
# Write out CMSSW config
cmsswCfg.writeFile(cfgOutputName)
## 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: CRAB configuration parameter Data.userInputFiles contains duplicated entries." % (colors.RED, colors.NORMAL)
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)
lumi_list = getLumiList(lumi_mask_name, logger = self.logger)
run_ranges = getattr(self.config.Data, 'runRange', None)
run_ranges_is_valid = run_ranges is not None and isinstance(run_ranges, str) and 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)
else:
if len(run_list) > 50000:
msg = "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)
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
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'module']
dataDirs = ['data']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug(" checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug(" adding directory %s to tarball" % fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath,'src')
self.logger.debug(" adding data directory %s to tarball" % root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException('The input file "%s" taken from parameter config.JobType.inputFiles cannot be found' % globName)
for filename in fileNames:
self.logger.debug(" adding file %s to tarball" % filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
# Adding the pset file to the tarfile
if cfgOutputName:
self.tarfile.add(cfgOutputName, arcname='PSet.py')
currentPath = os.getcwd()
# psetfile = getattr(self.config.JobType, 'psetName', None)
# self.tarfile.add(os.path.join(currentPath, psetfile), arcname='PSet.py')
def close(self):
"""
Calculate the checkum and clos
"""
self.calculateChecksum()
return self.tarfile.close()
def upload(self):
"""
Upload the tarball to the Panda Cache
"""
self.close()
archiveName = self.tarfile.name
serverUrl = ""
self.logger.debug(" uploading archive to cache %s " % archiveName)
status,out = PandaInterface.putFile(archiveName, verbose=False, useCacheSrv=True, reuseSandbox=True)
if out.startswith('NewFileName:'):
# found the same input sandbox to reuse
self.logger.debug("out: %s" % out)
self.logger.debug("status: %s" % status)
self.logger.debug("found the same input sandbox to reuse")
archiveName = out.split(':')[-1]
serverUrl = "https://%s:%s" % (out.split(':')[-2], '25443')
self.logger.debug("archiveName: %s" %archiveName)
elif out.startswith('True'):
archiveName = out.split(':')[-1]
serverUrl = "%s:%s:%s" % (out.split(':')[-4], out.split(':')[-3], out.split(':')[-2])
else:
self.logger.error( str(out) )
self.logger.error("failed to upload source files with %s" % status)
raise CachefileNotFoundException
return serverUrl, archiveName, self.checksum
def calculateChecksum(self):
"""
Calculate a checksum that doesn't depend on the tgz
creation data
"""
lsl = [(x.name, int(x.size), int(x.mtime), x.uname) for x in self.tarfile.getmembers()]
hasher = hashlib.md5(str(lsl))
self.logger.debug('tgz contents: %s' % lsl)
self.checksum = hasher.hexdigest()
self.logger.debug('MD5 checksum: %s' % self.checksum)
#Old way reads in the file again. May use for for non-tar files if needed.
#sha256sum = hashlib.sha256()
#with open(self.tarfile.name, 'rb') as f:
#while True:
#chunkdata = f.read(8192)
#if not chunkdata:
#break
#sha256sum.update(chunkdata)
#sha256sum.hexdigest()
def __getattr__(self, *args):
"""
Pass any unknown functions or attribute requests on to the TarFile object
"""
self.logger.debug("Passing getattr %s on to TarFile" % args)
return self.tarfile.__getattribute__(*args)
def __enter__(self):
"""
Allow use as context manager
"""
return self
def __exit__(self, excType, excValue, excTrace):
"""
Allow use as context manager
"""
self.tarfile.close()
if excType:
return False
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
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name, mode=mode, dereference=True)
self.checksum = None
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'module']
if getattr(
self.config.JobType, 'sendPythonFolder',
configParametersInfo['JobType.sendPythonFolder']['default']):
directories.append('python')
dataDirs = ['data']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug(" checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug(" adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath, 'src')
self.logger.debug(" adding data directory %s to tarball" %
root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException(
'The input file "%s" taken from parameter config.JobType.inputFiles cannot be found'
% globName)
for filename in fileNames:
self.logger.debug(" adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename,
os.path.basename(filename),
recursive=True)
scriptExe = getattr(self.config.JobType, 'scriptExe', None)
if scriptExe:
self.tarfile.add(scriptExe, arcname=os.path.basename(scriptExe))
# Adding the pset and crabconfig file to the tarfile
if cfgOutputName:
self.tarfile.add(cfgOutputName, arcname='PSet.py')
self.tarfile.add(os.path.splitext(cfgOutputName)[0] + '.pkl',
arcname='PSet.pkl')
configtmp = tempfile.NamedTemporaryFile(delete=True)
configtmp.write(str(self.config))
configtmp.flush()
psetfilename = getattr(self.config.JobType, 'psetName', None)
if not psetfilename == None:
self.tarfile.add(psetfilename, '/debug/originalPSet.py')
else:
self.logger.debug('Failed to add pset to tarball')
self.tarfile.add(configtmp.name, '/debug/crabConfig.py')
configtmp.close()
def close(self):
"""
Calculate the checkum and clos
"""
self.calculateChecksum()
return self.tarfile.close()
def upload(self, filecacheurl=None):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
self.logger.debug(" uploading archive to cache %s " % archiveName)
ufc = CRABClient.Emulator.getEmulator('ufc')({
'endpoint': filecacheurl
})
result = ufc.upload(archiveName)
if 'hashkey' not in result:
self.logger.error("Failed to upload source files: %s" %
str(result))
raise CachefileNotFoundException
return str(result['hashkey']) + '.tar.gz', self.checksum
def calculateChecksum(self):
"""
Calculate a checksum that doesn't depend on the tgz
creation data
"""
lsl = [(x.name, int(x.size), int(x.mtime), x.uname)
for x in self.tarfile.getmembers()]
hasher = hashlib.md5(str(lsl))
self.logger.debug('tgz contents: %s' % lsl)
self.checksum = hasher.hexdigest()
self.logger.debug('MD5 checksum: %s' % self.checksum)
#Old way reads in the file again. May use for for non-tar files if needed.
#sha256sum = hashlib.sha256()
#with open(self.tarfile.name, 'rb') as f:
#while True:
#chunkdata = f.read(8192)
#if not chunkdata:
#break
#sha256sum.update(chunkdata)
#sha256sum.hexdigest()
def checkdirectory(self, dir_):
#checking for infinite symbolic link loop
try:
for root, _, files in os.walk(dir_, followlinks=True):
for file_ in files:
os.stat(os.path.join(root, file_))
except OSError, msg:
err = '%sError %s:Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir_ , msg)
raise EnvironmentException(err)
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
class UserTarball(object):
"""
_UserTarball_
A subclass of TarFile for the user code tarballs. By default
creates a new tarball with the user libraries from lib, module,
and the data/ sections of the src/ area.
Also adds user specified files in the right place.
"""
def __init__(self, name=None, mode='w:gz', config=None, logger=None):
self.config = config
self.logger = logger
self.scram = ScramEnvironment(logger=self.logger)
self.logger.debug("Making tarball in %s" % name)
self.tarfile = tarfile.open(name=name , mode=mode, dereference=True)
self.checksum = None
PandaInterface.LOGGER = logging.getLogger('CRAB3:traceback')
def addFiles(self, userFiles=None, cfgOutputName=None):
"""
Add the necessary files to the tarball
"""
directories = ['lib', 'module']
dataDirs = ['data']
userFiles = userFiles or []
# Tar up whole directories
for directory in directories:
fullPath = os.path.join(self.scram.getCmsswBase(), directory)
self.logger.debug(" checking directory %s" % fullPath)
if os.path.exists(fullPath):
self.logger.debug(" adding directory %s to tarball" % fullPath)
self.checkdirectory(fullPath)
self.tarfile.add(fullPath, directory, recursive=True)
# Search for and tar up "data" directories in src/
srcPath = os.path.join(self.scram.getCmsswBase(), 'src')
for root, _dummy, _dummy in os.walk(srcPath):
if os.path.basename(root) in dataDirs:
directory = root.replace(srcPath,'src')
self.logger.debug(" adding data directory %s to tarball" % root)
self.checkdirectory(root)
self.tarfile.add(root, directory, recursive=True)
# Tar up extra files the user needs
for globName in userFiles:
fileNames = glob.glob(globName)
if not fileNames:
raise InputFileNotFoundException('The input file "%s" taken from parameter config.JobType.inputFiles cannot be found' % globName)
for filename in fileNames:
self.logger.debug(" adding file %s to tarball" % filename)
self.checkdirectory(filename)
self.tarfile.add(filename, os.path.basename(filename), recursive=True)
# Adding the pset file to the tarfile
if cfgOutputName:
self.tarfile.add(cfgOutputName, arcname='PSet.py')
currentPath = os.getcwd()
# psetfile = getattr(self.config.JobType, 'psetName', None)
# self.tarfile.add(os.path.join(currentPath, psetfile), arcname='PSet.py')
def close(self):
"""
Calculate the checkum and clos
"""
self.calculateChecksum()
return self.tarfile.close()
def upload(self):
"""
Upload the tarball to the File Cache
"""
self.close()
archiveName = self.tarfile.name
serverUrl = ""
self.logger.debug(" uploading archive to cache %s " % archiveName)
ufc = UserFileCache({'endpoint' : self.config.JobType.filecacheurl})
result = ufc.upload(archiveName)
if 'hashkey' not in result:
self.logger.error("Failed to upload source files: %s" % str(result))
raise CachefileNotFoundException
return self.config.JobType.filecacheurl, str(result['hashkey']) + '.tar.gz', self.checksum
def calculateChecksum(self):
"""
Calculate a checksum that doesn't depend on the tgz
creation data
"""
lsl = [(x.name, int(x.size), int(x.mtime), x.uname) for x in self.tarfile.getmembers()]
hasher = hashlib.md5(str(lsl))
self.logger.debug('tgz contents: %s' % lsl)
self.checksum = hasher.hexdigest()
self.logger.debug('MD5 checksum: %s' % self.checksum)
#Old way reads in the file again. May use for for non-tar files if needed.
#sha256sum = hashlib.sha256()
#with open(self.tarfile.name, 'rb') as f:
#while True:
#chunkdata = f.read(8192)
#if not chunkdata:
#break
#sha256sum.update(chunkdata)
#sha256sum.hexdigest()
def checkdirectory(self,dir):
#checking for infinite symbolic link loop
try :
for root , _ , files in os.walk(dir, followlinks = True ):
for file in files:
os.stat(os.path.join(root, file ))
except OSError , msg :
err = '%sError %s:Infinite directory loop found in: %s \nStderr: %s' % \
(colors.RED, colors.NORMAL, dir , msg)
raise EnvironmentException(err)
