def prepare(self,app,appconfig,appmasterconfig,jobmasterconfig):
'''prepare the subjob specific configuration'''
# PandaTools
from pandatools import Client
from pandatools import AthenaUtils
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('AthenaMCPandaRTHandler prepare called for %s', job.getFQID('.'))
try:
assert self.outsite
except:
logger.error("outsite not set. Aborting")
raise Exception()
job.backend.site = self.outsite
job.backend.actualCE = self.outsite
cloud = job._getRoot().backend.requirements.cloud
job.backend.requirements.cloud = cloud
# now just filling the job from AthenaMC data
jspec = JobSpec()
jspec.jobDefinitionID = job._getRoot().id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.AtlasRelease = 'Atlas-%s' % app.atlas_rel
if app.transform_archive:
jspec.homepackage = 'AnalysisTransforms'+app.transform_archive
elif app.prod_release:
jspec.homepackage = 'AnalysisTransforms-AtlasProduction_'+str(app.prod_release)
jspec.transformation = '%s/runAthena-00-00-11' % Client.baseURLSUB
#---->???? prodDBlock and destinationDBlock when facing several input / output datasets?
jspec.prodDBlock = 'NULL'
if job.inputdata and len(app.inputfiles)>0 and app.inputfiles[0] in app.dsetmap:
jspec.prodDBlock = app.dsetmap[app.inputfiles[0]]
# How to specify jspec.destinationDBlock when more than one type of output is available? Panda prod jobs seem to specify only the last output dataset
outdset=""
for type in ["EVNT","RDO","HITS","AOD","ESD","NTUP"]:
if type in app.outputpaths.keys():
outdset=string.replace(app.outputpaths[type],"/",".")
outdset=outdset[1:-1]
break
if not outdset:
try:
assert len(app.outputpaths.keys())>0
except:
logger.error("app.outputpaths is empty: check your output datasets")
raise
type=app.outputpaths.keys()[0]
outdset=string.replace(app.outputpaths[type],"/",".")
outdset=outdset[1:-1]
jspec.destinationDBlock = outdset
jspec.destinationSE = self.outsite
jspec.prodSourceLabel = 'user'
jspec.assignedPriority = 1000
jspec.cloud = cloud
# memory
if job.backend.requirements.memory != -1:
jspec.minRamCount = job.backend.requirements.memory
jspec.computingSite = self.outsite
jspec.cmtConfig = AthenaUtils.getCmtConfig(athenaVer=app.atlas_rel)
# library (source files)
flib = FileSpec()
flib.lfn = self.library
# flib.GUID =
flib.type = 'input'
# flib.status =
flib.dataset = self.libDataset
flib.dispatchDBlock = self.libDataset
jspec.addFile(flib)
# input files FIXME: many more input types
for lfn in app.inputfiles:
useguid=app.turls[lfn].replace("guid:","")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# add dbfiles if any:
for lfn in app.dbfiles:
useguid=app.dbturls[lfn].replace("guid:","")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# then minbias files
for lfn in app.mbfiles:
useguid=app.minbias_turls[lfn].replace("guid:","")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# then cavern files
for lfn in app.cavernfiles:
useguid=app.cavern_turls[lfn].replace("guid:","")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# output files( this includes the logfiles)
# Output files
jidtag=""
job = app._getParent() # Returns job or subjob object
if job._getRoot().subjobs:
jidtag = job._getRoot().id
else:
jidtag = "%d" % job.id
outfiles=app.subjobsOutfiles[job.id]
pandaOutfiles={}
for type in outfiles.keys():
pandaOutfiles[type]=outfiles[type]+"."+str(jidtag)
if type=="LOG":
pandaOutfiles[type]+=".tgz"
#print pandaOutfiles
for outtype in pandaOutfiles.keys():
fout = FileSpec()
dset=string.replace(app.outputpaths[outtype],"/",".")
dset=dset[1:-1]
fout.dataset=dset
fout.lfn=pandaOutfiles[outtype]
fout.type = 'output'
# fout.destinationDBlock = jspec.destinationDBlock
fout.destinationDBlock = fout.dataset
fout.destinationSE = jspec.destinationSE
if outtype=='LOG':
fout.type='log'
fout.destinationDBlock = fout.dataset
fout.destinationSE = job.backend.site
jspec.addFile(fout)
# job parameters
param = '-l %s ' % self.library # user tarball.
# use corruption checker
if job.backend.requirements.corCheck:
param += '--corCheck '
# disable to skip missing files
if job.backend.requirements.notSkipMissing:
param += '--notSkipMissing '
# transform parameters
# need to update arglist with final output file name...
newArgs=[]
if app.mode == "evgen":
app.args[3]=app.args[3]+" -t "
if app.verbosity:
app.args[3]=app.args[3]+" -l %s " % app.verbosity
for arg in app.args[3:]:
for type in outfiles.keys():
if arg.find(outfiles[type])>-1:
arg=arg.replace(outfiles[type],pandaOutfiles[type])
newArgs.append(arg)
arglist=string.join(newArgs," ")
# print "Arglist:",arglist
param += ' -r ./ '
param += ' -j "%s"' % urllib.quote(arglist)
allinfiles=app.inputfiles+app.dbfiles
# Input files.
param += ' -i "%s" ' % allinfiles
if len(app.mbfiles)>0:
param+= ' -m "%s" ' % app.mbfiles
if len(app.cavernfiles)>0:
param+= ' -n "%s" ' % app.cavernfiles
# param += '-m "[]" ' #%minList FIXME
# param += '-n "[]" ' #%cavList FIXME
del pandaOutfiles["LOG"] # logfiles do not appear in IROOT block, and this one is not needed anymore...
param += ' -o "{\'IROOT\':%s }"' % str(pandaOutfiles.items())
# source URL
matchURL = re.search("(http.*://[^/]+)/",Client.baseURLSSL)
if matchURL != None:
param += " --sourceURL %s " % matchURL.group(1)
param += " --trf"
jspec.jobParameters = param
jspec.metadata="--trf \"%s\"" % arglist
#print "SUBJOB DETAILS:",jspec.values()
if app.dryrun:
print "job.application.dryrun activated, printing out job parameters"
print jspec.values()
return
return jspec
def prepare(self,app,appsubconfig,appmasterconfig,jobmasterconfig):
'''prepare the subjob specific configuration'''
from pandatools import Client
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('AthenaPandaRTHandler prepare called for %s', job.getFQID('.'))
# in case of a simple job get the dataset content, otherwise subjobs are filled by the splitter
if job.inputdata and not job._getRoot().subjobs:
if not job.inputdata.names:
contents = job.inputdata.get_contents(overlap=False, size=True)
for ds in contents.keys():
for f in contents[ds]:
job.inputdata.guids.append( f[0] )
job.inputdata.names.append( f[1][0] )
job.inputdata.sizes.append( f[1][1] )
job.inputdata.checksums.append( f[1][2] )
job.inputdata.scopes.append( f[1][3] )
site = job._getRoot().backend.site
job.backend.site = site
job.backend.actualCE = site
cloud = job._getRoot().backend.requirements.cloud
job.backend.requirements.cloud = cloud
# if no outputdata are given
if not job.outputdata:
job.outputdata = DQ2OutputDataset()
job.outputdata.datasetname = job._getRoot().outputdata.datasetname
#if not job.outputdata.datasetname:
else:
job.outputdata.datasetname = job._getRoot().outputdata.datasetname
if not job.outputdata.datasetname:
raise ApplicationConfigurationError('DQ2OutputDataset has no datasetname')
jspec = JobSpec()
jspec.jobDefinitionID = job._getRoot().id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.transformation = '%s/runGen-00-00-02' % Client.baseURLSUB
if job.inputdata:
jspec.prodDBlock = job.inputdata.dataset[0]
else:
jspec.prodDBlock = 'NULL'
jspec.destinationDBlock = job.outputdata.datasetname
if job.outputdata.location:
if not job._getRoot().subjobs or job.id == 0:
logger.warning('You have specified outputdata.location. Note that Panda may not support writing to a user-defined output location.')
jspec.destinationSE = job.outputdata.location
else:
jspec.destinationSE = site
jspec.prodSourceLabel = configPanda['prodSourceLabelRun']
jspec.processingType = configPanda['processingType']
jspec.assignedPriority = configPanda['assignedPriorityRun']
jspec.cloud = cloud
# memory
if job.backend.requirements.memory != -1:
jspec.minRamCount = job.backend.requirements.memory
# cputime
if job.backend.requirements.cputime != -1:
jspec.maxCpuCount = job.backend.requirements.cputime
jspec.computingSite = site
# library (source files)
if job.backend.libds:
flib = FileSpec()
flib.lfn = self.fileBO.lfn
flib.GUID = self.fileBO.GUID
flib.type = 'input'
flib.status = self.fileBO.status
flib.dataset = self.fileBO.destinationDBlock
flib.dispatchDBlock = self.fileBO.destinationDBlock
jspec.addFile(flib)
elif job.backend.bexec:
flib = FileSpec()
flib.lfn = self.library
flib.type = 'input'
flib.dataset = self.libDataset
flib.dispatchDBlock = self.libDataset
jspec.addFile(flib)
# input files FIXME: many more input types
if job.inputdata:
for guid, lfn, size, checksum, scope in zip(job.inputdata.guids,job.inputdata.names,job.inputdata.sizes, job.inputdata.checksums, job.inputdata.scopes):
finp = FileSpec()
finp.lfn = lfn
finp.GUID = guid
finp.scope = scope
# finp.fsize =
# finp.md5sum =
finp.dataset = job.inputdata.dataset[0]
finp.prodDBlock = job.inputdata.dataset[0]
finp.dispatchDBlock = job.inputdata.dataset[0]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# output files
# outMap = {}
#FIXME: if options.outMeta != []:
self.rundirectory = "."
# log files
flog = FileSpec()
flog.lfn = '%s._$PANDAID.log.tgz' % job.outputdata.datasetname
flog.type = 'log'
flog.dataset = job.outputdata.datasetname
flog.destinationDBlock = job.outputdata.datasetname
flog.destinationSE = job.backend.site
jspec.addFile(flog)
# job parameters
param = ''
# source URL
matchURL = re.search("(http.*://[^/]+)/",Client.baseURLCSRVSSL)
srcURL = ""
if matchURL != None:
srcURL = matchURL.group(1)
param += " --sourceURL %s " % srcURL
param += '-r "%s" ' % self.rundirectory
exe_name = job.application.exe
if job.backend.bexec == '':
if hasattr(job.application.exe, "name"):
exe_name = os.path.basename(job.application.exe.name)
# set jobO parameter
if job.application.args:
param += ' -j "" -p "%s %s" '%(exe_name,urllib.quote(" ".join(job.application.args)))
else:
param += ' -j "" -p "%s" '%exe_name
if self.inputsandbox:
param += ' -a %s '%self.inputsandbox
else:
param += '-l %s ' % self.library
param += '-j "" -p "%s %s" ' % ( exe_name,urllib.quote(" ".join(job.application.args)))
if job.inputdata:
param += '-i "%s" ' % job.inputdata.names
# fill outfiles
outfiles = {}
for f in self.extOutFile:
tarnum = 1
if f.find('*') != -1:
# archive *
outfiles[f] = "outputbox%i.%s.%s.tar.gz" % (tarnum, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S") )
tarnum += 1
else:
outfiles[f] = "%s.%s.%s" %(f, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S"))
fout = FileSpec()
fout.lfn = outfiles[f]
fout.type = 'output'
fout.dataset = job.outputdata.datasetname
fout.destinationDBlock = job.outputdata.datasetname
fout.destinationSE = job.backend.site
jspec.addFile(fout)
param += '-o "%s" ' % (outfiles) # must be double quotes, because python prints strings in 'single quotes'
for file in jspec.Files:
if file.type in [ 'output', 'log'] and configPanda['chirpconfig']:
file.dispatchDBlockToken = configPanda['chirpconfig']
logger.debug('chirp file %s',file)
jspec.jobParameters = param
return jspec
def run(self, data):
datasetName = 'panda:panda.destDB.%s' % commands.getoutput('uuidgen')
destName = 'ANALY_RRC-KI-HPC'
site = 'ANALY_RRC-KI-HPC'
scope = config['DEFAULT_SCOPE']
distributive = data['distributive']
release = data['release']
parameters = data['parameters']
input_type = data['input_type']
input_params = data['input_params']
input_files = data['input_files']
output_type = data['output_type']
output_params = data['output_params']
output_files = data['output_files']
jobid = data['jobid']
_logger.debug('Jobid: ' + str(jobid))
job = JobSpec()
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = commands.getoutput('uuidgen')
job.transformation = config['DEFAULT_TRF']
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 1000
job.prodSourceLabel = 'user'
job.computingSite = site
job.cloud = 'RU'
job.prodDBlock = "%s:%s.%s" % (scope, scope, job.jobName)
job.jobParameters = '%s %s "%s"' % (release, distributive, parameters)
params = {}
_logger.debug('MoveData')
ec = 0
ec, uploaded_input_files = movedata(
params=params,
fileList=input_files,
fromType=input_type,
fromParams=input_params,
toType='hpc',
toParams={'dest': '/' + re.sub(':', '/', job.prodDBlock)})
if ec != 0:
_logger.error('Move data error: ' + ec[1])
return
for file in uploaded_input_files:
fileIT = FileSpec()
fileIT.lfn = file
fileIT.dataset = job.prodDBlock
fileIT.prodDBlock = job.prodDBlock
fileIT.type = 'input'
fileIT.scope = scope
fileIT.status = 'ready'
fileIT.GUID = commands.getoutput('uuidgen')
job.addFile(fileIT)
for file in output_files:
fileOT = FileSpec()
fileOT.lfn = file
fileOT.destinationDBlock = job.prodDBlock
fileOT.destinationSE = job.destinationSE
fileOT.dataset = job.prodDBlock
fileOT.type = 'output'
fileOT.scope = scope
fileOT.GUID = commands.getoutput('uuidgen')
job.addFile(fileOT)
fileOL = FileSpec()
fileOL.lfn = "%s.log.tgz" % job.jobName
fileOL.destinationDBlock = job.destinationDBlock
fileOL.destinationSE = job.destinationSE
fileOL.dataset = job.destinationDBlock
fileOL.type = 'log'
fileOL.scope = 'panda'
job.addFile(fileOL)
self.jobList.append(job)
#submitJob
o = self.submitJobs(self.jobList)
x = o[0]
#update PandaID
conn = MySQLdb.connect(
host=self.dbhost,
db=self.dbname,
# port=self.dbport, connect_timeout=self.dbtimeout,
user=self.dbuser,
passwd=self.dbpasswd)
cur = conn.cursor()
try:
varDict = {}
PandaID = int(x[0])
varDict['id'] = jobid
varDict['pandaId'] = PandaID
sql = "UPDATE %s SET %s.pandaId=%s WHERE %s.id=%s" % (
self.table_jobs, self.table_jobs, varDict['pandaId'],
self.table_jobs, varDict['id'])
cur.execute(sql, varDict)
except:
_logger.error('SENDJOB: Incorrect server response')
try:
conn.commit()
return True
except:
_logger.error("commit error")
return False
fileOZ.dataset = job.destinationDBlock
fileOZ.type = 'output'
job.addFile(fileOZ)
files = [
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00001.pool.root.1',
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00002.pool.root.1',
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00003.pool.root.1',
]
for lfn in files:
fileI = FileSpec()
fileI.dataset = job.prodDBlock
fileI.prodDBlock = job.prodDBlock
fileI.lfn = lfn
fileI.type = 'input'
fileI.status = 'ready'
job.addFile(fileI)
fileL = FileSpec()
fileL.dataset = 'user.TadashiMaeno.lib._000157'
fileL.prodDBlock = 'user.TadashiMaeno.lib._000157'
fileL.lfn = 'user.TadashiMaeno.lib._000157.lib.tgz'
fileL.type = 'input'
fileL.status = 'ready'
job.addFile(fileL)
job.jobParameters = """ -l user.TadashiMaeno.lib._000157.lib.tgz -r run/ -j " AnalysisSkeleton_jobOptions.py" -i "['testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00001.pool.root.1', 'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00002.pool.root.1', 'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00003.pool.root.1']" -o "{'AANT': [('AANTupleStream', 'AANT', '%s')]}" """ % fileOZ.lfn
jobList.append(job)
s, o = Client.submitJobs(jobList)
def send_job(jobid, siteid):
_logger.debug('Jobid: ' + str(jobid))
site = sites_.get(siteid)
job = jobs_.get(int(jobid))
cont = job.container
files_catalog = cont.files
fscope = getScope(job.owner.username)
datasetName = '{}:{}'.format(fscope, cont.guid)
distributive = job.distr.name
release = job.distr.release
# Prepare runScript
parameters = job.distr.command
parameters = parameters.replace("$COMMAND$", job.params)
parameters = parameters.replace("$USERNAME$", job.owner.username)
parameters = parameters.replace("$WORKINGGROUP$", job.owner.working_group)
# Prepare metadata
metadata = dict(user=job.owner.username)
# Prepare PanDA Object
pandajob = JobSpec()
pandajob.jobDefinitionID = int(time.time()) % 10000
pandajob.jobName = cont.guid
pandajob.transformation = client_config.DEFAULT_TRF
pandajob.destinationDBlock = datasetName
pandajob.destinationSE = site.se
pandajob.currentPriority = 1000
pandajob.prodSourceLabel = 'user'
pandajob.computingSite = site.ce
pandajob.cloud = 'RU'
pandajob.VO = 'atlas'
pandajob.prodDBlock = "%s:%s" % (fscope, pandajob.jobName)
pandajob.coreCount = job.corecount
pandajob.metadata = json.dumps(metadata)
#pandajob.workingGroup = job.owner.working_group
if site.encode_commands:
# It requires script wrapper on cluster side
pandajob.jobParameters = '%s %s %s "%s"' % (cont.guid, release,
distributive, parameters)
else:
pandajob.jobParameters = parameters
has_input = False
for fcc in files_catalog:
if fcc.type == 'input':
f = fcc.file
guid = f.guid
fileIT = FileSpec()
fileIT.lfn = f.lfn
fileIT.dataset = pandajob.prodDBlock
fileIT.prodDBlock = pandajob.prodDBlock
fileIT.type = 'input'
fileIT.scope = fscope
fileIT.status = 'ready'
fileIT.GUID = guid
pandajob.addFile(fileIT)
has_input = True
if fcc.type == 'output':
f = fcc.file
fileOT = FileSpec()
fileOT.lfn = f.lfn
fileOT.destinationDBlock = pandajob.prodDBlock
fileOT.destinationSE = pandajob.destinationSE
fileOT.dataset = pandajob.prodDBlock
fileOT.type = 'output'
fileOT.scope = fscope
fileOT.GUID = f.guid
pandajob.addFile(fileOT)
# Save replica meta
fc.new_replica(f, site)
if not has_input:
# Add fake input
fileIT = FileSpec()
fileIT.lfn = "fake.input"
fileIT.dataset = pandajob.prodDBlock
fileIT.prodDBlock = pandajob.prodDBlock
fileIT.type = 'input'
fileIT.scope = fscope
fileIT.status = 'ready'
fileIT.GUID = "fake.guid"
pandajob.addFile(fileIT)
# Prepare lof file
fileOL = FileSpec()
fileOL.lfn = "%s.log.tgz" % pandajob.jobName
fileOL.destinationDBlock = pandajob.destinationDBlock
fileOL.destinationSE = pandajob.destinationSE
fileOL.dataset = '{}:logs'.format(fscope)
fileOL.type = 'log'
fileOL.scope = 'panda'
pandajob.addFile(fileOL)
# Save log meta
log = File()
log.scope = fscope
log.lfn = fileOL.lfn
log.guid = getGUID(log.scope, log.lfn)
log.type = 'log'
log.status = 'defined'
files_.save(log)
# Save replica meta
fc.new_replica(log, site)
# Register file in container
fc.reg_file_in_cont(log, cont, 'log')
# Submit job
o = submitJobs([pandajob])
x = o[0]
try:
#update PandaID
PandaID = int(x[0])
job.pandaid = PandaID
job.ce = site.ce
except:
job.status = 'submit_error'
jobs_.save(job)
return 0
def prepare(self, app, appsubconfig, appmasterconfig, jobmasterconfig):
'''prepare the subjob specific configuration'''
from pandatools import Client
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('AthenaPandaRTHandler prepare called for %s',
job.getFQID('.'))
# in case of a simple job get the dataset content, otherwise subjobs are filled by the splitter
if job.inputdata and not job._getRoot().subjobs:
if not job.inputdata.names:
contents = job.inputdata.get_contents(overlap=False, size=True)
for ds in contents.keys():
for f in contents[ds]:
job.inputdata.guids.append(f[0])
job.inputdata.names.append(f[1][0])
job.inputdata.sizes.append(f[1][1])
job.inputdata.checksums.append(f[1][2])
job.inputdata.scopes.append(f[1][3])
site = job._getRoot().backend.site
job.backend.site = site
job.backend.actualCE = site
cloud = job._getRoot().backend.requirements.cloud
job.backend.requirements.cloud = cloud
# if no outputdata are given
if not job.outputdata:
job.outputdata = DQ2OutputDataset()
job.outputdata.datasetname = job._getRoot().outputdata.datasetname
#if not job.outputdata.datasetname:
else:
job.outputdata.datasetname = job._getRoot().outputdata.datasetname
if not job.outputdata.datasetname:
raise ApplicationConfigurationError(
None, 'DQ2OutputDataset has no datasetname')
jspec = JobSpec()
jspec.jobDefinitionID = job._getRoot().id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.transformation = '%s/runGen-00-00-02' % Client.baseURLSUB
if job.inputdata:
jspec.prodDBlock = job.inputdata.dataset[0]
else:
jspec.prodDBlock = 'NULL'
jspec.destinationDBlock = job.outputdata.datasetname
if job.outputdata.location:
if not job._getRoot().subjobs or job.id == 0:
logger.warning(
'You have specified outputdata.location. Note that Panda may not support writing to a user-defined output location.'
)
jspec.destinationSE = job.outputdata.location
else:
jspec.destinationSE = site
jspec.prodSourceLabel = configPanda['prodSourceLabelRun']
jspec.processingType = configPanda['processingType']
jspec.assignedPriority = configPanda['assignedPriorityRun']
jspec.cloud = cloud
# memory
if job.backend.requirements.memory != -1:
jspec.minRamCount = job.backend.requirements.memory
# cputime
if job.backend.requirements.cputime != -1:
jspec.maxCpuCount = job.backend.requirements.cputime
jspec.computingSite = site
# library (source files)
if job.backend.libds:
flib = FileSpec()
flib.lfn = self.fileBO.lfn
flib.GUID = self.fileBO.GUID
flib.type = 'input'
flib.status = self.fileBO.status
flib.dataset = self.fileBO.destinationDBlock
flib.dispatchDBlock = self.fileBO.destinationDBlock
jspec.addFile(flib)
elif job.backend.bexec:
flib = FileSpec()
flib.lfn = self.library
flib.type = 'input'
flib.dataset = self.libDataset
flib.dispatchDBlock = self.libDataset
jspec.addFile(flib)
# input files FIXME: many more input types
if job.inputdata:
for guid, lfn, size, checksum, scope in zip(
job.inputdata.guids, job.inputdata.names,
job.inputdata.sizes, job.inputdata.checksums,
job.inputdata.scopes):
finp = FileSpec()
finp.lfn = lfn
finp.GUID = guid
finp.scope = scope
# finp.fsize =
# finp.md5sum =
finp.dataset = job.inputdata.dataset[0]
finp.prodDBlock = job.inputdata.dataset[0]
finp.dispatchDBlock = job.inputdata.dataset[0]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# output files
# outMap = {}
#FIXME: if options.outMeta != []:
self.rundirectory = "."
# log files
flog = FileSpec()
flog.lfn = '%s._$PANDAID.log.tgz' % job.outputdata.datasetname
flog.type = 'log'
flog.dataset = job.outputdata.datasetname
flog.destinationDBlock = job.outputdata.datasetname
flog.destinationSE = job.backend.site
jspec.addFile(flog)
# job parameters
param = ''
# source URL
matchURL = re.search("(http.*://[^/]+)/", Client.baseURLCSRVSSL)
srcURL = ""
if matchURL != None:
srcURL = matchURL.group(1)
param += " --sourceURL %s " % srcURL
param += '-r "%s" ' % self.rundirectory
exe_name = job.application.exe
if job.backend.bexec == '':
if hasattr(job.application.exe, "name"):
exe_name = os.path.basename(job.application.exe.name)
# set jobO parameter
if job.application.args:
param += ' -j "" -p "%s %s" ' % (
exe_name, urllib.quote(" ".join(job.application.args)))
else:
param += ' -j "" -p "%s" ' % exe_name
if self.inputsandbox:
param += ' -a %s ' % self.inputsandbox
else:
param += '-l %s ' % self.library
param += '-j "" -p "%s %s" ' % (
exe_name, urllib.quote(" ".join(job.application.args)))
if job.inputdata:
param += '-i "%s" ' % job.inputdata.names
# fill outfiles
outfiles = {}
for f in self.extOutFile:
tarnum = 1
if f.find('*') != -1:
# archive *
outfiles[f] = "outputbox%i.%s.%s.tar.gz" % (
tarnum, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S"))
tarnum += 1
else:
outfiles[f] = "%s.%s.%s" % (f, job.getFQID('.'),
time.strftime("%Y%m%d%H%M%S"))
fout = FileSpec()
fout.lfn = outfiles[f]
fout.type = 'output'
fout.dataset = job.outputdata.datasetname
fout.destinationDBlock = job.outputdata.datasetname
fout.destinationSE = job.backend.site
jspec.addFile(fout)
param += '-o "%s" ' % (
outfiles
) # must be double quotes, because python prints strings in 'single quotes'
for file in jspec.Files:
if file.type in ['output', 'log'] and configPanda['chirpconfig']:
file.dispatchDBlockToken = configPanda['chirpconfig']
logger.debug('chirp file %s', file)
jspec.jobParameters = param
return jspec
file.destinationDBlock = job.destinationDBlock
file.destinationSE = job.destinationSE
file.dataset = job.destinationDBlock
file.type = 'output'
job.addFile(file)
fileOL = FileSpec()
fileOL.lfn = "%s.job.log.tgz" % job.jobName
fileOL.destinationDBlock = job.destinationDBlock
fileOL.destinationSE = job.destinationSE
fileOL.dataset = job.destinationDBlock
fileOL.type = 'log'
job.addFile(fileOL)
fileL = FileSpec()
fileL.dataset = 'user.TadashiMaeno.acas0003.lib._000134'
fileL.prodDBlock = fileL.dataset
fileL.lfn = 'user.TadashiMaeno.acas0003.lib._000134.lib.tgz'
fileL.type = 'input'
fileL.status = 'ready'
job.addFile(fileL)
job.jobParameters=("-l %s " % fileL.lfn) + """-r run/ -j "%20AnalysisSkeleton_topOptions.py" -i "[]" -m "[]" -n "[]" -o "{'AANT': [('AANTupleStream', 'AANT', """ + ("""'%s')]}" -c""" % file.lfn)
jobList.append(job)
s,o = Client.submitJobs(jobList)
print "---------------------"
print s
for x in o:
print "PandaID=%s" % x[0]
fileOZ.dataset = job.destinationDBlock
fileOZ.type = 'output'
job.addFile(fileOZ)
files = [
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00001.pool.root.1',
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00002.pool.root.1',
'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00003.pool.root.1',
]
for lfn in files:
fileI = FileSpec()
fileI.dataset = job.prodDBlock
fileI.prodDBlock = job.prodDBlock
fileI.lfn = lfn
fileI.type = 'input'
fileI.status = 'ready'
job.addFile(fileI)
fileL = FileSpec()
fileL.dataset = 'user.TadashiMaeno.lib._000157'
fileL.prodDBlock = 'user.TadashiMaeno.lib._000157'
fileL.lfn = 'user.TadashiMaeno.lib._000157.lib.tgz'
fileL.type = 'input'
fileL.status = 'ready'
job.addFile(fileL)
job.jobParameters=""" -l user.TadashiMaeno.lib._000157.lib.tgz -r run/ -j " AnalysisSkeleton_jobOptions.py" -i "['testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00001.pool.root.1', 'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00002.pool.root.1', 'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103._00003.pool.root.1']" -o "{'AANT': [('AANTupleStream', 'AANT', '%s')]}" """ % fileOZ.lfn
jobList.append(job)
file.type = 'output'
job.addFile(file)
fileOL = FileSpec()
fileOL.lfn = "%s.job.log.tgz" % job.jobName
fileOL.destinationDBlock = job.destinationDBlock
fileOL.destinationSE = job.destinationSE
fileOL.dataset = job.destinationDBlock
fileOL.type = 'log'
job.addFile(fileOL)
fileL = FileSpec()
fileL.dataset = 'user.TadashiMaeno.acas0003.lib._000134'
fileL.prodDBlock = fileL.dataset
fileL.lfn = 'user.TadashiMaeno.acas0003.lib._000134.lib.tgz'
fileL.type = 'input'
fileL.status = 'ready'
job.addFile(fileL)
job.jobParameters = (
"-l %s " % fileL.lfn
) + """-r run/ -j "%20AnalysisSkeleton_topOptions.py" -i "[]" -m "[]" -n "[]" -o "{'AANT': [('AANTupleStream', 'AANT', """ + (
"""'%s')]}" -c""" % file.lfn)
jobList.append(job)
s, o = Client.submitJobs(jobList)
print "---------------------"
print s
for x in o:
print "PandaID=%s" % x[0]
def send_job(jobid, siteid):
_logger.debug('Jobid: ' + str(jobid))
site = sites_.get(siteid)
job = jobs_.get(int(jobid))
cont = job.container
files_catalog = cont.files
fscope = getScope(job.owner.username)
datasetName = '{}:{}'.format(fscope, cont.guid)
distributive = job.distr.name
release = job.distr.release
# Prepare runScript
parameters = job.distr.command
parameters = parameters.replace("$COMMAND$", job.params)
parameters = parameters.replace("$USERNAME$", job.owner.username)
parameters = parameters.replace("$WORKINGGROUP$", job.owner.working_group)
# Prepare metadata
metadata = dict(user=job.owner.username)
# Prepare PanDA Object
pandajob = JobSpec()
pandajob.jobDefinitionID = int(time.time()) % 10000
pandajob.jobName = cont.guid
pandajob.transformation = client_config.DEFAULT_TRF
pandajob.destinationDBlock = datasetName
pandajob.destinationSE = site.se
pandajob.currentPriority = 1000
pandajob.prodSourceLabel = 'user'
pandajob.computingSite = site.ce
pandajob.cloud = 'RU'
pandajob.VO = 'atlas'
pandajob.prodDBlock = "%s:%s" % (fscope, pandajob.jobName)
pandajob.coreCount = job.corecount
pandajob.metadata = json.dumps(metadata)
#pandajob.workingGroup = job.owner.working_group
if site.encode_commands:
# It requires script wrapper on cluster side
pandajob.jobParameters = '%s %s %s "%s"' % (cont.guid, release, distributive, parameters)
else:
pandajob.jobParameters = parameters
has_input = False
for fcc in files_catalog:
if fcc.type == 'input':
f = fcc.file
guid = f.guid
fileIT = FileSpec()
fileIT.lfn = f.lfn
fileIT.dataset = pandajob.prodDBlock
fileIT.prodDBlock = pandajob.prodDBlock
fileIT.type = 'input'
fileIT.scope = fscope
fileIT.status = 'ready'
fileIT.GUID = guid
pandajob.addFile(fileIT)
has_input = True
if fcc.type == 'output':
f = fcc.file
fileOT = FileSpec()
fileOT.lfn = f.lfn
fileOT.destinationDBlock = pandajob.prodDBlock
fileOT.destinationSE = pandajob.destinationSE
fileOT.dataset = pandajob.prodDBlock
fileOT.type = 'output'
fileOT.scope = fscope
fileOT.GUID = f.guid
pandajob.addFile(fileOT)
# Save replica meta
fc.new_replica(f, site)
if not has_input:
# Add fake input
fileIT = FileSpec()
fileIT.lfn = "fake.input"
fileIT.dataset = pandajob.prodDBlock
fileIT.prodDBlock = pandajob.prodDBlock
fileIT.type = 'input'
fileIT.scope = fscope
fileIT.status = 'ready'
fileIT.GUID = "fake.guid"
pandajob.addFile(fileIT)
# Prepare lof file
fileOL = FileSpec()
fileOL.lfn = "%s.log.tgz" % pandajob.jobName
fileOL.destinationDBlock = pandajob.destinationDBlock
fileOL.destinationSE = pandajob.destinationSE
fileOL.dataset = '{}:logs'.format(fscope)
fileOL.type = 'log'
fileOL.scope = 'panda'
pandajob.addFile(fileOL)
# Save log meta
log = File()
log.scope = fscope
log.lfn = fileOL.lfn
log.guid = getGUID(log.scope, log.lfn)
log.type = 'log'
log.status = 'defined'
files_.save(log)
# Save replica meta
fc.new_replica(log, site)
# Register file in container
fc.reg_file_in_cont(log, cont, 'log')
# Submit job
o = submitJobs([pandajob])
x = o[0]
try:
#update PandaID
PandaID = int(x[0])
job.pandaid = PandaID
job.ce = site.ce
except:
job.status = 'submit_error'
jobs_.save(job)
return 0
def prepare(self, app, appconfig, appmasterconfig, jobmasterconfig):
'''prepare the subjob specific configuration'''
# PandaTools
from pandatools import Client
from pandatools import AthenaUtils
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('AthenaMCPandaRTHandler prepare called for %s',
job.getFQID('.'))
try:
assert self.outsite
except:
logger.error("outsite not set. Aborting")
raise Exception()
job.backend.site = self.outsite
job.backend.actualCE = self.outsite
cloud = job._getRoot().backend.requirements.cloud
job.backend.requirements.cloud = cloud
# now just filling the job from AthenaMC data
jspec = JobSpec()
jspec.jobDefinitionID = job._getRoot().id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.AtlasRelease = 'Atlas-%s' % app.atlas_rel
if app.transform_archive:
jspec.homepackage = 'AnalysisTransforms' + app.transform_archive
elif app.prod_release:
jspec.homepackage = 'AnalysisTransforms-AtlasProduction_' + str(
app.prod_release)
jspec.transformation = '%s/runAthena-00-00-11' % Client.baseURLSUB
#---->???? prodDBlock and destinationDBlock when facing several input / output datasets?
jspec.prodDBlock = 'NULL'
if job.inputdata and len(
app.inputfiles) > 0 and app.inputfiles[0] in app.dsetmap:
jspec.prodDBlock = app.dsetmap[app.inputfiles[0]]
# How to specify jspec.destinationDBlock when more than one type of output is available? Panda prod jobs seem to specify only the last output dataset
outdset = ""
for type in ["EVNT", "RDO", "HITS", "AOD", "ESD", "NTUP"]:
if type in app.outputpaths.keys():
outdset = string.replace(app.outputpaths[type], "/", ".")
outdset = outdset[1:-1]
break
if not outdset:
try:
assert len(app.outputpaths.keys()) > 0
except:
logger.error(
"app.outputpaths is empty: check your output datasets")
raise
type = app.outputpaths.keys()[0]
outdset = string.replace(app.outputpaths[type], "/", ".")
outdset = outdset[1:-1]
jspec.destinationDBlock = outdset
jspec.destinationSE = self.outsite
jspec.prodSourceLabel = 'user'
jspec.assignedPriority = 1000
jspec.cloud = cloud
# memory
if job.backend.requirements.memory != -1:
jspec.minRamCount = job.backend.requirements.memory
jspec.computingSite = self.outsite
jspec.cmtConfig = AthenaUtils.getCmtConfig(athenaVer=app.atlas_rel)
# library (source files)
flib = FileSpec()
flib.lfn = self.library
# flib.GUID =
flib.type = 'input'
# flib.status =
flib.dataset = self.libDataset
flib.dispatchDBlock = self.libDataset
jspec.addFile(flib)
# input files FIXME: many more input types
for lfn in app.inputfiles:
useguid = app.turls[lfn].replace("guid:", "")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# add dbfiles if any:
for lfn in app.dbfiles:
useguid = app.dbturls[lfn].replace("guid:", "")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# then minbias files
for lfn in app.mbfiles:
useguid = app.minbias_turls[lfn].replace("guid:", "")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# then cavern files
for lfn in app.cavernfiles:
useguid = app.cavern_turls[lfn].replace("guid:", "")
finp = FileSpec()
finp.lfn = lfn
finp.GUID = useguid
finp.dataset = app.dsetmap[lfn]
finp.prodDBlock = app.dsetmap[lfn]
finp.prodDBlockToken = 'local'
finp.dispatchDBlock = app.dsetmap[lfn]
finp.type = 'input'
finp.status = 'ready'
jspec.addFile(finp)
# output files( this includes the logfiles)
# Output files
jidtag = ""
job = app._getParent() # Returns job or subjob object
if job._getRoot().subjobs:
jidtag = job._getRoot().id
else:
jidtag = "%d" % job.id
outfiles = app.subjobsOutfiles[job.id]
pandaOutfiles = {}
for type in outfiles.keys():
pandaOutfiles[type] = outfiles[type] + "." + str(jidtag)
if type == "LOG":
pandaOutfiles[type] += ".tgz"
#print pandaOutfiles
for outtype in pandaOutfiles.keys():
fout = FileSpec()
dset = string.replace(app.outputpaths[outtype], "/", ".")
dset = dset[1:-1]
fout.dataset = dset
fout.lfn = pandaOutfiles[outtype]
fout.type = 'output'
# fout.destinationDBlock = jspec.destinationDBlock
fout.destinationDBlock = fout.dataset
fout.destinationSE = jspec.destinationSE
if outtype == 'LOG':
fout.type = 'log'
fout.destinationDBlock = fout.dataset
fout.destinationSE = job.backend.site
jspec.addFile(fout)
# job parameters
param = '-l %s ' % self.library # user tarball.
# use corruption checker
if job.backend.requirements.corCheck:
param += '--corCheck '
# disable to skip missing files
if job.backend.requirements.notSkipMissing:
param += '--notSkipMissing '
# transform parameters
# need to update arglist with final output file name...
newArgs = []
if app.mode == "evgen":
app.args[3] = app.args[3] + " -t "
if app.verbosity:
app.args[3] = app.args[3] + " -l %s " % app.verbosity
for arg in app.args[3:]:
for type in outfiles.keys():
if arg.find(outfiles[type]) > -1:
arg = arg.replace(outfiles[type], pandaOutfiles[type])
newArgs.append(arg)
arglist = string.join(newArgs, " ")
# print "Arglist:",arglist
param += ' -r ./ '
param += ' -j "%s"' % urllib.quote(arglist)
allinfiles = app.inputfiles + app.dbfiles
# Input files.
param += ' -i "%s" ' % allinfiles
if len(app.mbfiles) > 0:
param += ' -m "%s" ' % app.mbfiles
if len(app.cavernfiles) > 0:
param += ' -n "%s" ' % app.cavernfiles
# param += '-m "[]" ' #%minList FIXME
# param += '-n "[]" ' #%cavList FIXME
del pandaOutfiles[
"LOG"] # logfiles do not appear in IROOT block, and this one is not needed anymore...
param += ' -o "{\'IROOT\':%s }"' % str(pandaOutfiles.items())
# source URL
matchURL = re.search("(http.*://[^/]+)/", Client.baseURLSSL)
if matchURL != None:
param += " --sourceURL %s " % matchURL.group(1)
param += " --trf"
jspec.jobParameters = param
jspec.metadata = "--trf \"%s\"" % arglist
#print "SUBJOB DETAILS:",jspec.values()
if app.dryrun:
print "job.application.dryrun activated, printing out job parameters"
print jspec.values()
return
return jspec
