def createJobSpec(nodes, walltime, command, jobName, outputFile=None):
transformation = '#json#'
datasetName = 'panda.destDB.%s' % subprocess.check_output('uuidgen')
destName = 'local'
prodSourceLabel = 'user'
currentPriority = 1000
job = JobSpec()
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = jobName
job.VO = VO
job.transformation = transformation
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = currentPriority
job.prodSourceLabel = prodSourceLabel
job.computingSite = QUEUE_NAME
job.cmtConfig = json.dumps({'name': job.jobName, 'next': None})
lqcd_command = {
"nodes": nodes,
"walltime": walltime,
"name": job.jobName,
"command": command
}
if (outputFile):
lqcd_command['outputFile'] = outputFile
job.jobParameters = json.dumps(lqcd_command)
fileOL = FileSpec()
fileOL.lfn = "%s.job.log.tgz" % job.jobName.strip()
fileOL.destinationDBlock = job.destinationDBlock
fileOL.destinationSE = job.destinationSE
fileOL.dataset = job.destinationDBlock
fileOL.type = 'log'
job.addFile(fileOL)
#job.cmtConfig = None
return job
def defineEvgen16Job(self, i):
"""Define an Evgen16 job based on predefined values and randomly generated names
"""
job = JobSpec()
job.computingSite = self.__site
job.cloud = self.__cloud
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = "%s_%d" % (uuid.uuid1(), i)
job.AtlasRelease = 'Atlas-16.6.2'
job.homepackage = 'AtlasProduction/16.6.2.1'
job.transformation = 'Evgen_trf.py'
job.destinationDBlock = self.__datasetName
job.destinationSE = self.__destName
job.currentPriority = 10000
job.prodSourceLabel = 'test'
job.cmtConfig = 'i686-slc5-gcc43-opt'
#Output file
fileO = FileSpec()
fileO.lfn = "%s.evgen.pool.root" % job.jobName
fileO.destinationDBlock = job.destinationDBlock
fileO.destinationSE = job.destinationSE
fileO.dataset = job.destinationDBlock
fileO.destinationDBlockToken = 'ATLASDATADISK'
fileO.type = 'output'
job.addFile(fileO)
#Log file
fileL = FileSpec()
fileL.lfn = "%s.job.log.tgz" % job.jobName
fileL.destinationDBlock = job.destinationDBlock
fileL.destinationSE = job.destinationSE
fileL.dataset = job.destinationDBlock
fileL.destinationDBlockToken = 'ATLASDATADISK'
fileL.type = 'log'
job.addFile(fileL)
job.jobParameters = "2760 105048 19901 101 200 MC10.105048.PythiaB_ccmu3mu1X.py %s NONE NONE NONE MC10JobOpts-latest-test.tar.gz" % fileO.lfn
return job
def defineEvgen16Job(self, i):
"""Define an Evgen16 job based on predefined values and randomly generated names
"""
job = JobSpec()
job.computingSite = self.__site
job.cloud = self.__cloud
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = "%s_%d" % (uuid.uuid1(), i)
job.AtlasRelease = 'Atlas-16.6.2'
job.homepackage = 'AtlasProduction/16.6.2.1'
job.transformation = 'Evgen_trf.py'
job.destinationDBlock = self.__datasetName
job.destinationSE = self.__destName
job.currentPriority = 10000
job.prodSourceLabel = 'test'
job.cmtConfig = 'i686-slc5-gcc43-opt'
#Output file
fileO = FileSpec()
fileO.lfn = "%s.evgen.pool.root" % job.jobName
fileO.destinationDBlock = job.destinationDBlock
fileO.destinationSE = job.destinationSE
fileO.dataset = job.destinationDBlock
fileO.destinationDBlockToken = 'ATLASDATADISK'
fileO.type = 'output'
job.addFile(fileO)
#Log file
fileL = FileSpec()
fileL.lfn = "%s.job.log.tgz" % job.jobName
fileL.destinationDBlock = job.destinationDBlock
fileL.destinationSE = job.destinationSE
fileL.dataset = job.destinationDBlock
fileL.destinationDBlockToken = 'ATLASDATADISK'
fileL.type = 'log'
job.addFile(fileL)
job.jobParameters = "2760 105048 19901 101 200 MC10.105048.PythiaB_ccmu3mu1X.py %s NONE NONE NONE MC10JobOpts-latest-test.tar.gz" % fileO.lfn
return job
def run(self):
for i in range(1):
prodDBlock = 'rome.004201.evgen.ZeeJimmy'
destinationDBlock = 'pandatest.000123.test.simul'
destinationSE = 'BNL_SE'
jobs = []
#for i in range(self.interval):
for i in range(2):
job = JobSpec()
job.jobDefinitionID=self.jobDefinitionID
job.AtlasRelease='Atlas-11.0.1'
job.prodDBlock=prodDBlock
job.destinationDBlock=destinationDBlock
job.destinationSE=destinationSE
job.currentPriority=i
lfnI = 'rome.004201.evgen.ZeeJimmy._00001.pool.root'
file = FileSpec()
file.lfn = lfnI
file.dataset = 'rome.004201.evgen.ZeeJimmy'
file.type = 'input'
file.prodDBlock = prodDBlock
file.dataset = prodDBlock
job.addFile(file)
lfnO ='%s.pool.root.1' % commands.getoutput('uuidgen')
file = FileSpec()
file.lfn = lfnO
file.type = 'output'
file.destinationDBlock = destinationDBlock
file.dataset = destinationDBlock
file.destinationSE = destinationSE
job.addFile(file)
job.homepackage='JobTransforms-11-00-01-01'
job.transformation='share/rome.g4sim.standard.trf'
job.jobParameters='%s %s 1 2 14268' % (lfnI,lfnO)
jobs.append(job)
self.taskbuffer.storeJobs(jobs,None)
time.sleep(self.interval)
def createJob(self,
name,
nodes,
walltime,
command,
inputs=None,
queuename=None):
job = JobSpec()
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = "%s" % commands.getoutput('uuidgen')
job.VO = self.vo
job.transformation = self.transformation
job.destinationDBlock = self.datasetName
job.destinationSE = self.destName
job.currentPriority = self.currentPriority
job.prodSourceLabel = self.prodSourceLabel
job.computingSite = self.site if queuename is None else queuename
lqcd_command = {
"nodes": nodes,
"walltime": walltime,
"name": name,
"command": command
}
job.jobParameters = json.dumps(lqcd_command)
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)
job.cmtConfig = inputs
return job
job.homepackage = 'JobTransforms'
for ver in vers:
job.homepackage += "-%02d" % int(ver)
else:
job.homepackage = 'AtlasProduction/%s' % trfVer
# trf
job.transformation = trf
job.destinationDBlock = oDatasets[0]
# prod DBlock
job.prodDBlock = iDataset
# souce lavel
job.prodSeriesLabel = 'pandatest'
job.prodSourceLabel = 'managed'
# priority
job.assignedPriority = priority
job.currentPriority = priority
# CPU, memory,disk ### FIXME
# attempt number ### FIXME
# input files
if iDataset != 'NULL':
# remove _tidXXX
pat = re.sub('_tid\d+$', '', iDataset)
# search
m = re.search('(' + pat + '\S+)', line)
if m != None:
file = FileSpec()
file.lfn = m.group(1)
file.type = 'input'
file.dataset = iDataset
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
def send_merging_job(task, files_list, merge_chunk_number):
logger.info(
'Going to send merging job for task %s run number %s and merge chunk number %s'
% (task, files_list[0].run_number, merge_chunk_number))
input_files_copy = ''
input_files_rm = ''
for j in files_list:
TMPDUMPFILE = 'evtdump%(prodSlt)s-%(runChunk)s-%(runNumber)s.raw' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
if j.task.site == 'BW_COMPASS_MCORE':
input_files_copy += ' cp $dumpspath/' + TMPDUMPFILE + ' .;'
else:
input_files_copy += ' xrdcp -N -f $dumpspath/' + TMPDUMPFILE + ' .;'
datasetName = '%(prodNameOnly)s.%(runNumber)s-%(prodSlt)s-%(phastVer)s-merging-dump' % {
'prodNameOnly': task.production,
'runNumber': j.run_number,
'prodSlt': task.prodslt,
'phastVer': task.phastver
}
logger.info(datasetName)
destName = 'local' # PanDA will not try to move output data, data will be placed by pilot (based on schedconfig)
MERGEDDUMPFILE = 'evtdump%(prodSlt)s-%(runNumber)s.raw' % {
'runNumber': j.run_number,
'prodSlt': task.prodslt
}
if format(merge_chunk_number, '03d') != '000':
MERGEDDUMPFILE = MERGEDDUMPFILE + '.' + format(merge_chunk_number,
'03d')
logger.info(MERGEDDUMPFILE)
PRODSOFT = task.soft
ProdPathAndName = task.home + task.path + task.soft
if j.task.site == 'BW_COMPASS_MCORE':
dumpsPath = '/scratch/sciteam/criedl/projectdata/' + task.path + task.soft + '/evtdump/slot' + str(
task.prodslt)
else:
dumpsPath = 'root://eoscompass.cern.ch//eos/experiment/compass/' + task.path + task.soft + '/evtdump/slot' + str(
task.prodslt)
job = JobSpec()
job.VO = 'vo.compass.cern.ch'
job.taskID = task.id
job.jobDefinitionID = 0
job.jobName = '%(prodNameOnly)s-merge-dump-%(runNumber)s-ch%(mergeChunkNumber)s' % {
'prodNameOnly': task.production,
'runNumber': j.run_number,
'mergeChunkNumber': format(merge_chunk_number, '03d')
}
job.transformation = 'merging dump' # payload (can be URL as well)
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 5000
job.prodSourceLabel = 'prod_test'
job.computingSite = task.site
job.attemptNr = j.attempt_merging_evntdmp + 1
job.maxAttempt = j.task.max_attempts
if j.status_merging_evntdmp == 'failed':
job.parentID = j.panda_id_merging_evntdmp
if j.task.site == 'BW_COMPASS_MCORE':
job.jobParameters = 'ppwd=$(pwd);ppwd=$(pwd);export COMPASS_SW_PREFIX=/scratch/sciteam/criedl/projectdata/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export prodSlt=%(prodSlt)s;export MERGEDDUMPFILE=%(MERGEDDUMPFILE)s;export dumpspath=%(dumpsPath)s;export PRODSOFT=%(PRODSOFT)s;%(input_files_copy)scat evtdump%(prodSlt)s-*-*.raw > %(MERGEDDUMPFILE)s;rm evtdump%(prodSlt)s-*-*.raw;' % {
'MERGEDDUMPFILE': MERGEDDUMPFILE,
'dumpsPath': dumpsPath,
'PRODSOFT': PRODSOFT,
'input_files_copy': input_files_copy,
'ProdPathAndName': ProdPathAndName,
'prodPath': task.path,
'prodName': task.production,
'prodSlt': task.prodslt
}
else:
job.jobParameters = 'export EOS_MGM_URL=root://eoscompass.cern.ch;ppwd=$(pwd);ppwd=$(pwd);export COMPASS_SW_PREFIX=/eos/experiment/compass/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export prodSlt=%(prodSlt)s;export MERGEDDUMPFILE=%(MERGEDDUMPFILE)s;export dumpspath=%(dumpsPath)s;export PRODSOFT=%(PRODSOFT)s;%(input_files_copy)scat evtdump%(prodSlt)s-*-*.raw > %(MERGEDDUMPFILE)s;rm evtdump%(prodSlt)s-*-*.raw;' % {
'MERGEDDUMPFILE': MERGEDDUMPFILE,
'dumpsPath': dumpsPath,
'PRODSOFT': PRODSOFT,
'input_files_copy': input_files_copy,
'ProdPathAndName': ProdPathAndName,
'prodPath': task.path,
'prodName': task.production,
'prodSlt': task.prodslt
}
fileOLog = FileSpec()
fileOLog.lfn = "%s.job.log.tgz" % (job.jobName)
fileOLog.destinationDBlock = job.destinationDBlock
fileOLog.destinationSE = job.destinationSE
fileOLog.dataset = job.destinationDBlock
fileOLog.type = 'log'
job.addFile(fileOLog)
fileOdump = FileSpec()
fileOdump.lfn = "%s" % (MERGEDDUMPFILE)
fileOdump.destinationDBlock = job.destinationDBlock
fileOdump.destinationSE = job.destinationSE
fileOdump.dataset = job.destinationDBlock
fileOdump.type = 'output'
job.addFile(fileOdump)
# fileOstdout = FileSpec()
# fileOstdout.lfn = "payload_stdout.txt"
# fileOstdout.destinationDBlock = job.destinationDBlock
# fileOstdout.destinationSE = job.destinationSE
# fileOstdout.dataset = job.destinationDBlock
# fileOstdout.type = 'output'
# job.addFile(fileOstdout)
# fileOstderr = FileSpec()
# fileOstderr.lfn = "payload_stderr.txt"
# fileOstderr.destinationDBlock = job.destinationDBlock
# fileOstderr.destinationSE = job.destinationSE
# fileOstderr.dataset = job.destinationDBlock
# fileOstderr.type = 'output'
# job.addFile(fileOstderr)
s, o = Client.submitJobs([job], srvID=aSrvID)
logger.info(s)
for x in o:
logger.info("PandaID=%s" % x[0])
if x[0] != 0 and x[0] != 'NULL':
for j in files_list:
j_update = Job.objects.get(id=j.id)
j_update.panda_id_merging_evntdmp = x[0]
j_update.status_merging_evntdmp = 'sent'
j_update.attempt_merging_evntdmp = j_update.attempt_merging_evntdmp + 1
j_update.chunk_number_merging_evntdmp = merge_chunk_number
j_update.date_updated = today
try:
j_update.save()
logger.info('Job %s with PandaID %s updated' %
(j.id, x[0]))
except IntegrityError as e:
logger.exception('Unique together catched, was not saved')
except DatabaseError as e:
logger.exception('Something went wrong while saving: %s' %
e.message)
else:
logger.info('Job %s was not added to PanDA' % j.id)
def doBrokerage(self,inputList,vo,prodSourceLabel,workQueue):
# list with a lock
inputListWorld = ListWithLock([])
# variables for submission
maxBunchTask = 100
# make logger
tmpLog = MsgWrapper(logger)
tmpLog.debug('start doBrokerage')
# return for failure
retFatal = self.SC_FATAL
retTmpError = self.SC_FAILED
tmpLog.debug('vo={0} label={1} queue={2} nTasks={3}'.format(vo,prodSourceLabel,
workQueue.queue_name,
len(inputList)))
# loop over all tasks
allRwMap = {}
prioMap = {}
tt2Map = {}
expRWs = {}
jobSpecList = []
for tmpJediTaskID,tmpInputList in inputList:
for taskSpec,cloudName,inputChunk in tmpInputList:
# collect tasks for WORLD
if taskSpec.useWorldCloud():
inputListWorld.append((taskSpec,inputChunk))
continue
# make JobSpec to be submitted for TaskAssigner
jobSpec = JobSpec()
jobSpec.taskID = taskSpec.jediTaskID
jobSpec.jediTaskID = taskSpec.jediTaskID
# set managed to trigger TA
jobSpec.prodSourceLabel = 'managed'
jobSpec.processingType = taskSpec.processingType
jobSpec.workingGroup = taskSpec.workingGroup
jobSpec.metadata = taskSpec.processingType
jobSpec.assignedPriority = taskSpec.taskPriority
jobSpec.currentPriority = taskSpec.currentPriority
jobSpec.maxDiskCount = (taskSpec.getOutDiskSize() + taskSpec.getWorkDiskSize()) / 1024 / 1024
if taskSpec.useWorldCloud():
# use destinationSE to trigger task brokerage in WORLD cloud
jobSpec.destinationSE = taskSpec.cloud
prodDBlock = None
setProdDBlock = False
for datasetSpec in inputChunk.getDatasets():
prodDBlock = datasetSpec.datasetName
if datasetSpec.isMaster():
jobSpec.prodDBlock = datasetSpec.datasetName
setProdDBlock = True
for fileSpec in datasetSpec.Files:
tmpInFileSpec = fileSpec.convertToJobFileSpec(datasetSpec)
jobSpec.addFile(tmpInFileSpec)
# use secondary dataset name as prodDBlock
if setProdDBlock == False and prodDBlock != None:
jobSpec.prodDBlock = prodDBlock
# append
jobSpecList.append(jobSpec)
prioMap[jobSpec.taskID] = jobSpec.currentPriority
tt2Map[jobSpec.taskID] = jobSpec.processingType
# get RW for a priority
if not allRwMap.has_key(jobSpec.currentPriority):
tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(vo,prodSourceLabel,workQueue,
jobSpec.currentPriority)
if tmpRW == None:
tmpLog.error('failed to calculate RW with prio={0}'.format(jobSpec.currentPriority))
return retTmpError
allRwMap[jobSpec.currentPriority] = tmpRW
# get expected RW
expRW = self.taskBufferIF.calculateTaskRW_JEDI(jobSpec.jediTaskID)
if expRW == None:
tmpLog.error('failed to calculate RW for jediTaskID={0}'.format(jobSpec.jediTaskID))
return retTmpError
expRWs[jobSpec.taskID] = expRW
# for old clouds
if jobSpecList != []:
# get fullRWs
fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(vo,prodSourceLabel,None,None)
if fullRWs == None:
tmpLog.error('failed to calculate full RW')
return retTmpError
# set metadata
for jobSpec in jobSpecList:
rwValues = allRwMap[jobSpec.currentPriority]
jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (jobSpec.metadata,
str(rwValues),str(expRWs),
str(prioMap),str(fullRWs),
str(tt2Map))
tmpLog.debug('run task assigner for {0} tasks'.format(len(jobSpecList)))
nBunchTask = 0
while nBunchTask < len(jobSpecList):
# get a bunch
jobsBunch = jobSpecList[nBunchTask:nBunchTask+maxBunchTask]
strIDs = 'jediTaskID='
for tmpJobSpec in jobsBunch:
strIDs += '{0},'.format(tmpJobSpec.taskID)
strIDs = strIDs[:-1]
tmpLog.debug(strIDs)
# increment index
nBunchTask += maxBunchTask
# run task brokerge
stS,outSs = PandaClient.runTaskAssignment(jobsBunch)
tmpLog.debug('{0}:{1}'.format(stS,str(outSs)))
# for WORLD
if len(inputListWorld) > 0:
# thread pool
threadPool = ThreadPool()
# get full RW for WORLD
fullRWs = self.taskBufferIF.calculateWorldRWwithPrio_JEDI(vo,prodSourceLabel,None,None)
if fullRWs == None:
tmpLog.error('failed to calculate full WORLD RW')
return retTmpError
# get RW per priority
for taskSpec,inputChunk in inputListWorld:
if not taskSpec.currentPriority in allRwMap:
tmpRW = self.taskBufferIF.calculateWorldRWwithPrio_JEDI(vo,prodSourceLabel,workQueue,
taskSpec.currentPriority)
if tmpRW == None:
tmpLog.error('failed to calculate RW with prio={0}'.format(taskSpec.currentPriority))
return retTmpError
allRwMap[taskSpec.currentPriority] = tmpRW
# live counter for RWs
liveCounter = MapWithLock(allRwMap)
# make workers
ddmIF = self.ddmIF.getInterface(vo)
for iWorker in range(4):
thr = AtlasProdTaskBrokerThread(inputListWorld,threadPool,
self.taskBufferIF,ddmIF,
fullRWs,liveCounter)
thr.start()
threadPool.join(60*10)
# return
tmpLog.debug('doBrokerage done')
return self.SC_SUCCEEDED
def doBrokerage(self, inputList, vo, prodSourceLabel, workQueue):
# variables for submission
maxBunchTask = 100
# make logger
tmpLog = MsgWrapper(logger)
tmpLog.debug("start doBrokerage")
# return for failure
retFatal = self.SC_FATAL
retTmpError = self.SC_FAILED
tmpLog.debug("vo={0} label={1} queue={2}".format(vo, prodSourceLabel, workQueue.queue_name))
# loop over all tasks
allRwMap = {}
prioMap = {}
tt2Map = {}
expRWs = {}
jobSpecList = []
for tmpJediTaskID, tmpInputList in inputList:
for taskSpec, cloudName, inputChunk in tmpInputList:
# make JobSpec to be submitted for TaskAssigner
jobSpec = JobSpec()
jobSpec.taskID = taskSpec.jediTaskID
jobSpec.jediTaskID = taskSpec.jediTaskID
# set managed to trigger TA
jobSpec.prodSourceLabel = "managed"
jobSpec.processingType = taskSpec.processingType
jobSpec.workingGroup = taskSpec.workingGroup
jobSpec.metadata = taskSpec.processingType
jobSpec.assignedPriority = taskSpec.taskPriority
jobSpec.currentPriority = taskSpec.currentPriority
jobSpec.maxDiskCount = (taskSpec.getOutDiskSize() + taskSpec.getWorkDiskSize()) / 1024 / 1024
if taskSpec.useWorldCloud():
# use destinationSE to trigger task brokerage in WORLD cloud
jobSpec.destinationSE = taskSpec.cloud
prodDBlock = None
setProdDBlock = False
for datasetSpec in inputChunk.getDatasets():
prodDBlock = datasetSpec.datasetName
if datasetSpec.isMaster():
jobSpec.prodDBlock = datasetSpec.datasetName
setProdDBlock = True
for fileSpec in datasetSpec.Files:
tmpInFileSpec = fileSpec.convertToJobFileSpec(datasetSpec)
jobSpec.addFile(tmpInFileSpec)
# use secondary dataset name as prodDBlock
if setProdDBlock == False and prodDBlock != None:
jobSpec.prodDBlock = prodDBlock
# append
jobSpecList.append(jobSpec)
prioMap[jobSpec.taskID] = jobSpec.currentPriority
tt2Map[jobSpec.taskID] = jobSpec.processingType
# get RW for a priority
if not allRwMap.has_key(jobSpec.currentPriority):
tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(
vo, prodSourceLabel, workQueue, jobSpec.currentPriority
)
if tmpRW == None:
tmpLog.error("failed to calculate RW with prio={0}".format(jobSpec.currentPriority))
return retTmpError
allRwMap[jobSpec.currentPriority] = tmpRW
# get expected RW
expRW = self.taskBufferIF.calculateTaskRW_JEDI(jobSpec.jediTaskID)
if expRW == None:
tmpLog.error("failed to calculate RW for jediTaskID={0}".format(jobSpec.jediTaskID))
return retTmpError
expRWs[jobSpec.taskID] = expRW
# get fullRWs
fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(vo, prodSourceLabel, None, None)
if fullRWs == None:
tmpLog.error("failed to calculate full RW")
return retTmpError
# set metadata
for jobSpec in jobSpecList:
rwValues = allRwMap[jobSpec.currentPriority]
jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (
jobSpec.metadata,
str(rwValues),
str(expRWs),
str(prioMap),
str(fullRWs),
str(tt2Map),
)
tmpLog.debug("run task assigner for {0} tasks".format(len(jobSpecList)))
nBunchTask = 0
while nBunchTask < len(jobSpecList):
# get a bunch
jobsBunch = jobSpecList[nBunchTask : nBunchTask + maxBunchTask]
strIDs = "jediTaskID="
for tmpJobSpec in jobsBunch:
strIDs += "{0},".format(tmpJobSpec.taskID)
strIDs = strIDs[:-1]
tmpLog.debug(strIDs)
# increment index
nBunchTask += maxBunchTask
# run task brokerge
stS, outSs = PandaClient.runTaskAssignment(jobsBunch)
tmpLog.debug("{0}:{1}".format(stS, str(outSs)))
# return
tmpLog.debug("done")
return self.SC_SUCCEEDED
def prepare(self, app, appsubconfig, appmasterconfig, jobmasterconfig):
"""Prepare the specific aspec of each subjob.
Returns: subjobconfig list of objects understood by backends."""
from pandatools import Client
from pandatools import AthenaUtils
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
from GangaAtlas.Lib.ATLASDataset.DQ2Dataset import dq2_set_dataset_lifetime
from GangaPanda.Lib.Panda.Panda import refreshPandaSpecs
# make sure we have the correct siteType
refreshPandaSpecs()
job = app._getParent()
masterjob = job._getRoot()
logger.debug('ProdTransPandaRTHandler prepare called for %s',
job.getFQID('.'))
job.backend.actualCE = job.backend.site
job.backend.requirements.cloud = Client.PandaSites[
job.backend.site]['cloud']
# check that the site is in a submit-able status
if not job.splitter or job.splitter._name != 'DQ2JobSplitter':
allowed_sites = job.backend.list_ddm_sites()
try:
outDsLocation = Client.PandaSites[job.backend.site]['ddm']
tmpDsExist = False
if (configPanda['processingType'].startswith('gangarobot') or
configPanda['processingType'].startswith('hammercloud')):
#if Client.getDatasets(job.outputdata.datasetname):
if getDatasets(job.outputdata.datasetname):
tmpDsExist = True
logger.info('Re-using output dataset %s' %
job.outputdata.datasetname)
if not configPanda[
'specialHandling'] == 'ddm:rucio' and not configPanda[
'processingType'].startswith(
'gangarobot'
) and not configPanda['processingType'].startswith(
'hammercloud') and not configPanda[
'processingType'].startswith('rucio_test'):
Client.addDataset(job.outputdata.datasetname,
False,
location=outDsLocation,
allowProdDisk=True,
dsExist=tmpDsExist)
logger.info('Output dataset %s registered at %s' %
(job.outputdata.datasetname, outDsLocation))
dq2_set_dataset_lifetime(job.outputdata.datasetname, outDsLocation)
except exceptions.SystemExit:
raise BackendError(
'Panda', 'Exception in adding dataset %s: %s %s' %
(job.outputdata.datasetname, sys.exc_info()[0],
sys.exc_info()[1]))
# JobSpec.
jspec = JobSpec()
jspec.currentPriority = app.priority
jspec.jobDefinitionID = masterjob.id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.coreCount = app.core_count
jspec.AtlasRelease = 'Atlas-%s' % app.atlas_release
jspec.homepackage = app.home_package
jspec.transformation = app.transformation
jspec.destinationDBlock = job.outputdata.datasetname
if job.outputdata.location:
jspec.destinationSE = job.outputdata.location
else:
jspec.destinationSE = job.backend.site
if job.inputdata:
jspec.prodDBlock = job.inputdata.dataset[0]
else:
jspec.prodDBlock = 'NULL'
if app.prod_source_label:
jspec.prodSourceLabel = app.prod_source_label
else:
jspec.prodSourceLabel = configPanda['prodSourceLabelRun']
jspec.processingType = configPanda['processingType']
jspec.specialHandling = configPanda['specialHandling']
jspec.computingSite = job.backend.site
jspec.cloud = job.backend.requirements.cloud
jspec.cmtConfig = app.atlas_cmtconfig
if app.dbrelease == 'LATEST':
try:
latest_dbrelease = getLatestDBReleaseCaching()
except:
from pandatools import Client
latest_dbrelease = Client.getLatestDBRelease()
m = re.search('(.*):DBRelease-(.*)\.tar\.gz', latest_dbrelease)
if m:
self.dbrelease_dataset = m.group(1)
self.dbrelease = m.group(2)
else:
raise ApplicationConfigurationError(
None,
"Error retrieving LATEST DBRelease. Try setting application.dbrelease manually."
)
else:
self.dbrelease_dataset = app.dbrelease_dataset
self.dbrelease = app.dbrelease
jspec.jobParameters = app.job_parameters
if self.dbrelease:
if self.dbrelease == 'current':
jspec.jobParameters += ' --DBRelease=current'
else:
if jspec.transformation.endswith(
"_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --DBRelease=DBRelease-%s.tar.gz' % (
self.dbrelease, )
else:
jspec.jobParameters += ' DBRelease=DBRelease-%s.tar.gz' % (
self.dbrelease, )
dbspec = FileSpec()
dbspec.lfn = 'DBRelease-%s.tar.gz' % self.dbrelease
dbspec.dataset = self.dbrelease_dataset
dbspec.prodDBlock = jspec.prodDBlock
dbspec.type = 'input'
jspec.addFile(dbspec)
if job.inputdata:
m = re.search('(.*)\.(.*)\.(.*)\.(.*)\.(.*)\.(.*)',
job.inputdata.dataset[0])
if not m:
logger.error("Error retrieving run number from dataset name")
#raise ApplicationConfigurationError(None, "Error retrieving run number from dataset name")
runnumber = 105200
else:
runnumber = int(m.group(2))
if jspec.transformation.endswith(
"_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --runNumber %d' % runnumber
else:
jspec.jobParameters += ' RunNumber=%d' % runnumber
# Output files.
randomized_lfns = []
ilfn = 0
for lfn, lfntype in zip(app.output_files, app.output_type):
ofspec = FileSpec()
if app.randomize_lfns:
randomized_lfn = lfn + (
'.%s.%d.%s' %
(job.backend.site, int(time.time()),
commands.getoutput('uuidgen 2> /dev/null')[:4]))
else:
randomized_lfn = lfn
ofspec.lfn = randomized_lfn
randomized_lfns.append(randomized_lfn)
ofspec.destinationDBlock = jspec.destinationDBlock
ofspec.destinationSE = jspec.destinationSE
ofspec.dataset = jspec.destinationDBlock
ofspec.type = 'output'
jspec.addFile(ofspec)
if jspec.transformation.endswith(
"_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --output%sFile %s' % (
lfntype, randomized_lfns[ilfn])
else:
jspec.jobParameters += ' output%sFile=%s' % (
lfntype, randomized_lfns[ilfn])
ilfn = ilfn + 1
# Input files.
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):
ifspec = FileSpec()
ifspec.lfn = lfn
ifspec.GUID = guid
ifspec.fsize = size
ifspec.md5sum = checksum
ifspec.scope = scope
ifspec.dataset = jspec.prodDBlock
ifspec.prodDBlock = jspec.prodDBlock
ifspec.type = 'input'
jspec.addFile(ifspec)
if app.input_type:
itype = app.input_type
else:
itype = m.group(5)
if jspec.transformation.endswith(
"_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --input%sFile %s' % (itype, ','.join(
job.inputdata.names))
else:
jspec.jobParameters += ' input%sFile=%s' % (itype, ','.join(
job.inputdata.names))
# Log files.
lfspec = FileSpec()
lfspec.lfn = '%s.job.log.tgz' % jspec.jobName
lfspec.destinationDBlock = jspec.destinationDBlock
lfspec.destinationSE = jspec.destinationSE
lfspec.dataset = jspec.destinationDBlock
lfspec.type = 'log'
jspec.addFile(lfspec)
return jspec
def main():
logger.info('Getting tasks with status send and running')
# tasks_list = Task.objects.all().filter(Q(status='send') | Q(status='running'))
tasks_list = Task.objects.all().filter(name='dvcs2016P09t2r13v1_mu+')
logger.info('Got list of %s tasks' % len(tasks_list))
for t in tasks_list:
logger.info('Getting jobs in status defined or failed for task %s' % t)
jobs_list_count = Job.objects.all().filter(task=t).count()
if jobs_list_count > 50:
jobs_list = Job.objects.all().filter(
task=t).order_by('id')[:max_send_amount]
else:
jobs_list = Job.objects.all().filter(
task=t).order_by('id')[:jobs_list_count]
logger.info('Got list of %s jobs' % len(jobs_list))
i = 0
for j in jobs_list:
if i >= max_send_amount:
break
logger.info('Going to send job %s of %s task' %
(j.file, j.task.name))
umark = commands.getoutput('uuidgen')
datasetName = 'panda.destDB.%s' % umark
destName = 'COMPASSPRODDISK' # PanDA will not try to move output data, data will be placed by pilot (based on schedconfig)
TMPRAWFILE = j.file[j.file.rfind('/') + 1:]
logger.info(TMPRAWFILE)
TMPMDSTFILE = 'mDST-%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s.root' % {
'input_file': j.file,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
logger.info(TMPMDSTFILE)
TMPHISTFILE = '%(runNumber)s-%(runChunk)s-%(prodSlt)s.root' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(TMPHISTFILE)
TMPRICHFILE = 'gfile_%(runNumber)s-%(runChunk)s.gfile' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number
}
logger.info(TMPRICHFILE)
EVTDUMPFILE = 'evtdump%(prodSlt)s-%(runChunk)s-%(runNumber)s.raw' % {
'prodSlt': j.task.prodslt,
'runNumber': j.run_number,
'runChunk': j.chunk_number
}
logger.info(EVTDUMPFILE)
STDOUTFILE = '%(prodNameOnly)s.%(runNumber)s-%(runChunk)s-%(prodSlt)s.stdout' % {
'prodNameOnly': j.task.soft,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(STDOUTFILE)
STDERRFILE = '%(prodNameOnly)s.%(runNumber)s-%(runChunk)s-%(prodSlt)s.stderr' % {
'prodNameOnly': j.task.soft,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(STDERRFILE)
try:
file_year = j.file.split('/')[5]
logger.info(file_year)
except:
logger.error('Error while splitting file to get year')
sys.exit(1)
ProdPathAndName = j.task.home + j.task.path + j.task.soft
job = JobSpec()
job.taskID = j.task.id
job.jobDefinitionID = 0
job.jobName = '%(prodName)s-%(fileYear)s--%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s' % {
'prodName': j.task.soft,
'fileYear': file_year,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
job.transformation = j.task.type # payload (can be URL as well)
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 2000
job.prodSourceLabel = 'prod_test'
job.computingSite = site
job.attemptNr = j.attempt + 1
job.maxAttempt = j.task.max_attempts
if j.status == 'failed':
job.parentID = j.panda_id
head, tail = os.path.split(j.file)
# job.transferType = 'direct'
job.sourceSite = 'CERN_COMPASS_PROD'
# logs, and all files generated during execution will be placed in log (except output file)
#job.jobParameters='source /afs/cern.ch/project/eos/installation/compass/etc/setup.sh;export EOS_MGM_URL=root://eoscompass.cern.ch;export PATH=/afs/cern.ch/project/eos/installation/compass/bin:$PATH;ppwd=$(pwd);echo $ppwd;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;coralpath=%(ProdPathAndName)s/coral;echo $coralpath;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";echo $coralpathsetup;source $coralpathsetup;cd $ppwd;$CORAL/../phast/coral/coral.exe %(ProdPathAndName)s/template.opt;xrdcp -np $ppwd/%(TMPMDSTFILE)s xroot://eoscompass.cern.ch//eos/compass/%(prodName)s/mDST/%(TMPMDSTFILE)s;xrdcp -np $ppwd/%(TMPHISTFILE)s xroot://eoscompass.cern.ch//eos/compass/%(prodName)s/histos/%(TMPHISTFILE)s;metadataxml=$(ls metadata-*);echo $metadataxml;cp $metadataxml $metadataxml.PAYLOAD;' % {'TMPMDSTFILE': TMPMDSTFILE, 'TMPHISTFILE': TMPHISTFILE, 'TMPRICHFILE': TMPRICHFILE, 'input_file': input_file, 'ProdPathAndName': ProdPathAndName, 'prodName': prodName}
job.jobParameters = 'export EOS_MGM_URL=root://eoscompass.cern.ch;ppwd=$(pwd);export COMPASS_SW_PREFIX=/eos/experiment/compass/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export TMPRAWFILE=%(TMPRAWFILE)s;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;export prodSlt=%(prodSlt)s;export EVTDUMPFILE=%(EVTDUMPFILE)s;xrdcp -N -f root://castorpublic.cern.ch/%(input_file)s\?svcClass=compasscdr .;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;$CORAL/../phast/coral/coral.exe %(ProdPathAndName)s/%(template)s;if [ ! -s testevtdump.raw ]; then echo "PanDA message: the file is empty">testevtdump.raw; fi;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;rm %(tail)s' % {
'TMPRAWFILE': TMPRAWFILE,
'TMPMDSTFILE': TMPMDSTFILE,
'TMPHISTFILE': TMPHISTFILE,
'TMPRICHFILE': TMPRICHFILE,
'input_file': j.file,
'ProdPathAndName': ProdPathAndName,
'prodPath': j.task.path,
'prodName': j.task.soft,
'template': j.task.template,
'tail': tail,
'prodSlt': j.task.prodslt,
'EVTDUMPFILE': EVTDUMPFILE,
'STDOUTFILE': STDOUTFILE,
'STDERRFILE': STDERRFILE
}
fileIRaw = FileSpec()
fileIRaw.lfn = "%s" % (j.file)
fileIRaw.GUID = '5874a461-61d3-4543-8f34-6fd7a4624e78'
fileIRaw.fsize = 1073753368
fileIRaw.checksum = '671608be'
fileIRaw.destinationDBlock = job.destinationDBlock
fileIRaw.destinationSE = job.destinationSE
fileIRaw.dataset = job.destinationDBlock
fileIRaw.type = 'input'
job.addFile(fileIRaw)
fileOstdout = FileSpec()
fileOstdout.lfn = "payload_stdout.txt"
fileOstdout.destinationDBlock = job.destinationDBlock
fileOstdout.destinationSE = job.destinationSE
fileOstdout.dataset = job.destinationDBlock
fileOstdout.type = 'output'
job.addFile(fileOstdout)
fileOstderr = FileSpec()
fileOstderr.lfn = "payload_stderr.txt"
fileOstderr.destinationDBlock = job.destinationDBlock
fileOstderr.destinationSE = job.destinationSE
fileOstderr.dataset = job.destinationDBlock
fileOstderr.type = 'output'
job.addFile(fileOstderr)
fileOLog = FileSpec()
fileOLog.lfn = "%(prodName)s-%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s.job.log.tgz" % {
'prodName': j.task.soft,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
fileOLog.destinationDBlock = job.destinationDBlock
fileOLog.destinationSE = job.destinationSE
fileOLog.dataset = job.destinationDBlock
fileOLog.type = 'log'
job.addFile(fileOLog)
fileOmDST = FileSpec()
fileOmDST.lfn = "%s" % (TMPMDSTFILE)
fileOmDST.destinationDBlock = job.destinationDBlock
fileOmDST.destinationSE = job.destinationSE
fileOmDST.dataset = job.destinationDBlock
fileOmDST.type = 'output'
job.addFile(fileOmDST)
fileOTrafdic = FileSpec()
fileOTrafdic.lfn = "%s" % (TMPHISTFILE)
fileOTrafdic.destinationDBlock = job.destinationDBlock
fileOTrafdic.destinationSE = job.destinationSE
fileOTrafdic.dataset = job.destinationDBlock
fileOTrafdic.type = 'output'
job.addFile(fileOTrafdic)
fileOtestevtdump = FileSpec()
fileOtestevtdump.lfn = "testevtdump.raw"
fileOtestevtdump.destinationDBlock = job.destinationDBlock
fileOtestevtdump.destinationSE = job.destinationSE
fileOtestevtdump.dataset = job.destinationDBlock
fileOtestevtdump.type = 'output'
job.addFile(fileOtestevtdump)
s, o = Client.submitJobs([job], srvID=aSrvID)
logger.info(s)
logger.info(o)
# for x in o:
# logger.info("PandaID=%s" % x[0])
# today = datetime.datetime.today()
#
# if x[0] != 0 and x[0] != 'NULL':
# j_update = Job.objects.get(id=j.id)
# j_update.panda_id = x[0]
# j_update.status = 'sent'
# j_update.attempt = j_update.attempt + 1
# j_update.date_updated = today
#
# try:
# j_update.save()
# logger.info('Job %s with PandaID %s updated at %s' % (j.id, x[0], today))
# except IntegrityError as e:
# logger.exception('Unique together catched, was not saved')
# except DatabaseError as e:
# logger.exception('Something went wrong while saving: %s' % e.message)
# else:
# logger.info('Job %s was not added to PanDA' % j.id)
i += 1
logger.info('done')
def main():
logger.info('Getting tasks with status send and running')
# tasks_list = Task.objects.all().filter(Q(status='send') | Q(status='running'))
tasks_list = Task.objects.all().filter(name='TestTaskBW')
logger.info('Got list of %s tasks' % len(tasks_list))
for t in tasks_list:
logger.info('Getting jobs in status defined or failed for task %s' % t)
jobs_list_count = Job.objects.all().filter(task=t).count()
if jobs_list_count > 50:
jobs_list = Job.objects.all().filter(
task=t).order_by('id')[:max_send_amount]
else:
jobs_list = Job.objects.all().filter(
task=t).order_by('id')[:jobs_list_count]
logger.info('Got list of %s jobs' % len(jobs_list))
i = 0
for j in jobs_list:
if i >= max_send_amount:
break
logger.info('Going to send job %s of %s task' %
(j.file, j.task.name))
umark = commands.getoutput('uuidgen')
datasetName = 'panda.destDB.%s' % umark
destName = 'local' # PanDA will not try to move output data, data will be placed by pilot (based on schedconfig)
job = JobSpec()
job.taskID = j.task.id
job.jobDefinitionID = 0
job.jobName = 'hello world'
job.transformation = j.task.type # payload (can be URL as well)
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 2000
job.prodSourceLabel = 'test'
job.computingSite = site
job.attemptNr = 1
job.maxAttempt = 5
job.sourceSite = 'BW_COMPASS_MCORE'
job.VO = 'local'
# logs, and all files generated during execution will be placed in log (except output file)
job.jobParameters = 'python /u/sciteam/petrosya/panda/hello.py'
fileOLog = FileSpec()
fileOLog.lfn = "log.job.log.tgz"
fileOLog.destinationDBlock = job.destinationDBlock
fileOLog.destinationSE = job.destinationSE
fileOLog.dataset = job.destinationDBlock
fileOLog.type = 'log'
job.addFile(fileOLog)
s, o = Client.submitJobs([job], srvID=aSrvID)
logger.info(s)
logger.info(o)
# for x in o:
# logger.info("PandaID=%s" % x[0])
# today = datetime.datetime.today()
#
# if x[0] != 0 and x[0] != 'NULL':
# j_update = Job.objects.get(id=j.id)
# j_update.panda_id = x[0]
# j_update.status = 'sent'
# j_update.attempt = j_update.attempt + 1
# j_update.date_updated = today
#
# try:
# j_update.save()
# logger.info('Job %s with PandaID %s updated at %s' % (j.id, x[0], today))
# except IntegrityError as e:
# logger.exception('Unique together catched, was not saved')
# except DatabaseError as e:
# logger.exception('Something went wrong while saving: %s' % e.message)
# else:
# logger.info('Job %s was not added to PanDA' % j.id)
i += 1
logger.info('done')
def run(self):
try:
_logger.debug('%s Start %s' % (self.pandaID,self.job.jobStatus))
flagComplete = True
ddmJobs = []
topUserDsList = []
usingMerger = False
disableNotifier = False
firstIndvDS = True
finalStatusDS = []
for destinationDBlock in self.destinationDBlocks:
dsList = []
_logger.debug('%s start %s' % (self.pandaID,destinationDBlock))
# ignore tid datasets
if re.search('_tid[\d_]+$',destinationDBlock):
_logger.debug('%s skip %s' % (self.pandaID,destinationDBlock))
continue
# ignore HC datasets
if re.search('^hc_test\.',destinationDBlock) != None or re.search('^user\.gangarbt\.',destinationDBlock) != None:
if re.search('_sub\d+$',destinationDBlock) == None and re.search('\.lib$',destinationDBlock) == None:
_logger.debug('%s skip HC %s' % (self.pandaID,destinationDBlock))
continue
# query dataset
if self.datasetMap.has_key(destinationDBlock):
dataset = self.datasetMap[destinationDBlock]
else:
dataset = self.taskBuffer.queryDatasetWithMap({'name':destinationDBlock})
if dataset == None:
_logger.error('%s Not found : %s' % (self.pandaID,destinationDBlock))
flagComplete = False
continue
# skip tobedeleted/tobeclosed
if dataset.status in ['cleanup','tobeclosed','completed']:
_logger.debug('%s skip %s due to %s' % (self.pandaID,destinationDBlock,dataset.status))
continue
dsList.append(dataset)
# sort
dsList.sort()
# count number of completed files
notFinish = self.taskBuffer.countFilesWithMap({'destinationDBlock':destinationDBlock,
'status':'unknown'})
if notFinish < 0:
_logger.error('%s Invalid DB return : %s' % (self.pandaID,notFinish))
flagComplete = False
continue
# check if completed
_logger.debug('%s notFinish:%s' % (self.pandaID,notFinish))
if self.job.destinationSE == 'local' and self.job.prodSourceLabel in ['user','panda']:
# close non-DQ2 destinationDBlock immediately
finalStatus = 'closed'
elif self.job.lockedby == 'jedi' and self.isTopLevelDS(destinationDBlock):
# set it closed in order not to trigger DDM cleanup. It will be closed by JEDI
finalStatus = 'closed'
elif self.job.prodSourceLabel in ['user'] and "--mergeOutput" in self.job.jobParameters \
and self.job.processingType != 'usermerge':
# merge output files
if firstIndvDS:
# set 'tobemerged' to only the first dataset to avoid triggering many Mergers for --individualOutDS
finalStatus = 'tobemerged'
firstIndvDS = False
else:
finalStatus = 'tobeclosed'
# set merging to top dataset
usingMerger = True
# disable Notifier
disableNotifier = True
elif self.job.produceUnMerge():
finalStatus = 'doing'
else:
# set status to 'tobeclosed' to trigger DQ2 closing
finalStatus = 'tobeclosed'
if notFinish==0:
_logger.debug('%s set %s to dataset : %s' % (self.pandaID,finalStatus,destinationDBlock))
# set status
dataset.status = finalStatus
# update dataset in DB
retT = self.taskBuffer.updateDatasets(dsList,withLock=True,withCriteria="status<>:crStatus AND status<>:lockStatus ",
criteriaMap={':crStatus':finalStatus,':lockStatus':'locked'})
if len(retT) > 0 and retT[0]==1:
finalStatusDS += dsList
# close user datasets
if self.job.prodSourceLabel in ['user'] and self.job.destinationDBlock.endswith('/') \
and (dataset.name.startswith('user') or dataset.name.startswith('group')):
# get top-level user dataset
topUserDsName = re.sub('_sub\d+$','',dataset.name)
# update if it is the first attempt
if topUserDsName != dataset.name and not topUserDsName in topUserDsList and self.job.lockedby != 'jedi':
topUserDs = self.taskBuffer.queryDatasetWithMap({'name':topUserDsName})
if topUserDs != None:
# check status
if topUserDs.status in ['completed','cleanup','tobeclosed',
'tobemerged','merging']:
_logger.debug('%s skip %s due to status=%s' % (self.pandaID,topUserDsName,topUserDs.status))
else:
# set status
if self.job.processingType.startswith('gangarobot') or \
self.job.processingType.startswith('hammercloud'):
# not trigger freezing for HC datasets so that files can be appended
topUserDs.status = 'completed'
elif not usingMerger:
topUserDs.status = finalStatus
else:
topUserDs.status = 'merging'
# append to avoid repetition
topUserDsList.append(topUserDsName)
# update DB
retTopT = self.taskBuffer.updateDatasets([topUserDs],withLock=True,withCriteria="status<>:crStatus",
criteriaMap={':crStatus':topUserDs.status})
if len(retTopT) > 0 and retTopT[0]==1:
_logger.debug('%s set %s to top dataset : %s' % (self.pandaID,topUserDs.status,topUserDsName))
else:
_logger.debug('%s failed to update top dataset : %s' % (self.pandaID,topUserDsName))
# get parent dataset for merge job
if self.job.processingType == 'usermerge':
tmpMatch = re.search('--parentDS ([^ \'\"]+)',self.job.jobParameters)
if tmpMatch == None:
_logger.error('%s failed to extract parentDS' % self.pandaID)
else:
unmergedDsName = tmpMatch.group(1)
# update if it is the first attempt
if not unmergedDsName in topUserDsList:
unmergedDs = self.taskBuffer.queryDatasetWithMap({'name':unmergedDsName})
if unmergedDs == None:
_logger.error('%s failed to get parentDS=%s from DB' % (self.pandaID,unmergedDsName))
else:
# check status
if unmergedDs.status in ['completed','cleanup','tobeclosed']:
_logger.debug('%s skip %s due to status=%s' % (self.pandaID,unmergedDsName,unmergedDs.status))
else:
# set status
unmergedDs.status = finalStatus
# append to avoid repetition
topUserDsList.append(unmergedDsName)
# update DB
retTopT = self.taskBuffer.updateDatasets([unmergedDs],withLock=True,withCriteria="status<>:crStatus",
criteriaMap={':crStatus':unmergedDs.status})
if len(retTopT) > 0 and retTopT[0]==1:
_logger.debug('%s set %s to parent dataset : %s' % (self.pandaID,unmergedDs.status,unmergedDsName))
else:
_logger.debug('%s failed to update parent dataset : %s' % (self.pandaID,unmergedDsName))
if self.pandaDDM and self.job.prodSourceLabel=='managed':
# instantiate SiteMapper
if self.siteMapper == None:
self.siteMapper = SiteMapper(self.taskBuffer)
# get file list for PandaDDM
retList = self.taskBuffer.queryFilesWithMap({'destinationDBlock':destinationDBlock})
lfnsStr = ''
guidStr = ''
for tmpFile in retList:
if tmpFile.type in ['log','output']:
lfnsStr += '%s,' % tmpFile.lfn
guidStr += '%s,' % tmpFile.GUID
if lfnsStr != '':
guidStr = guidStr[:-1]
lfnsStr = lfnsStr[:-1]
# create a DDM job
ddmjob = JobSpec()
ddmjob.jobDefinitionID = int(time.time()) % 10000
ddmjob.jobName = "%s" % commands.getoutput('uuidgen')
ddmjob.transformation = 'http://pandaserver.cern.ch:25080/trf/mover/run_dq2_cr'
ddmjob.destinationDBlock = 'testpanda.%s' % ddmjob.jobName
ddmjob.computingSite = "BNL_ATLAS_DDM"
ddmjob.destinationSE = ddmjob.computingSite
ddmjob.currentPriority = 200000
ddmjob.prodSourceLabel = 'ddm'
ddmjob.transferType = 'sub'
# append log file
fileOL = FileSpec()
fileOL.lfn = "%s.job.log.tgz" % ddmjob.jobName
fileOL.destinationDBlock = ddmjob.destinationDBlock
fileOL.destinationSE = ddmjob.destinationSE
fileOL.dataset = ddmjob.destinationDBlock
fileOL.type = 'log'
ddmjob.addFile(fileOL)
# make arguments
dstDQ2ID = 'BNLPANDA'
srcDQ2ID = self.siteMapper.getSite(self.job.computingSite).ddm
callBackURL = 'https://%s:%s/server/panda/datasetCompleted?vuid=%s&site=%s' % \
(panda_config.pserverhost,panda_config.pserverport,
dataset.vuid,dstDQ2ID)
_logger.debug(callBackURL)
# set src/dest
ddmjob.sourceSite = srcDQ2ID
ddmjob.destinationSite = dstDQ2ID
# if src==dst, send callback without ddm job
if dstDQ2ID == srcDQ2ID:
comout = commands.getoutput('curl -k %s' % callBackURL)
_logger.debug(comout)
else:
# run dq2_cr
callBackURL = urllib.quote(callBackURL)
# get destination dir
destDir = brokerage.broker_util._getDefaultStorage(self.siteMapper.getSite(self.job.computingSite).dq2url)
argStr = "-s %s -r %s --guids %s --lfns %s --callBack %s -d %s/%s %s" % \
(srcDQ2ID,dstDQ2ID,guidStr,lfnsStr,callBackURL,destDir,
destinationDBlock,destinationDBlock)
# set job parameters
ddmjob.jobParameters = argStr
_logger.debug('%s pdq2_cr %s' % (self.pandaID,ddmjob.jobParameters))
ddmJobs.append(ddmjob)
# start Activator
if re.search('_sub\d+$',dataset.name) == None:
if self.job.prodSourceLabel=='panda' and self.job.processingType in ['merge','unmerge']:
# don't trigger Activator for merge jobs
pass
else:
if self.job.jobStatus == 'finished':
aThr = Activator(self.taskBuffer,dataset)
aThr.start()
aThr.join()
else:
# unset flag since another thread already updated
#flagComplete = False
pass
else:
# update dataset in DB
self.taskBuffer.updateDatasets(dsList,withLock=True,withCriteria="status<>:crStatus AND status<>:lockStatus ",
criteriaMap={':crStatus':finalStatus,':lockStatus':'locked'})
# unset flag
flagComplete = False
# end
_logger.debug('%s end %s' % (self.pandaID,destinationDBlock))
# start DDM jobs
if ddmJobs != []:
self.taskBuffer.storeJobs(ddmJobs,self.job.prodUserID,joinThr=True)
# change pending jobs to failed
finalizedFlag = True
if flagComplete and self.job.prodSourceLabel=='user':
_logger.debug('%s finalize %s %s' % (self.pandaID,self.job.prodUserName,self.job.jobDefinitionID))
finalizedFlag = self.taskBuffer.finalizePendingJobs(self.job.prodUserName,self.job.jobDefinitionID,waitLock=True)
_logger.debug('%s finalized with %s' % (self.pandaID,finalizedFlag))
# update unmerged datasets in JEDI to trigger merging
if flagComplete and self.job.produceUnMerge() and finalStatusDS != []:
if finalizedFlag:
self.taskBuffer.updateUnmergedDatasets(self.job,finalStatusDS)
# start notifier
_logger.debug('%s source:%s complete:%s' % (self.pandaID,self.job.prodSourceLabel,flagComplete))
if (self.job.jobStatus != 'transferring') and ((flagComplete and self.job.prodSourceLabel=='user') or \
(self.job.jobStatus=='failed' and self.job.prodSourceLabel=='panda')) and \
self.job.lockedby != 'jedi':
# don't send email for merge jobs
if (not disableNotifier) and not self.job.processingType in ['merge','unmerge']:
useNotifier = True
summaryInfo = {}
# check all jobDefIDs in jobsetID
if not self.job.jobsetID in [0,None,'NULL']:
useNotifier,summaryInfo = self.taskBuffer.checkDatasetStatusForNotifier(self.job.jobsetID,self.job.jobDefinitionID,
self.job.prodUserName)
_logger.debug('%s useNotifier:%s' % (self.pandaID,useNotifier))
if useNotifier:
_logger.debug('%s start Notifier' % self.pandaID)
nThr = Notifier.Notifier(self.taskBuffer,self.job,self.destinationDBlocks,summaryInfo)
nThr.run()
_logger.debug('%s end Notifier' % self.pandaID)
_logger.debug('%s End' % self.pandaID)
except:
errType,errValue = sys.exc_info()[:2]
_logger.error("%s %s" % (errType,errValue))
job.homepackage = 'JobTransforms'
for ver in vers:
job.homepackage += "-%02d" % int(ver)
else:
job.homepackage = 'AtlasProduction/%s' % trfVer
# trf
job.transformation = trf
job.destinationDBlock = oDatasets[0]
# prod DBlock
job.prodDBlock = iDataset
# souce lavel
job.prodSeriesLabel = 'pandatest'
job.prodSourceLabel = 'managed'
# priority
job.assignedPriority = priority
job.currentPriority = priority
# CPU, memory,disk ### FIXME
# attempt number ### FIXME
# input files
if iDataset != 'NULL':
# remove _tidXXX
pat = re.sub('_tid\d+$','',iDataset)
# search
m = re.search('('+pat+'\S+)',line)
if m != None:
file = FileSpec()
file.lfn = m.group(1)
file.type = 'input'
file.dataset = iDataset
def send_merging_job(task, files_list, merge_chunk_number):
logger.info(
'Going to send merging job for task %s run number %s and merge chunk number %s'
% (task, files_list[0].run_number, merge_chunk_number))
input_files = ''
input_files_copy = ''
for j in files_list:
TMPMDSTFILE = 'mDST-%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s.root' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
input_files += ' ' + TMPMDSTFILE
if j.task.site == 'BW_COMPASS_MCORE':
input_files_copy += ' cp /scratch/sciteam/criedl/projectdata/' + task.path + task.soft + '/mDST.chunks/' + TMPMDSTFILE + ' .;'
else:
input_files_copy += ' xrdcp -N -f root://eoscompass.cern.ch//eos/experiment/compass/' + task.path + task.soft + '/mDST.chunks/' + TMPMDSTFILE + ' .;'
datasetName = '%(prodNameOnly)s.%(runNumber)s-%(prodSlt)s-%(phastVer)s-merging-mdst' % {
'prodNameOnly': task.production,
'runNumber': j.run_number,
'prodSlt': task.prodslt,
'phastVer': task.phastver
}
logger.info(datasetName)
destName = 'local' # PanDA will not try to move output data, data will be placed by pilot (based on schedconfig)
MERGEDHISTFILE = '%(runNumber)s-%(prodSlt)s-%(phastVer)s.root' % {
'runNumber': j.run_number,
'prodSlt': task.prodslt,
'phastVer': task.phastver
}
if format(merge_chunk_number, '03d') != '000':
MERGEDHISTFILE = MERGEDHISTFILE + '.' + format(merge_chunk_number,
'03d')
logger.info(MERGEDHISTFILE)
MERGEDMDSTFILE = 'mDST-%(runNumber)s-%(prodSlt)s-%(phastVer)s.root' % {
'runNumber': j.run_number,
'prodSlt': task.prodslt,
'phastVer': task.phastver
}
if format(merge_chunk_number, '03d') != '000':
MERGEDMDSTFILE = MERGEDMDSTFILE + '.' + format(merge_chunk_number,
'03d')
logger.info(MERGEDMDSTFILE)
TMPHISTFILE = 'merge-%(runNumber)s-ch%(mergeChunkNumber)s.root' % {
'runNumber': j.run_number,
'mergeChunkNumber': format(merge_chunk_number, '03d')
}
logger.info(TMPHISTFILE)
PRODSOFT = task.soft
ProdPathAndName = task.home + task.path + task.soft
job = JobSpec()
job.VO = 'vo.compass.cern.ch'
job.taskID = task.id
job.jobDefinitionID = 0
job.jobName = '%(prodNameOnly)s-merge-mdst-%(runNumber)s-ch%(mergeChunkNumber)s' % {
'prodNameOnly': task.production,
'runNumber': j.run_number,
'mergeChunkNumber': format(merge_chunk_number, '03d')
}
job.transformation = 'merging mdst' # payload (can be URL as well)
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 5000
job.prodSourceLabel = 'prod_test'
job.computingSite = task.site
job.attemptNr = j.attempt_merging_mdst + 1
job.maxAttempt = j.task.max_attempts
if j.status_merging_mdst == 'failed':
job.parentID = j.panda_id_merging_mdst
if j.task.site == 'BW_COMPASS_MCORE':
job.jobParameters = 'ppwd=$(pwd);ppwd=$(pwd);export COMPASS_SW_PREFIX=/scratch/sciteam/criedl/projectdata/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export prodSlt=%(prodSlt)s;export MERGEDHISTFILE=%(MERGEDHISTFILE)s;export MERGEDMDSTFILE=%(MERGEDMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export PRODSOFT=%(PRODSOFT)s;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;%(input_files_copy)sexport PHAST_mDST_MAX_SIZE=6000000000;$CORAL/../phast/phast -m -o %(MERGEDMDSTFILE)s %(input_files)s;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;gzip payload_stdout.out;' % {
'MERGEDHISTFILE': MERGEDHISTFILE,
'MERGEDMDSTFILE': MERGEDMDSTFILE,
'PRODSOFT': PRODSOFT,
'input_files_copy': input_files_copy,
'input_files': input_files,
'ProdPathAndName': ProdPathAndName,
'prodPath': task.path,
'prodName': task.production,
'prodSlt': task.prodslt,
'TMPHISTFILE': TMPHISTFILE
}
else:
job.jobParameters = 'export EOS_MGM_URL=root://eoscompass.cern.ch;ppwd=$(pwd);ppwd=$(pwd);export COMPASS_SW_PREFIX=/eos/experiment/compass/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export prodSlt=%(prodSlt)s;export MERGEDHISTFILE=%(MERGEDHISTFILE)s;export MERGEDMDSTFILE=%(MERGEDMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export PRODSOFT=%(PRODSOFT)s;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;%(input_files_copy)sexport PHAST_mDST_MAX_SIZE=6000000000;$CORAL/../phast/phast -m -o %(MERGEDMDSTFILE)s %(input_files)s;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;gzip payload_stdout.out;' % {
'MERGEDHISTFILE': MERGEDHISTFILE,
'MERGEDMDSTFILE': MERGEDMDSTFILE,
'PRODSOFT': PRODSOFT,
'input_files_copy': input_files_copy,
'input_files': input_files,
'ProdPathAndName': ProdPathAndName,
'prodPath': task.path,
'prodName': task.production,
'prodSlt': task.prodslt,
'TMPHISTFILE': TMPHISTFILE
}
fileOLog = FileSpec()
fileOLog.lfn = "%s.job.log.tgz" % (job.jobName)
fileOLog.destinationDBlock = job.destinationDBlock
fileOLog.destinationSE = job.destinationSE
fileOLog.dataset = job.destinationDBlock
fileOLog.type = 'log'
job.addFile(fileOLog)
fileOmDST = FileSpec()
fileOmDST.lfn = "%s" % (MERGEDMDSTFILE)
fileOmDST.destinationDBlock = job.destinationDBlock
fileOmDST.destinationSE = job.destinationSE
fileOmDST.dataset = job.destinationDBlock
fileOmDST.type = 'output'
job.addFile(fileOmDST)
fileOstdout = FileSpec()
fileOstdout.lfn = "payload_stdout.out.gz"
fileOstdout.destinationDBlock = job.destinationDBlock
fileOstdout.destinationSE = job.destinationSE
fileOstdout.dataset = job.destinationDBlock
fileOstdout.type = 'output'
job.addFile(fileOstdout)
# fileOstderr = FileSpec()
# fileOstderr.lfn = "payload_stderr.txt"
# fileOstderr.destinationDBlock = job.destinationDBlock
# fileOstderr.destinationSE = job.destinationSE
# fileOstderr.dataset = job.destinationDBlock
# fileOstderr.type = 'output'
# job.addFile(fileOstderr)
s, o = Client.submitJobs([job], srvID=aSrvID)
logger.info(s)
for x in o:
logger.info("PandaID=%s" % x[0])
if x[0] != 0 and x[0] != 'NULL':
for j in files_list:
j_update = Job.objects.get(id=j.id)
j_update.panda_id_merging_mdst = x[0]
j_update.status_merging_mdst = 'sent'
j_update.attempt_merging_mdst = j_update.attempt_merging_mdst + 1
j_update.chunk_number_merging_mdst = merge_chunk_number
j_update.date_updated = today
try:
j_update.save()
logger.info('Job %s with PandaID %s updated' %
(j.id, x[0]))
except IntegrityError as e:
logger.exception('Unique together catched, was not saved')
except DatabaseError as e:
logger.exception('Something went wrong while saving: %s' %
e.message)
else:
logger.info('Job %s was not added to PanDA' % j.id)
def doBrokerage(self, inputList, vo, prodSourceLabel, workQueue):
# variables for submission
maxBunchTask = 100
# make logger
tmpLog = MsgWrapper(logger)
tmpLog.debug('start doBrokerage')
# return for failure
retFatal = self.SC_FATAL
retTmpError = self.SC_FAILED
tmpLog.debug('vo={0} label={1} queue={2}'.format(
vo, prodSourceLabel, workQueue.queue_name))
# loop over all tasks
allRwMap = {}
prioMap = {}
tt2Map = {}
expRWs = {}
jobSpecList = []
for tmpJediTaskID, tmpInputList in inputList:
for taskSpec, cloudName, inputChunk in tmpInputList:
# make JobSpec to be submitted for TaskAssigner
jobSpec = JobSpec()
jobSpec.taskID = taskSpec.jediTaskID
jobSpec.jediTaskID = taskSpec.jediTaskID
# set managed to trigger TA
jobSpec.prodSourceLabel = 'managed'
jobSpec.processingType = taskSpec.processingType
jobSpec.workingGroup = taskSpec.workingGroup
jobSpec.metadata = taskSpec.processingType
jobSpec.assignedPriority = taskSpec.taskPriority
jobSpec.currentPriority = taskSpec.currentPriority
jobSpec.maxDiskCount = (
taskSpec.getOutDiskSize() +
taskSpec.getWorkDiskSize()) / 1024 / 1024
if taskSpec.useWorldCloud():
# use destinationSE to trigger task brokerage in WORLD cloud
jobSpec.destinationSE = taskSpec.cloud
prodDBlock = None
setProdDBlock = False
for datasetSpec in inputChunk.getDatasets():
prodDBlock = datasetSpec.datasetName
if datasetSpec.isMaster():
jobSpec.prodDBlock = datasetSpec.datasetName
setProdDBlock = True
for fileSpec in datasetSpec.Files:
tmpInFileSpec = fileSpec.convertToJobFileSpec(
datasetSpec)
jobSpec.addFile(tmpInFileSpec)
# use secondary dataset name as prodDBlock
if setProdDBlock == False and prodDBlock != None:
jobSpec.prodDBlock = prodDBlock
# append
jobSpecList.append(jobSpec)
prioMap[jobSpec.taskID] = jobSpec.currentPriority
tt2Map[jobSpec.taskID] = jobSpec.processingType
# get RW for a priority
if not allRwMap.has_key(jobSpec.currentPriority):
tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(
vo, prodSourceLabel, workQueue,
jobSpec.currentPriority)
if tmpRW == None:
tmpLog.error(
'failed to calculate RW with prio={0}'.format(
jobSpec.currentPriority))
return retTmpError
allRwMap[jobSpec.currentPriority] = tmpRW
# get expected RW
expRW = self.taskBufferIF.calculateTaskRW_JEDI(
jobSpec.jediTaskID)
if expRW == None:
tmpLog.error(
'failed to calculate RW for jediTaskID={0}'.format(
jobSpec.jediTaskID))
return retTmpError
expRWs[jobSpec.taskID] = expRW
# get fullRWs
fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(
vo, prodSourceLabel, None, None)
if fullRWs == None:
tmpLog.error('failed to calculate full RW')
return retTmpError
# set metadata
for jobSpec in jobSpecList:
rwValues = allRwMap[jobSpec.currentPriority]
jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (
jobSpec.metadata, str(rwValues), str(expRWs), str(prioMap),
str(fullRWs), str(tt2Map))
tmpLog.debug('run task assigner for {0} tasks'.format(
len(jobSpecList)))
nBunchTask = 0
while nBunchTask < len(jobSpecList):
# get a bunch
jobsBunch = jobSpecList[nBunchTask:nBunchTask + maxBunchTask]
strIDs = 'jediTaskID='
for tmpJobSpec in jobsBunch:
strIDs += '{0},'.format(tmpJobSpec.taskID)
strIDs = strIDs[:-1]
tmpLog.debug(strIDs)
# increment index
nBunchTask += maxBunchTask
# run task brokerge
stS, outSs = PandaClient.runTaskAssignment(jobsBunch)
tmpLog.debug('{0}:{1}'.format(stS, str(outSs)))
# return
tmpLog.debug('done')
return self.SC_SUCCEEDED
destName = 'BNL_SE'
jobDefinitionID = int(time.time()) % 10000
jobList = []
for i in range(2):
job = JobSpec()
job.jobDefinitionID = jobDefinitionID
job.jobName = "%s_%d" % (commands.getoutput('uuidgen'),i)
job.AtlasRelease = 'Atlas-12.0.6'
job.homepackage = 'AnalysisTransforms'
job.transformation = 'https://gridui01.usatlas.bnl.gov:24443/dav/test/runAthenaXrd'
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 3000
job.assignedPriority = 3000
job.prodSourceLabel = 'user'
job.computingSite = site
file = FileSpec()
file.lfn = "%s.AANT._%05d.root" % (job.jobName,i)
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
site = None
jobList = []
for i in range(2):
datasetName = 'panda.destDB.%s' % commands.getoutput('uuidgen')
destName = 'ANALY_BNL_ATLAS_1'
job = JobSpec()
job.jobDefinitionID = 1
job.jobName = commands.getoutput('uuidgen')
job.AtlasRelease = 'Atlas-12.0.2'
job.homepackage = 'AnalysisTransforms'
job.transformation = 'https://gridui01.usatlas.bnl.gov:24443/dav/test/runAthena2'
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 3000
job.prodSourceLabel = 'user'
job.computingSite = site
job.prodDBlock = 'testIdeal_06.005001.pythia_minbias.recon.AOD.v12000103'
fileOL = FileSpec()
fileOL.lfn = "%s.job.log.tgz" % commands.getoutput('uuidgen')
fileOL.destinationDBlock = job.destinationDBlock
fileOL.destinationSE = job.destinationSE
fileOL.dataset = job.destinationDBlock
fileOL.type = 'log'
job.addFile(fileOL)
fileOZ = FileSpec()
fileOZ.lfn = "AANT.%s.root" % commands.getoutput('uuidgen')
fileOZ.destinationDBlock = job.destinationDBlock
def doBrokerage(self, inputList, vo, prodSourceLabel, workQueue,
resource_name):
# list with a lock
inputListWorld = ListWithLock([])
# variables for submission
maxBunchTask = 100
# make logger
tmpLog = MsgWrapper(logger)
tmpLog.debug('start doBrokerage')
# return for failure
retFatal = self.SC_FATAL
retTmpError = self.SC_FAILED
tmpLog.debug(
'vo={0} label={1} queue={2} resource_name={3} nTasks={4}'.format(
vo, prodSourceLabel, workQueue.queue_name, resource_name,
len(inputList)))
# loop over all tasks
allRwMap = {}
prioMap = {}
tt2Map = {}
expRWs = {}
jobSpecList = []
for tmpJediTaskID, tmpInputList in inputList:
for taskSpec, cloudName, inputChunk in tmpInputList:
# collect tasks for WORLD
if taskSpec.useWorldCloud():
inputListWorld.append((taskSpec, inputChunk))
continue
# make JobSpec to be submitted for TaskAssigner
jobSpec = JobSpec()
jobSpec.taskID = taskSpec.jediTaskID
jobSpec.jediTaskID = taskSpec.jediTaskID
# set managed to trigger TA
jobSpec.prodSourceLabel = 'managed'
jobSpec.processingType = taskSpec.processingType
jobSpec.workingGroup = taskSpec.workingGroup
jobSpec.metadata = taskSpec.processingType
jobSpec.assignedPriority = taskSpec.taskPriority
jobSpec.currentPriority = taskSpec.currentPriority
jobSpec.maxDiskCount = (
taskSpec.getOutDiskSize() +
taskSpec.getWorkDiskSize()) // 1024 // 1024
if taskSpec.useWorldCloud():
# use destinationSE to trigger task brokerage in WORLD cloud
jobSpec.destinationSE = taskSpec.cloud
prodDBlock = None
setProdDBlock = False
for datasetSpec in inputChunk.getDatasets():
prodDBlock = datasetSpec.datasetName
if datasetSpec.isMaster():
jobSpec.prodDBlock = datasetSpec.datasetName
setProdDBlock = True
for fileSpec in datasetSpec.Files:
tmpInFileSpec = fileSpec.convertToJobFileSpec(
datasetSpec)
jobSpec.addFile(tmpInFileSpec)
# use secondary dataset name as prodDBlock
if setProdDBlock is False and prodDBlock is not None:
jobSpec.prodDBlock = prodDBlock
# append
jobSpecList.append(jobSpec)
prioMap[jobSpec.taskID] = jobSpec.currentPriority
tt2Map[jobSpec.taskID] = jobSpec.processingType
# get RW for a priority
if jobSpec.currentPriority not in allRwMap:
tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(
vo, prodSourceLabel, workQueue,
jobSpec.currentPriority)
if tmpRW is None:
tmpLog.error(
'failed to calculate RW with prio={0}'.format(
jobSpec.currentPriority))
return retTmpError
allRwMap[jobSpec.currentPriority] = tmpRW
# get expected RW
expRW = self.taskBufferIF.calculateTaskRW_JEDI(
jobSpec.jediTaskID)
if expRW is None:
tmpLog.error(
'failed to calculate RW for jediTaskID={0}'.format(
jobSpec.jediTaskID))
return retTmpError
expRWs[jobSpec.taskID] = expRW
# for old clouds
if jobSpecList != []:
# get fullRWs
fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(
vo, prodSourceLabel, None, None)
if fullRWs is None:
tmpLog.error('failed to calculate full RW')
return retTmpError
# set metadata
for jobSpec in jobSpecList:
rwValues = allRwMap[jobSpec.currentPriority]
jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (
jobSpec.metadata, str(rwValues), str(expRWs), str(prioMap),
str(fullRWs), str(tt2Map))
tmpLog.debug('run task assigner for {0} tasks'.format(
len(jobSpecList)))
nBunchTask = 0
while nBunchTask < len(jobSpecList):
# get a bunch
jobsBunch = jobSpecList[nBunchTask:nBunchTask + maxBunchTask]
strIDs = 'jediTaskID='
for tmpJobSpec in jobsBunch:
strIDs += '{0},'.format(tmpJobSpec.taskID)
strIDs = strIDs[:-1]
tmpLog.debug(strIDs)
# increment index
nBunchTask += maxBunchTask
# run task brokerge
stS, outSs = PandaClient.runTaskAssignment(jobsBunch)
tmpLog.debug('{0}:{1}'.format(stS, str(outSs)))
# for WORLD
if len(inputListWorld) > 0:
# thread pool
threadPool = ThreadPool()
# get full RW for WORLD
fullRWs = self.taskBufferIF.calculateWorldRWwithPrio_JEDI(
vo, prodSourceLabel, None, None)
if fullRWs is None:
tmpLog.error('failed to calculate full WORLD RW')
return retTmpError
# get RW per priority
for taskSpec, inputChunk in inputListWorld:
if taskSpec.currentPriority not in allRwMap:
tmpRW = self.taskBufferIF.calculateWorldRWwithPrio_JEDI(
vo, prodSourceLabel, workQueue,
taskSpec.currentPriority)
if tmpRW is None:
tmpLog.error(
'failed to calculate RW with prio={0}'.format(
taskSpec.currentPriority))
return retTmpError
allRwMap[taskSpec.currentPriority] = tmpRW
# live counter for RWs
liveCounter = MapWithLock(allRwMap)
# make workers
ddmIF = self.ddmIF.getInterface(vo)
for iWorker in range(4):
thr = AtlasProdTaskBrokerThread(inputListWorld, threadPool,
self.taskBufferIF, ddmIF,
fullRWs, liveCounter,
workQueue)
thr.start()
threadPool.join(60 * 10)
# return
tmpLog.debug('doBrokerage done')
return self.SC_SUCCEEDED
def prepare(self, app, appsubconfig, appmasterconfig, jobmasterconfig):
"""Prepare the specific aspec of each subjob.
Returns: subjobconfig list of objects understood by backends."""
from pandatools import Client
from pandatools import AthenaUtils
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
from GangaAtlas.Lib.ATLASDataset.DQ2Dataset import dq2_set_dataset_lifetime
from GangaPanda.Lib.Panda.Panda import refreshPandaSpecs
# make sure we have the correct siteType
refreshPandaSpecs()
job = app._getParent()
masterjob = job._getRoot()
logger.debug('ProdTransPandaRTHandler prepare called for %s',
job.getFQID('.'))
job.backend.actualCE = job.backend.site
job.backend.requirements.cloud = Client.PandaSites[job.backend.site]['cloud']
# check that the site is in a submit-able status
if not job.splitter or job.splitter._name != 'DQ2JobSplitter':
allowed_sites = job.backend.list_ddm_sites()
try:
outDsLocation = Client.PandaSites[job.backend.site]['ddm']
tmpDsExist = False
if (configPanda['processingType'].startswith('gangarobot') or configPanda['processingType'].startswith('hammercloud')):
#if Client.getDatasets(job.outputdata.datasetname):
if getDatasets(job.outputdata.datasetname):
tmpDsExist = True
logger.info('Re-using output dataset %s'%job.outputdata.datasetname)
if not configPanda['specialHandling']=='ddm:rucio' and not configPanda['processingType'].startswith('gangarobot') and not configPanda['processingType'].startswith('hammercloud') and not configPanda['processingType'].startswith('rucio_test'):
Client.addDataset(job.outputdata.datasetname,False,location=outDsLocation,allowProdDisk=True,dsExist=tmpDsExist)
logger.info('Output dataset %s registered at %s'%(job.outputdata.datasetname,outDsLocation))
dq2_set_dataset_lifetime(job.outputdata.datasetname, outDsLocation)
except exceptions.SystemExit:
raise BackendError('Panda','Exception in adding dataset %s: %s %s'%(job.outputdata.datasetname,sys.exc_info()[0],sys.exc_info()[1]))
# JobSpec.
jspec = JobSpec()
jspec.currentPriority = app.priority
jspec.jobDefinitionID = masterjob.id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.coreCount = app.core_count
jspec.AtlasRelease = 'Atlas-%s' % app.atlas_release
jspec.homepackage = app.home_package
jspec.transformation = app.transformation
jspec.destinationDBlock = job.outputdata.datasetname
if job.outputdata.location:
jspec.destinationSE = job.outputdata.location
else:
jspec.destinationSE = job.backend.site
if job.inputdata:
jspec.prodDBlock = job.inputdata.dataset[0]
else:
jspec.prodDBlock = 'NULL'
if app.prod_source_label:
jspec.prodSourceLabel = app.prod_source_label
else:
jspec.prodSourceLabel = configPanda['prodSourceLabelRun']
jspec.processingType = configPanda['processingType']
jspec.specialHandling = configPanda['specialHandling']
jspec.computingSite = job.backend.site
jspec.cloud = job.backend.requirements.cloud
jspec.cmtConfig = app.atlas_cmtconfig
if app.dbrelease == 'LATEST':
try:
latest_dbrelease = getLatestDBReleaseCaching()
except:
from pandatools import Client
latest_dbrelease = Client.getLatestDBRelease()
m = re.search('(.*):DBRelease-(.*)\.tar\.gz', latest_dbrelease)
if m:
self.dbrelease_dataset = m.group(1)
self.dbrelease = m.group(2)
else:
raise ApplicationConfigurationError(None, "Error retrieving LATEST DBRelease. Try setting application.dbrelease manually.")
else:
self.dbrelease_dataset = app.dbrelease_dataset
self.dbrelease = app.dbrelease
jspec.jobParameters = app.job_parameters
if self.dbrelease:
if self.dbrelease == 'current':
jspec.jobParameters += ' --DBRelease=current'
else:
if jspec.transformation.endswith("_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --DBRelease=DBRelease-%s.tar.gz' % (self.dbrelease,)
else:
jspec.jobParameters += ' DBRelease=DBRelease-%s.tar.gz' % (self.dbrelease,)
dbspec = FileSpec()
dbspec.lfn = 'DBRelease-%s.tar.gz' % self.dbrelease
dbspec.dataset = self.dbrelease_dataset
dbspec.prodDBlock = jspec.prodDBlock
dbspec.type = 'input'
jspec.addFile(dbspec)
if job.inputdata:
m = re.search('(.*)\.(.*)\.(.*)\.(.*)\.(.*)\.(.*)',
job.inputdata.dataset[0])
if not m:
logger.error("Error retrieving run number from dataset name")
#raise ApplicationConfigurationError(None, "Error retrieving run number from dataset name")
runnumber = 105200
else:
runnumber = int(m.group(2))
if jspec.transformation.endswith("_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --runNumber %d' % runnumber
else:
jspec.jobParameters += ' RunNumber=%d' % runnumber
# Output files.
randomized_lfns = []
ilfn = 0
for lfn, lfntype in zip(app.output_files,app.output_type):
ofspec = FileSpec()
if app.randomize_lfns:
randomized_lfn = lfn + ('.%s.%d.%s' % (job.backend.site, int(time.time()), commands.getoutput('uuidgen 2> /dev/null')[:4] ) )
else:
randomized_lfn = lfn
ofspec.lfn = randomized_lfn
randomized_lfns.append(randomized_lfn)
ofspec.destinationDBlock = jspec.destinationDBlock
ofspec.destinationSE = jspec.destinationSE
ofspec.dataset = jspec.destinationDBlock
ofspec.type = 'output'
jspec.addFile(ofspec)
if jspec.transformation.endswith("_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --output%sFile %s' % (lfntype, randomized_lfns[ilfn])
else:
jspec.jobParameters += ' output%sFile=%s' % (lfntype, randomized_lfns[ilfn])
ilfn=ilfn+1
# Input files.
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):
ifspec = FileSpec()
ifspec.lfn = lfn
ifspec.GUID = guid
ifspec.fsize = size
ifspec.md5sum = checksum
ifspec.scope = scope
ifspec.dataset = jspec.prodDBlock
ifspec.prodDBlock = jspec.prodDBlock
ifspec.type = 'input'
jspec.addFile(ifspec)
if app.input_type:
itype = app.input_type
else:
itype = m.group(5)
if jspec.transformation.endswith("_tf.py") or jspec.transformation.endswith("_tf"):
jspec.jobParameters += ' --input%sFile %s' % (itype, ','.join(job.inputdata.names))
else:
jspec.jobParameters += ' input%sFile=%s' % (itype, ','.join(job.inputdata.names))
# Log files.
lfspec = FileSpec()
lfspec.lfn = '%s.job.log.tgz' % jspec.jobName
lfspec.destinationDBlock = jspec.destinationDBlock
lfspec.destinationSE = jspec.destinationSE
lfspec.dataset = jspec.destinationDBlock
lfspec.type = 'log'
jspec.addFile(lfspec)
return jspec
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
index = 0
for lfn in files.keys():
index += 1
job = JobSpec()
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = "%s_%d" % (commands.getoutput('uuidgen'), index)
job.AtlasRelease = 'Atlas-13.0.40'
job.homepackage = 'AtlasProduction/13.0.40.3'
job.transformation = 'csc_simul_trf.py'
job.destinationDBlock = datasetName
job.destinationSE = destName
job.computingSite = site
job.prodDBlock = 'valid1.005001.pythia_minbias.evgen.EVNT.e306_tid019128'
job.prodSourceLabel = 'test'
job.currentPriority = 10000
job.cloud = 'IT'
fileI = FileSpec()
fileI.dataset = job.prodDBlock
fileI.prodDBlock = job.prodDBlock
fileI.lfn = lfn
fileI.type = 'input'
job.addFile(fileI)
fileD = FileSpec()
fileD.dataset = 'ddo.000001.Atlas.Ideal.DBRelease.v040701'
fileD.prodDBlock = 'ddo.000001.Atlas.Ideal.DBRelease.v030101'
fileD.lfn = 'DBRelease-4.7.1.tar.gz'
fileD.type = 'input'
job.addFile(fileD)
def main():
logger.info('Getting tasks with status send and running')
tasks_list = Task.objects.all().filter(
Q(status='send') | Q(status='running'))
#tasks_list = Task.objects.all().filter(name='dvcs2017align7_mu-')
logger.info('Got list of %s tasks' % len(tasks_list))
cdbServerArr = ['compassvm23.cern.ch', 'compassvm24.cern.ch']
cdbServer = cdbServerArr[0]
for t in tasks_list:
max_send_amount = 1000
logger.info('Getting jobs in status staged or failed for task %s' % t)
jobs_list_count = Job.objects.all().filter(task=t).filter(
attempt__lt=t.max_attempts).filter(
Q(status='staged') | Q(status='failed')).count()
if jobs_list_count > 50:
jobs_list = Job.objects.all().filter(task=t).filter(
attempt__lt=t.max_attempts).filter(
Q(status='staged') | Q(status='failed')).order_by(
'-number_of_events')[:max_send_amount]
else:
jobs_list = Job.objects.all().filter(task=t).filter(
attempt__lt=t.max_attempts).filter(
Q(status='staged') | Q(status='failed')).order_by(
'-number_of_events')[:jobs_list_count]
logger.info('Got list of %s jobs' % len(jobs_list))
# jobs_list = Job.objects.all().filter(task=t).filter(file='/castor/cern.ch/compass/data/2017/raw/W04/cdr12116-278485.raw')
i = 0
for j in jobs_list:
if j.attempt >= j.task.max_attempts:
logger.info(
'Number of retry attempts has reached for job %s of task %s'
% (j.file, j.task.name))
continue
if i > max_send_amount:
break
logger.info('Job %s of %s' % (i, max_send_amount))
logger.info('Going to send job %s of %s task' %
(j.file, j.task.name))
umark = commands.getoutput('uuidgen')
datasetName = 'panda.destDB.%s' % umark
destName = 'local' # PanDA will not try to move output data, data will be placed by pilot (based on schedconfig)
TMPRAWFILE = j.file[j.file.rfind('/') + 1:]
logger.info(TMPRAWFILE)
TMPMDSTFILE = 'mDST-%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s.root' % {
'input_file': j.file,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
logger.info(TMPMDSTFILE)
TMPHISTFILE = '%(runNumber)s-%(runChunk)s-%(prodSlt)s.root' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(TMPHISTFILE)
TMPRICHFILE = 'gfile_%(runNumber)s-%(runChunk)s.gfile' % {
'runNumber': j.run_number,
'runChunk': j.chunk_number
}
logger.info(TMPRICHFILE)
EVTDUMPFILE = 'evtdump%(prodSlt)s-%(runChunk)s-%(runNumber)s.raw' % {
'prodSlt': j.task.prodslt,
'runNumber': j.run_number,
'runChunk': j.chunk_number
}
logger.info(EVTDUMPFILE)
STDOUTFILE = '%(prodNameOnly)s.%(runNumber)s-%(runChunk)s-%(prodSlt)s.stdout' % {
'prodNameOnly': j.task.production,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(STDOUTFILE)
STDERRFILE = '%(prodNameOnly)s.%(runNumber)s-%(runChunk)s-%(prodSlt)s.stderr' % {
'prodNameOnly': j.task.production,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt
}
logger.info(STDERRFILE)
PRODSOFT = j.task.soft
logger.info(PRODSOFT)
ProdPathAndName = j.task.home + j.task.path + j.task.soft
job = JobSpec()
job.VO = 'vo.compass.cern.ch'
job.taskID = j.task.id
job.jobDefinitionID = 0
job.jobName = '%(prodName)s-%(fileYear)s--%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s' % {
'prodName': j.task.production,
'fileYear': j.task.year,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
job.transformation = j.task.type # payload (can be URL as well)
job.destinationDBlock = datasetName
job.destinationSE = destName
job.currentPriority = 2000
if j.task.type == 'DDD filtering':
job.currentPriority = 1000
job.prodSourceLabel = 'prod_test'
job.computingSite = j.task.site
job.attemptNr = j.attempt + 1
job.maxAttempt = j.task.max_attempts
if j.status == 'failed':
job.parentID = j.panda_id
head, tail = os.path.split(j.file)
cdbServer = cdbServerArr[random.randrange(len(cdbServerArr))]
# logs, and all files generated during execution will be placed in log (except output file)
#job.jobParameters='source /afs/cern.ch/project/eos/installation/compass/etc/setup.sh;export EOS_MGM_URL=root://eoscompass.cern.ch;export PATH=/afs/cern.ch/project/eos/installation/compass/bin:$PATH;ppwd=$(pwd);echo $ppwd;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;coralpath=%(ProdPathAndName)s/coral;echo $coralpath;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";echo $coralpathsetup;source $coralpathsetup;cd $ppwd;$CORAL/../phast/coral/coral.exe %(ProdPathAndName)s/template.opt;xrdcp -np $ppwd/%(TMPMDSTFILE)s xroot://eoscompass.cern.ch//eos/compass/%(prodName)s/mDST/%(TMPMDSTFILE)s;xrdcp -np $ppwd/%(TMPHISTFILE)s xroot://eoscompass.cern.ch//eos/compass/%(prodName)s/histos/%(TMPHISTFILE)s;metadataxml=$(ls metadata-*);echo $metadataxml;cp $metadataxml $metadataxml.PAYLOAD;' % {'TMPMDSTFILE': TMPMDSTFILE, 'TMPHISTFILE': TMPHISTFILE, 'TMPRICHFILE': TMPRICHFILE, 'input_file': input_file, 'ProdPathAndName': ProdPathAndName, 'prodName': prodName}
if j.task.type == 'test production' or j.task.type == 'mass production' or j.task.type == 'technical production':
if j.task.site == 'BW_COMPASS_MCORE':
job.jobParameters = 'ppwd=$(pwd);export COMPASS_SW_PREFIX=/scratch/sciteam/criedl/projectdata/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export TMPRAWFILE=%(TMPRAWFILE)s;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;export prodSlt=%(prodSlt)s;export EVTDUMPFILE=%(EVTDUMPFILE)s;export PRODSOFT=%(PRODSOFT)s;cp %(input_file)s .;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;$CORAL/../phast/coral/coral.exe %(ProdPathAndName)s/%(template)s;if [ ! -s testevtdump.raw ]; then echo "PanDA message: the file is empty">testevtdump.raw; fi;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;gzip payload_stderr.out;gzip payload_stdout.out;rm %(tail)s' % {
'TMPRAWFILE': TMPRAWFILE,
'TMPMDSTFILE': TMPMDSTFILE,
'TMPHISTFILE': TMPHISTFILE,
'TMPRICHFILE': TMPRICHFILE,
'PRODSOFT': PRODSOFT,
'input_file': j.file,
'ProdPathAndName': ProdPathAndName,
'prodPath': j.task.path,
'prodName': j.task.production,
'template': j.task.template,
'tail': tail,
'prodSlt': j.task.prodslt,
'EVTDUMPFILE': EVTDUMPFILE,
'STDOUTFILE': STDOUTFILE,
'STDERRFILE': STDERRFILE
}
else:
job.jobParameters = 'export EOS_MGM_URL=root://eoscompass.cern.ch;ppwd=$(pwd);export COMPASS_SW_PREFIX=/eos/experiment/compass/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export TMPRAWFILE=%(TMPRAWFILE)s;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;export prodSlt=%(prodSlt)s;export EVTDUMPFILE=%(EVTDUMPFILE)s;export PRODSOFT=%(PRODSOFT)s;xrdcp -N -f root://castorpublic.cern.ch/%(input_file)s\?svcClass=compasscdr .;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;export CDBSERVER=%(cdbServer)s;$CORAL/../phast/coral/coral.exe %(ProdPathAndName)s/%(template)s;if [ ! -s testevtdump.raw ]; then echo "PanDA message: the file is empty">testevtdump.raw; fi;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;gzip payload_stderr.out;gzip payload_stdout.out;rm %(tail)s' % {
'TMPRAWFILE': TMPRAWFILE,
'TMPMDSTFILE': TMPMDSTFILE,
'TMPHISTFILE': TMPHISTFILE,
'TMPRICHFILE': TMPRICHFILE,
'PRODSOFT': PRODSOFT,
'input_file': j.file,
'ProdPathAndName': ProdPathAndName,
'prodPath': j.task.path,
'prodName': j.task.production,
'template': j.task.template,
'tail': tail,
'prodSlt': j.task.prodslt,
'EVTDUMPFILE': EVTDUMPFILE,
'STDOUTFILE': STDOUTFILE,
'STDERRFILE': STDERRFILE,
'cdbServer': cdbServer
}
if j.task.type == 'DDD filtering':
job.jobParameters = 'export EOS_MGM_URL=root://eoscompass.cern.ch;ppwd=$(pwd);export COMPASS_SW_PREFIX=/eos/experiment/compass/;export COMPASS_SW_PATH=%(prodPath)s;export COMPASS_PROD_NAME=%(prodName)s;export TMPRAWFILE=%(TMPRAWFILE)s;export TMPMDSTFILE=%(TMPMDSTFILE)s;export TMPHISTFILE=%(TMPHISTFILE)s;export TMPRICHFILE=%(TMPRICHFILE)s;export prodSlt=%(prodSlt)s;export EVTDUMPFILE=%(EVTDUMPFILE)s;export PRODSOFT=%(PRODSOFT)s;xrdcp -N -f root://castorpublic.cern.ch/%(input_file)s\?svcClass=compasscdr .;coralpath=%(ProdPathAndName)s/coral;cd -P $coralpath;export coralpathsetup=$coralpath"/setup.sh";source $coralpathsetup;cd $ppwd;$CORAL/src/DaqDataDecoding/examples/how-to/ddd --filter-CAL --out=testevtdump.raw %(TMPRAWFILE)s;if [ ! -s testevtdump.raw ]; then echo "PanDA message: the file is empty">testevtdump.raw; fi;cp payload_stderr.txt payload_stderr.out;cp payload_stdout.txt payload_stdout.out;gzip payload_stderr.out;gzip payload_stdout.out;rm %(tail)s' % {
'TMPRAWFILE': TMPRAWFILE,
'TMPMDSTFILE': TMPMDSTFILE,
'TMPHISTFILE': TMPHISTFILE,
'TMPRICHFILE': TMPRICHFILE,
'PRODSOFT': PRODSOFT,
'input_file': j.file,
'ProdPathAndName': ProdPathAndName,
'prodPath': j.task.path,
'prodName': j.task.production,
'template': j.task.template,
'tail': tail,
'prodSlt': j.task.prodslt,
'EVTDUMPFILE': EVTDUMPFILE,
'STDOUTFILE': STDOUTFILE,
'STDERRFILE': STDERRFILE
}
# fileIRaw = FileSpec()
# fileIRaw.lfn = "%s" % (input_file)
# fileIRaw.destinationDBlock = job.destinationDBlock
# fileIRaw.destinationSE = job.destinationSE
# fileIRaw.dataset = job.destinationDBlock
# fileIRaw.type = 'input'
# job.addFile(fileIRaw)
fileOstdout = FileSpec()
fileOstdout.lfn = "payload_stdout.out.gz"
fileOstdout.destinationDBlock = job.destinationDBlock
fileOstdout.destinationSE = job.destinationSE
fileOstdout.dataset = job.destinationDBlock
fileOstdout.type = 'output'
job.addFile(fileOstdout)
fileOstderr = FileSpec()
fileOstderr.lfn = "payload_stderr.out.gz"
fileOstderr.destinationDBlock = job.destinationDBlock
fileOstderr.destinationSE = job.destinationSE
fileOstderr.dataset = job.destinationDBlock
fileOstderr.type = 'output'
job.addFile(fileOstderr)
fileOLog = FileSpec()
fileOLog.lfn = "%(prodName)s-%(runNumber)s-%(runChunk)s-%(prodSlt)s-%(phastVer)s.job.log.tgz" % {
'prodName': j.task.production,
'runNumber': j.run_number,
'runChunk': j.chunk_number,
'prodSlt': j.task.prodslt,
'phastVer': j.task.phastver
}
fileOLog.destinationDBlock = job.destinationDBlock
fileOLog.destinationSE = job.destinationSE
fileOLog.dataset = job.destinationDBlock
fileOLog.type = 'log'
job.addFile(fileOLog)
if j.task.type == 'test production' or j.task.type == 'mass production' or j.task.type == 'technical production':
fileOmDST = FileSpec()
fileOmDST.lfn = "%s" % (TMPMDSTFILE)
fileOmDST.destinationDBlock = job.destinationDBlock
fileOmDST.destinationSE = job.destinationSE
fileOmDST.dataset = job.destinationDBlock
fileOmDST.type = 'output'
job.addFile(fileOmDST)
fileOTrafdic = FileSpec()
fileOTrafdic.lfn = "%s" % (TMPHISTFILE)
fileOTrafdic.destinationDBlock = job.destinationDBlock
fileOTrafdic.destinationSE = job.destinationSE
fileOTrafdic.dataset = job.destinationDBlock
fileOTrafdic.type = 'output'
job.addFile(fileOTrafdic)
if j.task.type == 'test production' or j.task.type == 'mass production' or j.task.type == 'technical production' or j.task.type == 'DDD filtering':
fileOtestevtdump = FileSpec()
fileOtestevtdump.lfn = "testevtdump.raw"
fileOtestevtdump.destinationDBlock = job.destinationDBlock
fileOtestevtdump.destinationSE = job.destinationSE
fileOtestevtdump.dataset = job.destinationDBlock
fileOtestevtdump.type = 'output'
job.addFile(fileOtestevtdump)
s, o = Client.submitJobs([job], srvID=aSrvID)
logger.info(s)
for x in o:
logger.info("PandaID=%s" % x[0])
if x[0] != 0 and x[0] != 'NULL':
j_update = Job.objects.get(id=j.id)
j_update.panda_id = x[0]
j_update.status = 'sent'
j_update.attempt = j_update.attempt + 1
j_update.date_updated = timezone.now()
try:
j_update.save()
logger.info('Job %s with PandaID %s updated at %s' %
(j.id, x[0], timezone.now()))
if j_update.task.status == 'send':
logger.info(
'Going to update status of task %s from send to running'
% j_update.task.name)
t_update = Task.objects.get(id=j_update.task.id)
t_update.status = 'running'
t_update.date_updated = timezone.now()
try:
t_update.save()
logger.info('Task %s updated' % t_update.name)
except IntegrityError as e:
logger.exception(
'Unique together catched, was not saved')
except DatabaseError as e:
logger.exception(
'Something went wrong while saving: %s' %
e.message)
except IntegrityError as e:
logger.exception(
'Unique together catched, was not saved')
except DatabaseError as e:
logger.exception(
'Something went wrong while saving: %s' %
e.message)
else:
logger.info('Job %s was not added to PanDA' % j.id)
i += 1
logger.info('done')
for i in range(1):
job = JobSpec()
job.jobDefinitionID = int(time.time()) % 10000
job.jobName = "%s_%d" % (commands.getoutput('uuidgen'),i)
# job.AtlasRelease = 'Atlas-12.0.6'
# job.homepackage = 'AtlasProduction/12.0.6.5'
job.AtlasRelease = 'Atlas-12.0.7'
job.homepackage = 'AtlasProduction/12.0.7.1'
job.transformation = 'csc_evgen_trf.py'
job.destinationDBlock = datasetName
# job.destinationSE = destName
# job.cloud = 'CA'
job.cloud = cloud
job.taskID = taskid
job.currentPriority = 1000
job.prodSourceLabel = 'test'
# job.prodSourceLabel = 'cloudtest'
job.computingSite = site
file = FileSpec()
file.lfn = "%s.evgen.pool.root" % job.jobName
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
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
