def addDatasetsToContainer(container, datasets):
from pandatools import Client
# HC datasets don't use containers
if not configPanda['processingType'].startswith(
'gangarobot') and not configPanda['processingType'].startswith(
'hammercloud'):
Client.addDatasetsToContainer(container, datasets, False)
def addDatasetsToContainer(container, datasets):
from pandatools import Client
# HC datasets don't use containers
if not configPanda["processingType"].startswith("gangarobot") and not configPanda["processingType"].startswith(
"hammercloud"
):
Client.addDatasetsToContainer(container, datasets, False)
def master_resubmit(self,jobs):
'''Resubmit failed Jedi job'''
from pandatools import Client
jobIDs = {}
for job in jobs:
jobIDs[job.backend.id] = job
allJobIDs = jobIDs.keys()
pandaJobIDs = {}
for jID in allJobIDs:
with inject_proxy(self.credential_requirements):
status, jediTaskDict = Client.getJediTaskDetails({'jediTaskID': jID},False,True,verbose=False)
if status != 0:
logger.error("Failed to get task details for %s" % jID)
raise BackendError('Jedi','Return code %d retrieving job status information.' % status)
# Retrieve job
job = jobIDs[jediTaskDict['jediTaskID']]
newJobsetID = -1 # get jobset
retryJobs = [] # jspecs
resubmittedJobs = [] # ganga jobs
if jediTaskDict['status'] in ['failed', 'killed', 'cancelled', 'aborted', 'broken', 'finished' ]:
retryJobs.append(job)
resubmittedJobs.append(jID)
#elif jediTaskDict['status'] == 'finished':
# pass
else:
logger.warning("Cannot resubmit. Jedi task %s is status %s." %(jID, jediTaskDict['status'] ))
return False
# submit
if len(retryJobs)==0:
logger.warning("No failed jobs to resubmit")
return False
with inject_proxy(self.credential_requirements):
status,out = Client.retryTask(jID, verbose=False)
if status != 0:
logger.error(status)
logger.error(out)
logger.error("Failed to retry JobID=%s" % jID)
return False
tmpStat,tmpDiag = out
if not tmpStat:
logger.error(tmpDiag)
logger.error("Failed to retry JobID=%s" % jID)
return False
logger.info(tmpDiag)
job.backend.status = None
job.backend.jobSpec = {}
job.updateStatus('submitted')
logger.info('Resubmission successful')
return True
def createContainer(name):
from pandatools import Client
# don't create containers for HC datasets
if not configPanda['processingType'].startswith('gangarobot') and not configPanda['processingType'].startswith('hammercloud'):
try:
Client.createContainer(name,False)
logger.info('Created output container %s' %name)
except exceptions.SystemExit:
raise BackendError('Panda','Exception in Client.createContainer %s: %s %s'%(name,sys.exc_info()[0],sys.exc_info()[1]))
def kill_processing(self, processing):
try:
if processing:
from pandatools import Client
task_id = processing.workload_id
Client.killTask(task_id)
except Exception as ex:
msg = "Failed to check the processing (%s) status: %s" % (str(
processing['processing_id']), str(ex))
raise exceptions.IDDSException(msg)
def createContainer(name):
from pandatools import Client
# don't create containers for HC datasets
if not configPanda['processingType'].startswith(
'gangarobot') and not configPanda['processingType'].startswith(
'hammercloud'):
try:
Client.createContainer(name, False)
logger.info('Created output container %s' % name)
except exceptions.SystemExit:
raise BackendError(
'Panda', 'Exception in Client.createContainer %s: %s %s' %
(name, sys.exc_info()[0], sys.exc_info()[1]))
def kill_processing_force(self, processing):
try:
if processing:
from pandatools import Client
proc = processing['processing_metadata']['processing']
task_id = proc.workload_id
# task_id = processing['processing_metadata']['task_id']
Client.killTask(task_id)
# Client.finishTask(task_id, soft=True)
except Exception as ex:
msg = "Failed to check the processing (%s) status: %s" % (str(
processing['processing_id']), str(ex))
raise exceptions.IDDSException(msg)
def createContainer(name):
from pandatools import Client
# don't create containers for HC datasets
if not configPanda["processingType"].startswith("gangarobot") and not configPanda["processingType"].startswith(
"hammercloud"
):
try:
Client.createContainer(name, False)
logger.info("Created output container %s" % name)
except exceptions.SystemExit:
raise BackendError(
"Panda", "Exception in Client.createContainer %s: %s %s" % (name, sys.exc_info()[0], sys.exc_info()[1])
)
def ELG_jediState(sample):
try:
from pandatools import PandaToolsPkgInfo # noqa: F401
except ImportError:
print("prun needs additional setup, try:")
print(" lsetup panda")
return 99
jediTaskID = int(sample.getMetaDouble("nc_jediTaskID", 0))
if jediTaskID < 100:
print("Sample " + sample.name() + " does not have a jediTaskID")
return ''
from pandatools import Client
taskDict = {}
taskDict['jediTaskID'] = jediTaskID
ret = Client.getJediTaskDetails(taskDict, False, True)
if ret[0] != 0:
print("Problem checking status of task %s with id %s" %
(sample.name(), jediTaskID))
return ''
return ret[1]['status'].encode("ascii")
def master_kill(self):
'''Kill jobs'''
from pandatools import Client
job = self.getJobObject()
logger.debug('Killing job %s' % job.getFQID('.'))
active_status = [
None, 'registered', 'waiting', 'defined', 'pending', 'assigning',
'ready', 'scouting', 'running', 'holding', 'merging', 'prepared',
'aborting', 'finishing'
]
#active_status = [ None, 'defined', 'unknown', 'assigned', 'waiting', 'activated', 'sent', 'starting', 'running', 'holding', 'transferring' ]
if self.id and self.status in active_status:
status, output = Client.killTask(self.id)
if status:
logger.error('Failed killing job (status = %d)', status)
return False
else:
logger.info('Killing Jedi task %s, Server returned: %s' %
(self.id, output))
else:
logger.error(
'Cannot kill Jedi job %s since it is not in active status',
self.id)
return True
def map_panda_ids(self, unregistered_job_ids, input_output_maps):
self.logger.debug("map_panda_ids, unregistered_job_ids: %s" %
str(unregistered_job_ids))
from pandatools import Client
# updated_map_ids = []
full_update_contents = []
chunksize = 2000
chunks = [
unregistered_job_ids[i:i + chunksize]
for i in range(0, len(unregistered_job_ids), chunksize)
]
for chunk in chunks:
jobs_list = Client.getJobStatus(chunk, verbose=0)[1]
for job_info in jobs_list:
if job_info and job_info.Files and len(job_info.Files) > 0:
for job_file in job_info.Files:
# if job_file.type in ['log']:
if job_file.type not in ['pseudo_input']:
continue
if ':' in job_file.lfn:
pos = job_file.lfn.find(":")
input_file = job_file.lfn[pos + 1:]
# input_file = job_file.lfn.split(':')[1]
else:
input_file = job_file.lfn
map_id = self.get_map_id_from_input(
input_output_maps, input_file)
if map_id:
update_contents = self.get_update_contents_from_map_id(
map_id, input_output_maps, job_info)
full_update_contents += update_contents
return full_update_contents
def getDBDatasets(jobO,trf,dbrelease):
from pandatools import Client
# get DB datasets
dbrFiles = {}
dbrDsList = []
if trf or dbrelease != '':
if trf:
# parse jobO for TRF
tmpItems = jobO.split()
else:
# mimic a trf parameter to reuse following algorithm
tmpItems = ['%DB='+dbrelease]
# look for DBRelease
for tmpItem in tmpItems:
match = re.search('%DB=([^:]+):(.+)$',tmpItem)
if match:
tmpDbrDS = match.group(1)
tmpDbrLFN = match.group(2)
# get files in the dataset
if not tmpDbrDS in dbrDsList:
logger.info("Querying files in dataset:%s" % tmpDbrDS)
try:
tmpList = Client.queryFilesInDataset(tmpDbrDS,False)
except:
raise ApplicationConfigurationError("ERROR : error while looking up dataset %s. Perhaps this dataset does not exist?"%tmpDbrDS)
# append
for tmpLFN,tmpVal in tmpList.iteritems():
dbrFiles[tmpLFN] = tmpVal
dbrDsList.append(tmpDbrDS)
# check
if tmpDbrLFN not in dbrFiles:
raise ApplicationConfigurationError("ERROR : %s is not in %s"%(tmpDbrLFN,tmpDbrDS))
return dbrFiles,dbrDsList
def get_panda_task_id(self, processing):
from pandatools import Client
start_time = datetime.datetime.utcnow() - datetime.timedelta(hours=10)
start_time = start_time.strftime('%Y-%m-%d %H:%M:%S')
status, results = Client.getJobIDsJediTasksInTimeRange(
start_time, task_type=self.task_type, verbose=False)
if status != 0:
self.logger.warn(
"Error to poll latest tasks in last ten hours: %s, %s" %
(status, results))
return None
proc = processing['processing_metadata']['processing']
task_id = None
for req_id in results:
task_name = results[req_id]['taskName']
if proc.workload_id is None and task_name == self.task_name:
task_id = results[req_id]['jediTaskID']
# processing['processing_metadata']['task_id'] = task_id
# processing['processing_metadata']['workload_id'] = task_id
proc.workload_id = task_id
if task_id:
proc.submitted_at = datetime.datetime.utcnow()
return task_id
def getLatestDBReleaseCaching():
import tempfile
import cPickle as pickle
from pandatools import Client
from GangaAtlas.Lib.Credentials.ProxyHelper import getNickname
TMPDIR = tempfile.gettempdir()
nickname = getNickname(allowMissingNickname=False)
DBRELCACHE = '%s/ganga.latestdbrel.%s'%(TMPDIR,nickname)
try:
fh = open(DBRELCACHE)
dbrelCache = pickle.load(fh)
fh.close()
if dbrelCache['mtime'] > time.time() - 3600:
logger.debug('Loading LATEST DBRelease from local cache')
return dbrelCache['atlas_dbrelease']
else:
raise Exception()
except:
logger.debug('Updating local LATEST DBRelease cache')
atlas_dbrelease = Client.getLatestDBRelease(False)
dbrelCache = {}
dbrelCache['mtime'] = time.time()
dbrelCache['atlas_dbrelease'] = atlas_dbrelease
fh = open(DBRELCACHE,'w')
pickle.dump(dbrelCache,fh)
fh.close()
return atlas_dbrelease
def getLatestDBReleaseCaching():
import tempfile
import cPickle as pickle
from pandatools import Client
from GangaAtlas.Lib.Credentials.ProxyHelper import getNickname
TMPDIR = tempfile.gettempdir()
nickname = getNickname(allowMissingNickname=False)
DBRELCACHE = '%s/ganga.latestdbrel.%s' % (TMPDIR, nickname)
try:
fh = open(DBRELCACHE)
dbrelCache = pickle.load(fh)
fh.close()
if dbrelCache['mtime'] > time.time() - 3600:
logger.debug('Loading LATEST DBRelease from local cache')
return dbrelCache['atlas_dbrelease']
else:
raise Exception()
except:
logger.debug('Updating local LATEST DBRelease cache')
atlas_dbrelease = Client.getLatestDBRelease(False)
dbrelCache = {}
dbrelCache['mtime'] = time.time()
dbrelCache['atlas_dbrelease'] = atlas_dbrelease
fh = open(DBRELCACHE, 'w')
pickle.dump(dbrelCache, fh)
fh.close()
return atlas_dbrelease
def status(self, JobID, forceUpdate=False):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase, self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase, False, self.verbose)
# get job info from local repository
job = self.getJobInfo(JobID)
if job == None:
# not found
return None
# update if needed
if job.dbStatus != 'frozen' or forceUpdate:
tmpLog.info("Getting status for JobID=%s ..." % JobID)
# get status from Panda server
status, pandaIDstatus = Client.getPandIDsWithJobID(
JobID, verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get status for JobID=%s" % JobID)
return None
# get one job to set computingSite which may have changed due to rebrokerage
pandaJob = None
if pandaIDstatus != {}:
tmpPandaIDs = pandaIDstatus.keys()
tmpPandaIDs.sort()
status, tmpPandaJobs = Client.getFullJobStatus(
tmpPandaIDs[:1], verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get PandaJobs for %s" % JobID)
return None
pandaJob = tmpPandaJobs[0]
# convert to local job spec
job = PdbUtils.convertPtoD([],
pandaIDstatus,
job,
pandaJobForSiteID=pandaJob)
# update DB
try:
PdbUtils.updateJobDB(job, self.verbose)
except:
tmpLog.error("Failed to update local repository for JobID=%s" %
JobID)
return None
tmpLog.info("Updated JobID=%s" % JobID)
# return
return job
def convertQueueNameToDQ2Names(self, queue, test):
from pandatools import Client
from dq2.info.TiersOfATLAS import ToACache
sites = []
if 'PFT' in test.template.description:
#Client.PandaSites = Client.getSiteSpecs('production')[1]
Client.PandaSites = Client.getSiteSpecs('all')[1]
for site in Client.PandaSites[queue]['setokens'].values():
sites.append(Client.convSrmV2ID(site))
allowed_sites = []
for site in ToACache.sites:
if (site not in allowed_sites
and Client.convSrmV2ID(site) in sites
and site.find('TAPE') == -1 and 'DISK' in site):
allowed_sites.append(site)
return allowed_sites
def retrievePandaJobs(job, jIDs):
'''
methods for retrieving panda job ids of panda jobs given a jobDefId
'''
from pandatools import Client
ick = False
jstatus = ''
num_pjobs = 0
logger.debug("retrievePandaJobs jIDs=%s" % jIDs)
# get status from Panda server
rc, jobsStatus = Client.getFullJobStatus(jIDs, False)
if rc:
logger.error('Return code %d retrieving job status information.', rc)
raise BackendError(
'Jedi', 'Return code %d retrieving job status information.' % rc)
for status in jobsStatus:
if not status: continue
jstatus = status.jobStatus
if status.jobStatus == None:
logger.warning('No panda jobs expected')
job.backend.pandajobs = []
elif status.jobStatus in [
"defined", "activated", "running", "failed", "finished",
"holding", "assigned"
]:
logger.debug('Panda jobs are running')
logger.debug("PandaID: %d" % status.PandaID)
pjobj = JediPandaJob()
pjobj.id = status.PandaID
pjobj.url = 'http://panda.cern.ch/?job=%d' % status.PandaID
pjobj.jopSpec = dict(zip(status._attributes, status.values()))
for k in pjobj.jobSpec.keys():
if type(pjobj.jobSpec[k]) not in [type(''), type(1)]:
pjobj.jobSpec[k] = str(pjobj.jobSpec[k])
if pjobj not in job.backend.pandajobs:
job.backend.pandajobs.append(pjobj)
else:
logger.debug("Panda job %s already exists locally" % pjobj.id)
num_pjobs += 1
else:
logger.warning(
"getFullJobStatus returned unsupported status %s for Panda job %s "
% (status.jobStatus, status.PandaID))
ick = True
return (ick, jstatus, num_pjobs)
def poll_panda_task(self, processing):
if 'panda_task_id' in processing['processing_metadata']:
from pandatools import Client
status, task_status = Client.getTaskStatus(processing.workload_id)
if status == 0:
return task_status
else:
return 'failed'
return None
def master_submit(self, rjobs, subjobspecs, buildjobspec):
'''Submit jobs'''
from pandatools import Client
from pandatools import MiscUtils
from Ganga.Core.exceptions import IncompleteJobSubmissionError
from Ganga.Utility.logging import log_user_exception
job = self.getJobObject()
# job name
jobName = 'ganga.%s' % MiscUtils.wrappedUuidGen()
jobspecs = {}
if buildjobspec:
jobspecs = buildjobspec
else:
jobspecs = subjobspecs
logger.debug(jobspecs)
# submit task
for subjob in rjobs:
subjob.updateStatus('submitting')
logger.info("Submitting to Jedi ...")
verbose = logger.isEnabledFor(10)
status, tmpOut = Client.insertTaskParams(jobspecs, verbose)
logger.debug(tmpOut)
if status != 0:
logger.error("Task submission to Jedi failed with %s " % status)
return False
if tmpOut[0] == False:
logger.error("Task submission to Jedi failed %s" % tmpOut[1])
return False
logger.info("Task submission to Jedi suceeded with new jediTaskID=%s" %
tmpOut[1])
#if buildjobspec:
# job.backend.buildjob = PandaBuildJob()
# job.backend.buildjob.id = jobids[0][0]
# job.backend.buildjob.url = 'http://panda.cern.ch/?job=%d'%jobids[0][0]
# del jobids[0]
for subjob in rjobs:
subjob.backend.id = tmpOut[1]
subjob.backend.url = 'http://pandamon.cern.ch/jedi/taskinfo?days=20&task=%d' % tmpOut[
1]
subjob.updateStatus('submitted')
logger.info("Panda monitor url: %s" % subjob.backend.url)
return True
def master_submit(self,rjobs,subjobspecs,buildjobspec):
'''Submit jobs'''
from pandatools import Client
from pandatools import MiscUtils
from Ganga.Core import IncompleteJobSubmissionError
from Ganga.Utility.logging import log_user_exception
job = self.getJobObject()
# job name
jobName = 'ganga.%s' % MiscUtils.wrappedUuidGen()
jobspecs = {}
if buildjobspec:
jobspecs = buildjobspec
else:
jobspecs = subjobspecs
logger.debug(jobspecs)
# submit task
for subjob in rjobs:
subjob.updateStatus('submitting')
logger.info("Submitting to Jedi ...")
verbose = logger.isEnabledFor(10)
with inject_proxy(self.credential_requirements):
status, tmpOut = Client.insertTaskParams(jobspecs, verbose)
logger.debug(tmpOut)
if status != 0:
logger.error("Task submission to Jedi failed with %s " %status)
return False
if tmpOut[0] == False:
logger.error("Task submission to Jedi failed %s" %tmpOut[1])
return False
logger.info("Task submission to Jedi suceeded with new jediTaskID=%s" %tmpOut[1])
#if buildjobspec:
# job.backend.buildjob = PandaBuildJob()
# job.backend.buildjob.id = jobids[0][0]
# job.backend.buildjob.url = 'http://panda.cern.ch/?job=%d'%jobids[0][0]
# del jobids[0]
for subjob in rjobs:
subjob.backend.id = tmpOut[1]
subjob.backend.url = 'http://pandamon.cern.ch/jedi/taskinfo?days=20&task=%d'%tmpOut[1]
subjob.updateStatus('submitted')
logger.info("Panda monitor url: %s" %subjob.backend.url)
return True
def getJobIDs(self, timestamp):
date = time.strftime('%Y-%m-%d %H:%M:%S', time.gmtime(timestamp))
status, output = Client.getJobIDsInTimeRange(date)
if status != 0:
print 'Error: panda !'
return []
output.sort()
return output
def finish(self,JobID,soft=False):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# force update just in case
self.status(JobID,True)
# get jobset
jobList = self.getJobIDsWithSetID(JobID)
if jobList == None:
# works only for jobsetID
if useJobsetID:
return
# works with jobID
jobList = [JobID]
else:
tmpMsg = "ID=%s is composed of JobID=" % JobID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for tmpJobID in jobList:
# get job info from local repository
job = self.getJobInfo(tmpJobID)
if job == None:
tmpLog.warning("JobID=%s not found in local repository. Synchronization may be needed" % tmpJobID)
continue
# skip frozen job
if job.dbStatus == 'frozen':
tmpLog.info('All subJobs in JobID=%s already finished/failed' % tmpJobID)
continue
# finish JEDI task
tmpLog.info('Sending finishTask command ...')
status,output = Client.finishTask(job.jediTaskID,soft,self.verbose)
# communication error
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to finish JobID=%s" % tmpJobID)
return False
tmpStat,tmpDiag = output
if not tmpStat:
tmpLog.error(tmpDiag)
tmpLog.error("Failed to finish JobID=%s" % tmpJobID)
return False
tmpLog.info(tmpDiag)
# done
tmpLog.info('Done. TaskID=%s will be finished soon' % job.jediTaskID)
return True
def kill(self, JobID):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase, self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase, False, self.verbose)
# get jobset
jobList = self.getJobIDsWithSetID(JobID)
if jobList == None:
jobList = [JobID]
else:
tmpMsg = "JobsetID=%s is composed of JobID=" % JobID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for tmpJobID in jobList:
# get job info from local repository
job = self.getJobInfo(tmpJobID)
if job == None:
tmpLog.warning(
"JobID=%s not found in local repository. Synchronization may be needed"
% tmpJobID)
continue
# skip frozen job
if job.dbStatus == 'frozen':
tmpLog.info('All subJobs in JobID=%s already finished/failed' %
tmpJobID)
continue
# get PandaID list
killJobs = job.PandaID.split(',')
# kill
tmpLog.info('Sending kill command ...')
status, output = Client.killJobs(killJobs, self.verbose)
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to kill JobID=%s" % tmpJobID)
return False
# update database
job.commandToPilot = 'tobekilled'
# update DB
try:
PdbUtils.updateJobDB(job, self.verbose)
except:
tmpLog.error("Failed to update local repository for JobID=%s" %
tmpJobID)
return False
# done
tmpLog.info('Done. JobID=%s will be killed in 30min' % tmpJobID)
return True
def retrievePandaJobs(job, jIDs):
'''
methods for retrieving panda job ids of panda jobs given a jobDefId
'''
from pandatools import Client
ick = False
jstatus = ''
num_pjobs = 0
logger.debug("retrievePandaJobs jIDs=%s" %jIDs)
# get status from Panda server
rc, jobsStatus = Client.getFullJobStatus(jIDs,False)
if rc:
logger.error('Return code %d retrieving job status information.',rc)
raise BackendError('Jedi','Return code %d retrieving job status information.' % rc)
for status in jobsStatus:
if not status: continue
jstatus = status.jobStatus
if status.jobStatus == None:
logger.warning('No panda jobs expected')
job.backend.pandajobs = []
elif status.jobStatus in [ "defined", "activated", "running", "failed", "finished", "holding", "assigned"]:
logger.debug('Panda jobs are running')
logger.debug("PandaID: %d" % status.PandaID)
pjobj = JediPandaJob()
pjobj.id = status.PandaID
pjobj.url = 'http://panda.cern.ch/?job=%d' % status.PandaID
pjobj.jopSpec = dict(zip(status._attributes,status.values()))
for k in pjobj.jobSpec.keys():
if type(pjobj.jobSpec[k]) not in [type(''),type(1)]:
pjobj.jobSpec[k]=str(pjobj.jobSpec[k])
if pjobj not in job.backend.pandajobs:
job.backend.pandajobs.append(pjobj)
else:
logger.debug("Panda job %s already exists locally" % pjobj.id)
num_pjobs += 1
else:
logger.warning("getFullJobStatus returned unsupported status %s for Panda job %s " %(status.jobStatus, status.PandaID) )
ick = True
return (ick, jstatus, num_pjobs)
def jobIsFinished(self, ID): L = Client.getPandIDsWithJobID(ID)[1] if L is None: result = False else: result = True for item in L: if L[item][0] != 'finished': result = False pass return result
def jobIsCancelled(self, ID): L = Client.getPandIDsWithJobID(ID)[1] if L is None: result = False else: result = False for item in L: if L[item][0] == 'cancelled': result = True pass return result
def submit_panda_task(self, processing):
try:
from pandatools import Client
proc = processing['processing_metadata']['processing']
task_param = proc.processing_metadata['task_param']
return_code = Client.insertTaskParams(task_param, verbose=True)
if return_code[0] == 0:
return return_code[1][1]
else:
self.logger.warn("submit_panda_task, return_code: %s" %
str(return_code))
except Exception as ex:
self.logger.error(ex)
self.logger.error(traceback.format_exc())
# raise exceptions.AgentPluginError('%s: %s' % (str(ex), traceback.format_exc()))
return None
def debug(self,pandaID,modeOn):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# rebrokerage
status,output = Client.setDebugMode(pandaID,modeOn,self.verbose)
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to set debug mode for %s" % pandaID)
return
# done
tmpLog.info(output)
return
def change_site_status(self, site, new_status):
if new_status not in ('test', 'online'):
return False
self.add_log('Changing %s status to %s:' % (site, new_status))
if self.debug:
self.add_log('DEBUG mode')
return True
now = int(time.time())
if new_status == 'test':
try:
last_exclusion = int(Site.objects.filter(name=site)[0].getSiteOptions_for_site.filter(option_name='last_exclusion')[0].option_value)
except:
last_exclusion = 0
if now - last_exclusion < 21600:
self.add_log('%s was recently auto-excluded. Skipping...' % site)
return False
cmd = "curl -s -k --cert $X509_USER_PROXY 'https://panda.cern.ch:25943/server/controller/query?tpmes=setmanual&queue=%s&moduser=HammerCloud&comment=HC.Blacklist.set.manual'" % site
self.add_log('> ' + cmd)
if not self.debug:
o = commands.getoutput(cmd)
self.add_log(o)
cmd = "curl -s -k --cert $X509_USER_PROXY 'https://panda.cern.ch:25943/server/controller/query?tpmes=set%s&queue=%s&moduser=HammerCloud&comment=HC.Blacklist.set.%s'" % (new_status, site, new_status)
self.add_log('> ' + cmd)
if not self.debug:
o = commands.getoutput(cmd)
self.add_log(o)
if new_status == 'test':
try:
option = Site.objects.filter(name=site)[0].getSiteOptions_for_site.filter(option_name='last_exclusion')[0]
option.option_value = now
except:
option = SiteOption()
option.option_name = 'last_exclusion'
option.option_value = now
option.site = Site.objects.get(name=site)
option.save()
Client.PandaSites = Client.getSiteSpecs(SITETYPE)[1]
if not self.debug and Client.PandaSites[site]['status'] != new_status:
self.store_log('Error setting %s to %s' % (site, new_status), 'error')
return False
return True
def reactivate_processing(self, processing):
try:
if processing:
from pandatools import Client
# task_id = processing['processing_metadata']['task_id']
proc = processing['processing_metadata']['processing']
task_id = proc.workload_id
# Client.retryTask(task_id)
status, out = Client.retryTask(task_id, newParams={})
self.logger.warn(
"Retry processing(%s) with task id(%s): %s, %s" %
(processing['processing_id'], task_id, status, out))
# Client.reactivateTask(task_id)
# Client.resumeTask(task_id)
except Exception as ex:
msg = "Failed to check the processing (%s) status: %s" % (str(
processing['processing_id']), str(ex))
raise exceptions.IDDSException(msg)
def submit_panda_task(self, processing):
try:
from pandatools import Client
status, tmpOut = Client.insertTaskParams(self.panda_task_paramsmap,
False, True)
if status == 0:
tmp_status, tmp_output = tmpOut
m = re.search("jediTaskID=(\d+)", tmp_output) # noqa W605
task_id = int(m.group(1))
processing.workload_id = task_id
else:
self.add_errors(tmpOut)
raise Exception(tmpOut)
except Exception as ex:
self.logger.error(ex)
self.logger.error(traceback.format_exc())
raise exceptions.IDDSException('%s: %s' %
(str(ex), traceback.format_exc()))
def getUserJobMetadata(self, jobID, output_filename):
job = self.getJobInfo(jobID)
# get logger
tmpLog = PLogger.getPandaLogger()
if job is None:
tmpLog.error('cannot find a task with {0}. May need to sync first'.format(jobID))
return False
# get metadata
task_id = job.jediTaskID
tmpLog.info('getting metadata')
status, metadata = Client.getUserJobMetadata(task_id, verbose=self.verbose)
if status != 0:
tmpLog.error(metadata)
tmpLog.error("Failed to get metadata")
return False
with open(output_filename, 'w') as f:
json.dump(metadata, f)
tmpLog.info('dumped to {0}'.format(output_filename))
# return
return True
def setMergeJobStatus(self,job,forceUpdate=False):
# only whenmerge job generation is active
if not forceUpdate and not job.activeMergeGen():
return True
# get logger
tmpLog = PLogger.getPandaLogger()
# check status of merge job generation
status,genStauts = Client.checkMergeGenerationStatus(job.JobID,verbose=self.verbose)
if status != 0:
tmpLog.error(genStauts)
tmpLog.error("Failed to check status of merge job generation for JobID=%s" % job.JobID)
return False
# set status
job.mergeJobStatus = genStauts['status']
# set merge job IDs
if genStauts['mergeIDs'] != []:
job.mergeJobID = ''
for tmpID in genStauts['mergeIDs']:
job.mergeJobID += '%s,' % tmpID
job.mergeJobID = job.mergeJobID[:-1]
# return
return True
def rebrokerage(self,JobsetID,cloud):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# get jobset
jobList = self.getJobIDsWithSetID(JobsetID)
if jobList == None:
jobList = [JobsetID]
else:
tmpMsg = "JobsetID=%s is composed of JobID=" % JobsetID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for JobID in jobList:
# get job info using status
job = self.status(JobID)
if job == None:
# not found
continue
# skip frozen job
if job.dbStatus == 'frozen':
tmpLog.info('All subJobs in JobID=%s already finished/failed' % JobID)
continue
# rebrokerage
tmpLog.info('Sending rebrokerage request ...')
status,output = Client.runReBrokerage(JobID,job.libDS,cloud,self.verbose)
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to reassign JobID=%s" % JobID)
return
# done
tmpLog.info('Done for %s' % JobID)
return
def master_kill(self):
'''Kill jobs'''
from pandatools import Client
job = self.getJobObject()
logger.debug('Killing job %s' % job.getFQID('.'))
active_status = [ None, 'registered', 'waiting', 'defined', 'pending', 'assigning', 'ready', 'scouting', 'running', 'holding', 'merging', 'prepared', 'aborting', 'finishing' ]
#active_status = [ None, 'defined', 'unknown', 'assigned', 'waiting', 'activated', 'sent', 'starting', 'running', 'holding', 'transferring' ]
if self.id and self.status in active_status:
status, output = Client.killTask(self.id)
if status:
logger.error('Failed killing job (status = %d)',status)
return False
else:
logger.info('Killing Jedi task %s, Server returned: %s' %(self.id, output))
else:
logger.error('Cannot kill Jedi job %s since it is not in active status', self.id)
return True
def get_status_changed_contents(self, unterminated_job_ids,
input_output_maps, panda_id_to_map_ids):
self.logger.debug(
"get_status_changed_contents, unterminated_job_ids: %s" %
str(unterminated_job_ids))
from pandatools import Client
full_update_contents = []
chunksize = 2000
chunks = [
unterminated_job_ids[i:i + chunksize]
for i in range(0, len(unterminated_job_ids), chunksize)
]
for chunk in chunks:
jobs_list = Client.getJobStatus(chunk, verbose=0)[1]
for job_info in jobs_list:
panda_id = job_info.PandaID
map_id = panda_id_to_map_ids[panda_id]
update_contents = self.get_update_contents_from_map_id(
map_id, input_output_maps, job_info)
full_update_contents += update_contents
return full_update_contents
def getDBDatasets(jobO, trf, dbrelease):
from pandatools import Client
# get DB datasets
dbrFiles = {}
dbrDsList = []
if trf or dbrelease != '':
if trf:
# parse jobO for TRF
tmpItems = jobO.split()
else:
# mimic a trf parameter to reuse following algorithm
tmpItems = ['%DB=' + dbrelease]
# look for DBRelease
for tmpItem in tmpItems:
match = re.search('%DB=([^:]+):(.+)$', tmpItem)
if match:
tmpDbrDS = match.group(1)
tmpDbrLFN = match.group(2)
# get files in the dataset
if not tmpDbrDS in dbrDsList:
logger.info("Querying files in dataset:%s" % tmpDbrDS)
try:
tmpList = Client.queryFilesInDataset(tmpDbrDS, False)
except:
raise ApplicationConfigurationError(
None,
"ERROR : error while looking up dataset %s. Perhaps this dataset does not exist?"
% tmpDbrDS)
# append
for tmpLFN, tmpVal in tmpList.iteritems():
dbrFiles[tmpLFN] = tmpVal
dbrDsList.append(tmpDbrDS)
# check
if tmpDbrLFN not in dbrFiles:
raise ApplicationConfigurationError(
None,
"ERROR : %s is not in %s" % (tmpDbrLFN, tmpDbrDS))
return dbrFiles, dbrDsList
def rebrokerage(self, JobsetID, cloud):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase, self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase, False, self.verbose)
# get jobset
jobList = self.getJobIDsWithSetID(JobsetID)
if jobList == None:
jobList = [JobsetID]
else:
tmpMsg = "JobsetID=%s is composed of JobID=" % JobsetID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for JobID in jobList:
# get job info using status
job = self.status(JobID)
if job == None:
# not found
continue
# skip frozen job
if job.dbStatus == 'frozen':
tmpLog.info('All subJobs in JobID=%s already finished/failed' %
JobID)
continue
# rebrokerage
tmpLog.info('Sending rebrokerage request ...')
status, output = Client.runReBrokerage(JobID, job.libDS, cloud,
self.verbose)
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to reassign JobID=%s" % JobID)
return
# done
tmpLog.info('Done for %s' % JobID)
return
def ELG_jediState(sample) :
from pandatools import PandaToolsPkgInfo
if int(float(PandaToolsPkgInfo.release_version[2])) < 4 :
print "Need prun with JEDI support, try:"
print " localSetupPandaClient currentJedi --noAthenaCheck"
return ''
jediTaskID = int(sample.getMetaDouble("nc_jediTaskID", 0))
if jediTaskID < 100 :
print "Sample " + sample.name() + " does not have a jediTaskID"
return ''
from pandatools import Client
taskDict = {}
taskDict['jediTaskID'] = jediTaskID
ret = Client.getJediTaskDetails(taskDict, False, True)
if ret[0] != 0 :
print "Problem checking status of task %s with id %s" % (sample.name(), jediTaskID)
return ''
return ret[1]['status']
def expandExcludedSiteList( job ):
'''Expand a site list taking wildcards into account'''
# first, check if there's anything to be done
check_ddm = False
wildcard = False
excl_sites = []
for s in job.backend.requirements.excluded_sites:
if s.find("ANALY_") == -1:
check_ddm = True
if s.find("*") != -1:
wildcard = True
if s.find("ANALY_") != -1 and s.find("*") == -1:
excl_sites.append(s)
if not check_ddm and not wildcard:
return excl_sites
# we have either wildcards or DDM sites listed
# First, find the allowed sites for this job and ensure no duplicates anywhere
from pandatools import Client
logger.info("Excluding DDM and wildcarded sites from Jedi job. Please wait....")
orig_ddm_list = []
new_ddm_list = []
for s in job.inputdata.get_locations():
if not s in orig_ddm_list:
orig_ddm_list.append(s)
new_ddm_list.append(s)
orig_panda_list = []
for s in [Client.convertDQ2toPandaID(x) for x in new_ddm_list]:
for s2 in Client.PandaSites.keys():
if s2.find(s) != -1 and not s2 in orig_panda_list:
orig_panda_list.append(s2)
if check_ddm:
# remove any DDM sites that are referenced, including wildcards
for s in job.backend.requirements.excluded_sites:
if s in orig_ddm_list:
new_ddm_list.remove(s)
if s.find("*") != -1:
for s2 in orig_ddm_list:
if fnmatch.fnmatch(s2, s):
new_ddm_list.remove(s2)
# now recreate the panda list and see if any have been dropped
new_panda_list = []
for s in [Client.convertDQ2toPandaID(x) for x in new_ddm_list]:
for s2 in Client.PandaSites.keys():
if s2.find(s) != -1 and not s2 in new_panda_list:
new_panda_list.append(s)
for s in orig_panda_list:
if not s in new_panda_list and not s in excl_sites:
excl_sites.append(s)
if wildcard:
# find wilcarded ANALY_* sites and exclude any that match good sites
for s in job.backend.requirements.excluded_sites:
if s.find("*") != -1:
for s2 in orig_panda_list:
if fnmatch.fnmatch(s2, s) and not s2 in excl_sites:
excl_sites.append(s2)
return excl_sites
def get_pick_dataset(self, verbose=False):
job = self._getParent()
if job and job.inputdata and job.inputdata.pick_event_list.name != '' and len(job.inputdata.dataset) !=0 :
raise ApplicationConfigurationError('Cannot use event pick list and input dataset at the same time.')
#parametr check for event picking
if job and job.inputdata and job.inputdata.pick_event_list.name != '':
if job.inputdata.pick_data_type == '':
raise ApplicationConfigurationError('Event pick data type (pick_data_type) must be specified.')
# set X509_USER_PROXY
from pandatools import Client
if 'X509_USER_PROXY' not in os.environ or os.environ['X509_USER_PROXY'] == '':
os.environ['X509_USER_PROXY'] = Client._x509()
# setup eventLookup
from pandatools.eventLookupClient import eventLookupClient
elssiIF = eventLookupClient()
# open run/event txt
runEvtList = []
if os.path.exists( self.pick_event_list.name ):
logger.info("Event pick list file %s selected" % self.pick_event_list.name)
runevttxt = open(self.pick_event_list.name)
for line in runevttxt:
items = line.split()
if len(items) != 2:
continue
runNr,evtNr = items
runEvtList.append([runNr,evtNr])
# close
runevttxt.close()
else:
raise ApplicationConfigurationError('Could not read event pick list file %s.' %self.pick_event_list.name)
# convert self.pick_data_type to Athena stream ref
if self.pick_data_type == 'AOD':
streamRef = 'StreamAOD_ref'
elif self.pick_data_type == 'ESD':
streamRef = 'StreamESD_ref'
elif self.pick_data_type == 'RAW':
streamRef = 'StreamRAW_ref'
else:
errStr = 'invalid data type %s for event picking. ' % self.pick_data_type
errStr += ' Must be one of AOD,ESD,RAW'
raise ApplicationConfigurationError(errStr)
logger.info('Getting dataset names and LFNs from ELSSI for event picking')
# read
guids = []
guidRunEvtMap = {}
runEvtGuidMap = {}
# bulk lookup
nEventsPerLoop = 500
iEventsTotal = 0
while iEventsTotal < len(runEvtList):
tmpRunEvtList = runEvtList[iEventsTotal:iEventsTotal+nEventsPerLoop]
iEventsTotal += nEventsPerLoop
paramStr = 'Run, Evt: %s, Stream: %s, Dataset pattern: %s' % (tmpRunEvtList,self.pick_stream_name, self.pick_dataset_pattern)
logger.debug(paramStr)
logger.info('.')
# check with ELSSI
if self.pick_stream_name == '':
guidListELSSI = elssiIF.doLookup(tmpRunEvtList,tokens=streamRef,amitag=self.event_pick_amitag,extract=True)
else:
guidListELSSI = elssiIF.doLookup(tmpRunEvtList,stream=self.pick_stream_name,tokens=streamRef,amitag=self.event_pick_amitag,extract=True)
if len(guidListELSSI) == 0 or guidListELSSI == None:
errStr = ''
for tmpLine in elssiIF.output:
errStr += tmpLine + '\n'
errStr = "GUID was not found in ELSSI.\n" + errStr
raise ApplicationConfigurationError(errStr)
# check attribute
attrNames, attrVals = guidListELSSI
def getAttributeIndex(attr):
for tmpIdx,tmpAttrName in enumerate(attrNames):
if tmpAttrName.strip() == attr:
return tmpIdx
errStr = "cannot find attribute=%s in %s provided by ELSSI" % (attr,str(attrNames))
raise ApplicationConfigurationError(errStr)
# get index
indexEvt = getAttributeIndex('EventNumber')
indexRun = getAttributeIndex('RunNumber')
indexTag = getAttributeIndex(streamRef)
# check events
for runNr,evtNr in tmpRunEvtList:
paramStr = 'Run:%s Evt:%s Stream:%s' % (runNr,evtNr,self.pick_stream_name)
# collect GUIDs
tmpguids = []
for attrVal in attrVals:
if runNr == attrVal[indexRun] and evtNr == attrVal[indexEvt]:
tmpGuid = attrVal[indexTag]
# check non existing
if tmpGuid == 'NOATTRIB':
continue
if not tmpGuid in tmpguids:
tmpguids.append(tmpGuid)
# not found
if tmpguids == []:
errStr = "no GUIDs were found in ELSSI for %s" % paramStr
raise ApplicationConfigurationError(errStr)
# append
for tmpguid in tmpguids:
if not tmpguid in guids:
guids.append(tmpguid)
guidRunEvtMap[tmpguid] = []
guidRunEvtMap[tmpguid].append((runNr,evtNr))
runEvtGuidMap[(runNr,evtNr)] = tmpguids
# convert to dataset names and LFNs
dsLFNs,allDSMap = Client.listDatasetsByGUIDs(guids,self.pick_dataset_pattern,verbose)
logger.debug(dsLFNs)
#populate DQ2Datase
if job and job.inputdata:
job.inputdata.dataset = []
job.inputdata.names = []
job.inputdata.guids = []
# check duplication
for runNr,evtNr in runEvtGuidMap.keys():
tmpLFNs = []
tmpAllDSs = {}
for tmpguid in runEvtGuidMap[(runNr,evtNr)]:
if tmpguid in dsLFNs:
tmpLFNs.append(dsLFNs[tmpguid])
job.inputdata.guids.append(tmpguid)
job.inputdata.names.append(dsLFNs[tmpguid][1])
if not ((dsLFNs[tmpguid][0]) in job.inputdata.dataset):
job.inputdata.dataset.append(dsLFNs[tmpguid][0])
else:
tmpAllDSs[tmpguid] = allDSMap[tmpguid]
if tmpguid in guidRunEvtMap:
del guidRunEvtMap[tmpguid]
# empty
if tmpLFNs == []:
paramStr = 'Run:%s Evt:%s Stream:%s' % (runNr,evtNr,self.pick_stream_name)
errStr = "Dataset pattern '%s' didn't pick up a file for %s\n" % (self.pick_dataset_pattern,paramStr)
for tmpguid,tmpAllDS in tmpAllDSs.iteritems():
errStr += " GUID:%s dataset:%s\n" % (tmpguid,str(tmpAllDS))
raise ApplicationConfigurationError(errStr)
# duplicated
if len(tmpLFNs) != 1:
paramStr = 'Run:%s Evt:%s Stream:%s' % (runNr,evtNr,self.pick_stream_name)
errStr = "Multiple LFNs %s were found in ELSSI for %s. Please set pick_dataset_pattern and/or pick_stream_name and or event_pick_amitag correctly." %(str(tmpLFNs),paramStr)
raise ApplicationConfigurationError(errStr)
# return
return guidRunEvtMap
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 master_resubmit(self, jobs):
'''Resubmit failed Jedi job'''
from pandatools import Client
jobIDs = {}
for job in jobs:
jobIDs[job.backend.id] = job
allJobIDs = jobIDs.keys()
pandaJobIDs = {}
for jID in allJobIDs:
status, jediTaskDict = Client.getJediTaskDetails(
{'jediTaskID': jID}, False, True, verbose=False)
if status != 0:
logger.error("Failed to get task details for %s" % jID)
raise BackendError(
'Jedi',
'Return code %d retrieving job status information.' %
status)
# Retrieve job
job = jobIDs[jediTaskDict['jediTaskID']]
newJobsetID = -1 # get jobset
retryJobs = [] # jspecs
resubmittedJobs = [] # ganga jobs
if jediTaskDict['status'] in [
'failed', 'killed', 'cancelled', 'aborted', 'broken',
'finished'
]:
retryJobs.append(job)
resubmittedJobs.append(jID)
#elif jediTaskDict['status'] == 'finished':
# pass
else:
logger.warning("Cannot resubmit. Jedi task %s is status %s." %
(jID, jediTaskDict['status']))
return False
# submit
if len(retryJobs) == 0:
logger.warning("No failed jobs to resubmit")
return False
status, out = Client.retryTask(jID, verbose=False)
if status != 0:
logger.error(status)
logger.error(out)
logger.error("Failed to retry JobID=%s" % jID)
return False
tmpStat, tmpDiag = out
if not tmpStat:
logger.error(tmpDiag)
logger.error("Failed to retry JobID=%s" % jID)
return False
logger.info(tmpDiag)
job.backend.status = None
job.backend.jobSpec = {}
job.updateStatus('submitted')
logger.info('Resubmission successful')
return True
def master_prepare(self, app, appconfig):
'''Prepare the master job'''
from pandatools import Client
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('ExecutablePandaRTHandler master_prepare called for %s',
job.getFQID('.'))
# set chirp variables
if configPanda['chirpconfig'] or configPanda['chirpserver']:
setChirpVariables()
# Pack inputsandbox
inputsandbox = 'sources.%s.tar' % commands.getoutput(
'uuidgen 2> /dev/null')
inpw = job.getInputWorkspace()
# add user script to inputsandbox
if hasattr(job.application.exe, "name"):
if not job.application.exe in job.inputsandbox:
job.inputsandbox.append(job.application.exe)
for fname in [f.name for f in job.inputsandbox]:
fname.rstrip(os.sep)
path = fname[:fname.rfind(os.sep)]
f = fname[fname.rfind(os.sep) + 1:]
rc, output = commands.getstatusoutput(
'tar rf %s -C %s %s' % (inpw.getPath(inputsandbox), path, f))
if rc:
logger.error('Packing inputsandbox failed with status %d', rc)
logger.error(output)
raise ApplicationConfigurationError(
None, 'Packing inputsandbox failed.')
if len(job.inputsandbox) > 0:
rc, output = commands.getstatusoutput('gzip %s' %
(inpw.getPath(inputsandbox)))
if rc:
logger.error('Packing inputsandbox failed with status %d', rc)
logger.error(output)
raise ApplicationConfigurationError(
None, 'Packing inputsandbox failed.')
inputsandbox += ".gz"
else:
inputsandbox = None
# Upload Inputsandbox
if inputsandbox:
logger.debug('Uploading source tarball ...')
uploadSources(inpw.getPath(), os.path.basename(inputsandbox))
self.inputsandbox = inputsandbox
else:
self.inputsandbox = None
# input dataset
if job.inputdata:
if job.inputdata._name != 'DQ2Dataset':
raise ApplicationConfigurationError(
None, 'PANDA application supports only DQ2Datasets')
# run brokerage here if not splitting
if not job.splitter:
from GangaPanda.Lib.Panda.Panda import runPandaBrokerage
runPandaBrokerage(job)
elif job.splitter._name not in [
'DQ2JobSplitter', 'ArgSplitter', 'ArgSplitterTask'
]:
raise ApplicationConfigurationError(
None, 'Panda splitter must be DQ2JobSplitter or ArgSplitter')
if job.backend.site == 'AUTO':
raise ApplicationConfigurationError(
None, 'site is still AUTO after brokerage!')
# output dataset
if job.outputdata:
if job.outputdata._name != 'DQ2OutputDataset':
raise ApplicationConfigurationError(
None, 'Panda backend supports only DQ2OutputDataset')
else:
logger.info('Adding missing DQ2OutputDataset')
job.outputdata = DQ2OutputDataset()
job.outputdata.datasetname, outlfn = dq2outputdatasetname(
job.outputdata.datasetname, job.id, job.outputdata.isGroupDS,
job.outputdata.groupname)
self.outDsLocation = Client.PandaSites[job.backend.site]['ddm']
try:
Client.addDataset(job.outputdata.datasetname,
False,
location=self.outDsLocation)
logger.info('Output dataset %s registered at %s' %
(job.outputdata.datasetname, self.outDsLocation))
dq2_set_dataset_lifetime(job.outputdata.datasetname,
location=self.outDsLocation)
except exceptions.SystemExit:
raise BackendError(
'Panda', 'Exception in Client.addDataset %s: %s %s' %
(job.outputdata.datasetname, sys.exc_info()[0],
sys.exc_info()[1]))
# handle the libds
if job.backend.libds:
self.libDataset = job.backend.libds
self.fileBO = getLibFileSpecFromLibDS(self.libDataset)
self.library = self.fileBO.lfn
elif job.backend.bexec:
self.libDataset = job.outputdata.datasetname + '.lib'
self.library = '%s.tgz' % self.libDataset
try:
Client.addDataset(self.libDataset,
False,
location=self.outDsLocation)
dq2_set_dataset_lifetime(self.libDataset,
location=self.outDsLocation)
logger.info('Lib dataset %s registered at %s' %
(self.libDataset, self.outDsLocation))
except exceptions.SystemExit:
raise BackendError(
'Panda', 'Exception in Client.addDataset %s: %s %s' %
(self.libDataset, sys.exc_info()[0], sys.exc_info()[1]))
# collect extOutFiles
self.extOutFile = []
for tmpName in job.outputdata.outputdata:
if tmpName != '':
self.extOutFile.append(tmpName)
for tmpName in job.outputsandbox:
if tmpName != '':
self.extOutFile.append(tmpName)
for tmpName in job.backend.extOutFile:
if tmpName != '':
self.extOutFile.append(tmpName)
# create build job
if job.backend.bexec != '':
jspec = JobSpec()
jspec.jobDefinitionID = job.id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.transformation = '%s/buildGen-00-00-01' % Client.baseURLSUB
if Client.isDQ2free(job.backend.site):
jspec.destinationDBlock = '%s/%s' % (
job.outputdata.datasetname, self.libDataset)
jspec.destinationSE = 'local'
else:
jspec.destinationDBlock = self.libDataset
jspec.destinationSE = job.backend.site
jspec.prodSourceLabel = configPanda['prodSourceLabelBuild']
jspec.processingType = configPanda['processingType']
jspec.assignedPriority = configPanda['assignedPriorityBuild']
jspec.computingSite = job.backend.site
jspec.cloud = job.backend.requirements.cloud
jspec.jobParameters = '-o %s' % (self.library)
if self.inputsandbox:
jspec.jobParameters += ' -i %s' % (self.inputsandbox)
else:
raise ApplicationConfigurationError(
None,
'Executable on Panda with build job defined, but inputsandbox is emtpy !'
)
matchURL = re.search('(http.*://[^/]+)/', Client.baseURLCSRVSSL)
if matchURL:
jspec.jobParameters += ' --sourceURL %s ' % matchURL.group(1)
if job.backend.bexec != '':
jspec.jobParameters += ' --bexec "%s" ' % urllib.quote(
job.backend.bexec)
jspec.jobParameters += ' -r %s ' % '.'
fout = FileSpec()
fout.lfn = self.library
fout.type = 'output'
fout.dataset = self.libDataset
fout.destinationDBlock = self.libDataset
jspec.addFile(fout)
flog = FileSpec()
flog.lfn = '%s.log.tgz' % self.libDataset
flog.type = 'log'
flog.dataset = self.libDataset
flog.destinationDBlock = self.libDataset
jspec.addFile(flog)
return jspec
else:
return None
def status(self,JobID,forceUpdate=False):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# get job info from local repository
job = self.getJobInfo(JobID)
if job == None:
# not found
return None
# update if needed
if job.dbStatus != 'frozen' or forceUpdate:
if not job.isJEDI():
tmpLog.info("Getting status for JobID=%s ..." % JobID)
# get status from Panda server
status,pandaIDstatus = Client.getPandIDsWithJobID(JobID,verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get status for ID=%s" % JobID)
return None
# get one job to set computingSite which may have changed due to rebrokerage
pandaJob = None
if pandaIDstatus != {}:
tmpPandaIDs = pandaIDstatus.keys()
tmpPandaIDs.sort()
status,tmpPandaJobs = Client.getFullJobStatus(
tmpPandaIDs[:1],
verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get PandaJobs for %s" % JobID)
return None
pandaJob = tmpPandaJobs[0]
# convert to local job spec
job = PdbUtils.convertPtoD([],pandaIDstatus,job,pandaJobForSiteID=pandaJob)
# check merge job generation
status = self.setMergeJobStatus(job,forceUpdate)
if not status:
return None
else:
tmpLog.info("Getting status for TaskID=%s ..." % job.jediTaskID)
# get JEDI task
status,jediTaskDict = Client.getJediTaskDetails(
{'jediTaskID':job.jediTaskID},
False,
True,
verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get task details for %s" % JobID)
return
# convert JEDI task
job = PdbUtils.convertJTtoD(jediTaskDict,job)
# update DB
try:
PdbUtils.updateJobDB(job,self.verbose)
except:
tmpLog.error("Failed to update local repository for JobID=%s" % JobID)
return None
if not job.isJEDI():
tmpLog.info("Updated JobID=%s" % JobID)
else:
tmpLog.info("Updated TaskID=%s ..." % job.jediTaskID)
# return
return job
def printGUIDsWithDatasets(guids):
""" list datasets by GUIDs
"""
try:
from pandatools import Client
except ImportError:
if os.environ.has_key('PANDA_SYS'):
pandapath = os.environ['PANDA_SYS']
else:
pandapath = PandaSys
sys.path.append(pandapath + '/lib/python2.4/site-packages')
try:
from pandatools import Client
except ImportError:
print "EventLookup failed to import PanDA client, GUID->dataset name resolution disabled"
return False
# instantiate curl
curl = Client._Curl()
curl.verbose = False
iLookUp = 0
guidLfnMap = {}
checkedDSList = []
# loop over all GUIDs
for guid in guids.keys():
# check existing map to avid redundant lookup
if guidLfnMap.has_key(guid):
continue
iLookUp += 1
if iLookUp % 20 == 0:
time.sleep(1)
# get vuids
url = Client.baseURLDQ2 + '/ws_content/rpc'
data = {
'operation': 'queryDatasetsWithFileByGUID',
'guid': guid,
'API': '0_3_0',
'tuid': Client.MiscUtils.wrappedUuidGen()
}
status, out = curl.get(url, data, rucioAccount=True)
if status != 0 or re.search('Exception', str(out)) != None:
# failed
guidLfnMap[guid] = "DQ2 query ERROR: " + str(out)
continue
if out == '\x00' or out == ():
guidLfnMap[
guid] = "DQ2.queryDatasetsWithFileByGUID() returned no results"
continue
tmpVUIDs = list(out)
# get dataset name
url = Client.baseURLDQ2 + '/ws_repository/rpc'
data = {
'operation': 'queryDatasetByVUIDs',
'vuids': tmpVUIDs,
'API': '0_3_0',
'tuid': Client.MiscUtils.wrappedUuidGen()
}
status, out = curl.post(url, data, rucioAccount=True)
if status != 0 or re.search('Exception', str(out)) != None:
# failed
guidLfnMap[guid] = "DQ2 query ERROR: " + str(out)
continue
if out == '\x00':
guidLfnMap[guid] = "DQ2.queryDatasetByVUIDs() returned no results"
continue
for tmpDsName in out.keys():
# ignore junk datasets
if not (tmpDsName.startswith('panda') or \
tmpDsName.startswith('user') or \
tmpDsName.startswith('group') or \
re.search('_sub\d+$',tmpDsName) != None or \
re.search('_dis\d+$',tmpDsName) != None or \
re.search('_shadow$',tmpDsName) != None \
or tmpDsName in checkedDSList ):
tmpMap = Client.queryFilesInDataset(tmpDsName)
for tmpLFN, tmpVal in tmpMap.iteritems():
guidLfnMap.setdefault(tmpVal['guid'],
[]).append([tmpLFN, tmpDsName])
checkedDSList.append(tmpDsName)
for guid in guids.keys():
print guid, guids[guid], guidLfnMap.setdefault(guid, "")
return True
def retrieveMergeJobs(job, pandaJobDefId):
'''
methods for retrieving panda job ids of merging jobs given a jobDefId
'''
from pandatools import Client
ick = False
status = ''
num_mjobs = 0
(ec, info) = Client.checkMergeGenerationStatus(pandaJobDefId)
if ec == 0:
try:
status = info['status']
mergeJobDefIds = info['mergeIDs']
if status == 'NA':
logger.warning('No merging jobs expected')
job.backend.mergejobs = []
elif status == 'generating':
logger.debug('merging jobs are generating')
job.backend.mergejobs = []
elif status == 'standby':
logger.debug('merging jobs to be created')
job.backend.mergejobs = []
elif status == 'generated':
logger.debug('merging jobs are generated')
for id in mergeJobDefIds:
logger.debug("merging jobDefId: %d" % id)
## retrieve merging job id,status given the jobDefId
(ec2, mjs) = Client.getPandIDsWithJobID(id)
if ec2 == 0:
for jid, jinfo in mjs.items():
mjobj = PandaMergeJob()
mjobj.id = jid
#mjobj.status = jinfo[0]
mjobj.url = 'http://panda.cern.ch/?job=%d' % jid
if mjobj not in job.backend.mergejobs:
job.backend.mergejobs.append(mjobj)
else:
logger.debug(
"merging job %s already exists locally" %
mjobj.id)
num_mjobs += 1
else:
logger.warning(
"getPandIDsWithJobID returns non-zero exit code: %d"
% ec2)
ick = True
except KeyError:
logger.error('unexpected job information: %s' % repr(info))
except Exception as e:
logger.error('general merge job information retrieval error')
raise e
else:
logger.error(
'checkMergeGenerationStatus returns non-zero exit code: %d' % ec)
return (ick, status, num_mjobs)
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 poll_panda_task(self, processing=None, input_output_maps=None):
task_id = None
try:
from pandatools import Client
jobs_ids = None
if processing:
proc = processing['processing_metadata']['processing']
task_id = proc.workload_id
if task_id is None:
task_id = self.get_panda_task_id(processing)
if task_id:
# ret_ids = Client.getPandaIDsWithTaskID(task_id, verbose=False)
task_info = Client.getJediTaskDetails(
{'jediTaskID': task_id}, True, True, verbose=False)
self.logger.info("poll_panda_task, task_info: %s" %
str(task_info))
if task_info[0] != 0:
self.logger.warn(
"poll_panda_task %s, error getting task status, task_info: %s"
% (task_id, str(task_info)))
return ProcessingStatus.Submitting, {}
task_info = task_info[1]
processing_status = self.get_processing_status_from_panda_status(
task_info["status"])
if processing_status in [ProcessingStatus.SubFinished]:
if self.retry_number < self.num_retries:
self.reactivate_processing(processing)
processing_status = ProcessingStatus.Submitted
self.retry_number += 1
jobs_ids = task_info['PandaID']
ret_get_registered_panda_jobids = self.get_registered_panda_jobids(
input_output_maps)
terminated_job_ids, unterminated_job_ids, map_id_without_panda_ids, panda_id_to_map_ids = ret_get_registered_panda_jobids
registered_job_ids = terminated_job_ids + unterminated_job_ids
unregistered_job_ids = []
for job_id in jobs_ids:
if job_id not in registered_job_ids:
unregistered_job_ids.append(job_id)
map_update_contents = self.map_panda_ids(
unregistered_job_ids, input_output_maps)
status_changed_update_contents = self.get_status_changed_contents(
unterminated_job_ids, input_output_maps,
panda_id_to_map_ids)
final_update_contents = []
if processing_status in [
ProcessingStatus.SubFinished,
ProcessingStatus.Finished, ProcessingStatus.Failed
]:
if (unregistered_job_ids or unterminated_job_ids):
# there are still polling contents, should not terminate the task.
log_warn = "Processing (%s) with panda id (%s) is %s, however there are still unregistered_job_ids(%s) or unterminated_job_ids(%s)" % (
processing['processing_id'], task_id,
processing_status, str(unregistered_job_ids),
str(unterminated_job_ids))
log_warn = log_warn + ". Keep the processing status as running now."
self.logger.warn(log_warn)
processing_status = ProcessingStatus.Running
else:
final_update_contents = self.get_final_update_contents(
input_output_maps)
if final_update_contents:
processing_status = ProcessingStatus.Running
return processing_status, map_update_contents + status_changed_update_contents + final_update_contents
else:
return ProcessingStatus.Failed, {}
except Exception as ex:
msg = "Failed to check the processing (%s) status: %s" % (str(
processing['processing_id']), str(ex))
self.logger.error(msg)
self.logger.error(ex)
self.logger.error(traceback.format_exc())
# raise exceptions.IDDSException(msg)
return ProcessingStatus.Submitting, []
def sync(self):
# get logger
tmpLog = PLogger.getPandaLogger()
tmpLog.info("Synchronizing local repository ...")
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# get nickname
nickName = PsubUtils.getNickname()
# set Rucio accounting
PsubUtils.setRucioAccount(nickName,'pbook',True)
# get JobIDs in local repository
localJobIDs = PdbUtils.getListOfJobIDs()
# get recent JobIDs from panda server
syncTimeRaw = datetime.datetime.utcnow()
syncTime = syncTimeRaw.strftime('%Y-%m-%d %H:%M:%S')
# set sync time for the first attempt
bookConf = BookConfig.getConfig()
if self.restoreDB:
# reset last_synctime to restore database
bookConf.last_synctime = ''
# disable
self.restoreDB = False
tmpLog.info("It may take several minutes to restore local repository ...")
if bookConf.last_synctime == '':
bookConf.last_synctime = datetime.datetime.utcnow()-datetime.timedelta(days=180)
bookConf.last_synctime = bookConf.last_synctime.strftime('%Y-%m-%d %H:%M:%S')
maxTaskID = None
while True:
status, jediTaskDicts = Client.getJobIDsJediTasksInTimeRange(bookConf.last_synctime,
minTaskID=maxTaskID,
verbose=self.verbose)
if status != 0:
tmpLog.error("Failed to get tasks from panda server")
return
if len(jediTaskDicts) == 0:
break
tmpLog.info("Got %s tasks to be updated" % len(jediTaskDicts))
# insert if missing
for remoteJobID in jediTaskDicts.keys():
taskID = jediTaskDicts[remoteJobID]['jediTaskID']
# get max
if maxTaskID is None or taskID > maxTaskID:
maxTaskID = taskID
# check local status
job = None
if remoteJobID in localJobIDs:
# get job info from local repository
job = PdbUtils.readJobDB(remoteJobID, self.verbose)
# skip if frozen
if job.dbStatus == 'frozen':
continue
tmpLog.info("Updating taskID=%s ..." % taskID)
# convert JEDI task
localJob = PdbUtils.convertJTtoD(jediTaskDicts[remoteJobID],job)
# update database
if not remoteJobID in localJobIDs:
# insert to DB
try:
PdbUtils.insertJobDB(localJob,self.verbose)
except:
tmpLog.error("Failed to insert taskID=%s to local repository" % taskID)
return
else:
# update
try:
PdbUtils.updateJobDB(localJob,self.verbose,syncTimeRaw)
except:
tmpLog.error("Failed to update local repository for taskID=%s" % taskID)
return
# update sync time
bookConf = BookConfig.getConfig()
bookConf.last_synctime = syncTime
BookConfig.updateConfig(bookConf)
self.updateTaskJobsetMap()
tmpLog.info("Synchronization Completed")
def master_updateMonitoringInformation(jobs):
'''Monitor jobs'''
from pandatools import Client
#active_status = [ None, 'defined', 'unknown', 'assigned', 'waiting', 'activated', 'sent', 'starting', 'running', 'holding', 'transferring' ]
submitting_status = []
active_status = [
None, 'registered', 'waiting', 'defined', 'pending', 'assigning',
'ready', 'scouting', 'running', 'holding', 'merging', 'prepared',
'aborting', 'finishing'
]
inactive_status = ['finished', 'aborted', 'broken', 'failed', 'done']
# Find jobs to be monitored
jobdict = {}
for job in jobs:
# add a delay as Panda can be a little slow in sorting out a new Task
if job.backend.id and job.backend.status in active_status and (
(datetime.datetime.utcnow() -
job.time.timestamps["submitted"]).seconds > 120):
jobdict[job.backend.id] = job
logger.debug("jobdict = %s" % jobdict)
# Monitor active Jedi tasks
allJobIDs = jobdict.keys()
pandaJobIDs = {}
for jID in allJobIDs:
status, jediTaskDict = Client.getJediTaskDetails(
{'jediTaskID': jID}, False, True, verbose=False)
if status != 0:
logger.error("Failed to get task details for %s" % jID)
#raise BackendError('Jedi','Return code %d retrieving job status information.' % status)
continue
# Retrieve job
job = jobdict[jediTaskDict['jediTaskID']]
# Store associated Panda jobs
if job.backend.pandajobs:
pandaJobIDs[job.backend.id] = [
pj.id for pj in job.backend.pandajobs
]
else:
pandaJobIDs[
jediTaskDict['jediTaskID']] = jediTaskDict['PandaID']
logger.debug("jID = %s, pandaJobIDs = %s" % (jID, pandaJobIDs))
# Fill the output data dataset list
if 'outDS' in jediTaskDict and jediTaskDict['outDS'] != '':
for ds in jediTaskDict['outDS'].split(','):
if not ds in job.outputdata.datasetList:
job.outputdata.datasetList.append(ds)
# Jedi job status has changed
if job.backend.status != jediTaskDict['status']:
logger.debug('Job %s has changed status from %s to %s',
job.getFQID('.'), job.backend.status,
jediTaskDict['status'])
job.backend.status = jediTaskDict['status']
job.backend.reason = jediTaskDict['statistics']
# Now update Jedi job status
if jediTaskDict['status'] in [
'registered', 'waiting', 'defined', 'pending',
'assigning', 'ready'
]:
job.updateStatus('submitted')
elif jediTaskDict['status'] in [
'scouting', 'running', 'holding', 'merging', 'prepared'
]:
job.updateStatus('running')
elif jediTaskDict['status'] in ['done']:
job.updateStatus('completed')
elif jediTaskDict['status'] in ['failed', 'finished']:
job.updateStatus('failed')
elif jediTaskDict['status'] in [
'aborted', 'broken', 'cancelled'
] and job.status not in ['completed', 'failed']:
job.updateStatus('killed')
else:
logger.warning('Unexpected Jedi task status %s',
jediTaskDict['status'])
# Check if associated Panda job exist and monitor them
if not job.backend.pandajobs:
jdefids = pandaJobIDs[jID]
# skip if there are no Panda jobs yet
if not jdefids:
continue
tot_num_mjobs = 0
do_master_update = True
ick, status, num_mjobs = retrievePandaJobs(job, jdefids)
logger.debug('retrievePandaJobs returns: %s %s' %
(repr(ick), status))
if not ick:
logger.debug(
'Panda job retrival failure for Jedi task %s with PandaIds %s'
% (job.backend.id, jdefids))
do_master_update = False
tot_num_mjobs += num_mjobs
logger.debug('Job %s retrieved %d Panda jobs' %
(job.getFQID('.'), tot_num_mjobs))
# Now monitor the already attached Panda jobs
else:
jdefids = [pj.id for pj in job.backend.pandajobs]
rc, jobsStatus = Client.getFullJobStatus(jdefids, False)
if rc:
logger.error(
'Return code %d retrieving job status information.',
rc)
raise BackendError(
'Jedi',
'Return code %d retrieving job status information.' %
rc)
for status in jobsStatus:
if not status: continue
for pjob in job.backend.pandajobs:
if pjob.id == status.PandaID:
# skip if no status change
if pjob.status == status.jobStatus:
continue
# Else update job record
pjob.jobSpec = dict(
zip(status._attributes, status.values()))
for k in pjob.jobSpec.keys():
if type(pjob.jobSpec[k]) not in [
type(''), type(1)
]:
pjob.jobSpec[k] = str(pjob.jobSpec[k])
logger.debug(
'Job %s with Panda job %s has changed status from %s to %s',
job.getFQID('.'), pjob.id, pjob.status,
status.jobStatus)
pjob.status = status.jobStatus
pjob.exitcode = str(status.transExitCode)
pjob.piloterrorcode = str(status.pilotErrorCode)
pjob.reason = ''
for k in pjob.jobSpec.keys():
if k.endswith('ErrorDiag'
) and pjob.jobSpec[k] != 'NULL':
pjob.reason += '%s: %s, ' % (
k, str(pjob.jobSpec[k]))
#if job.backend.jobSpec['transExitCode'] != 'NULL':
pjob.reason += 'transExitCode: %s' % pjob.jobSpec[
'transExitCode']
if status.jobStatus in [
'defined', 'unknown', 'assigned',
'waiting', 'activated', 'sent'
]:
logger.debug('Panda job %s %s' %
(pjob.id, status.jobStatus))
elif status.jobStatus in [
'starting', 'running', 'holding',
'transferring', 'merging'
]:
logger.debug('Panda job %s %s ' %
(pjob.id, status.jobStatus))
elif status.jobStatus in ['finished']:
logger.debug('Panda job %s %s ' %
(pjob.id, status.jobStatus))
elif status.jobStatus == 'failed':
logger.debug('Panda job %s %s ' %
(pjob.id, status.jobStatus))
# check for server side retry
if 'taskBufferErrorDiag' in pjob.jobSpec and pjob.jobSpec[
'taskBufferErrorDiag'].find(
"PandaID=") != -1:
# grab the new panda ID
newPandaID = long(
pjob.jobSpec['taskBufferErrorDiag'].
split("=")[1])
pjob.id = newPandaID
pjob.status = None
pjob.url = 'http://panda.cern.ch/?job=%d' % newPandaID
elif status.jobStatus == 'cancelled' and pjob.status not in [
'completed', 'failed'
]: # bug 67716
logger.debug('Panda job %s cancelled' %
pjob.id)
if 'taskBufferErrorDiag' in pjob.jobSpec and "rebrokerage" in pjob.jobSpec[
'taskBufferErrorDiag']:
newPandaID = checkForRebrokerage(
pjob.jobSpec['taskBufferErrorDiag'])
logger.warning(
"Subjob rebrokered by Panda server. Job %d moved to %d."
% (pjob.id, newPandaID))
pjob.id = newPandaID
pjob.status = None
else:
logger.warning('Unexpected job status %s',
status.jobStatus)
# history support
pconfDir = os.path.expanduser(os.environ['PANDA_CONFIG_ROOT'])
if not os.path.exists(pconfDir):
os.makedirs(pconfDir)
# set grid source file
globalConf = GlobalConfig.getConfig()
if globalConf.grid_src != '' and not os.environ.has_key(
'PATHENA_GRID_SETUP_SH'):
os.environ['PATHENA_GRID_SETUP_SH'] = globalConf.grid_src
# make tmp dir
tmpDir = tempfile.mkdtemp()
# set tmp dir in Client
Client.setGlobalTmpDir(tmpDir)
# look for PandaTools package
for path in sys.path:
if path == '':
path = '.'
if os.path.exists(path) and os.path.isdir(path) and 'pandatools' in os.listdir(path) \
and os.path.exists('%s/pandatools/__init__.py' % path):
# make symlink for module name
os.symlink('%s/pandatools' % path, '%s/taskbuffer' % tmpDir)
break
sys.path = [tmpDir] + sys.path
# total time
TotalCPUTime = 0
if not PsubUtils.checkOutDsName(options.outDS,
distinguishedName,
options.official,
nickName,
verbose=options.verbose):
tmpStr = "invalid output dataset name: %s" % options.outDS
tmpLog.error(tmpStr)
sys.exit(1)
# full execution string
fullExecString = PsubUtils.convSysArgv()
fullExecString += jsonExecStr
# use INTR server
if options.intrSrv:
Client.useIntrServer()
# create tmp dir
curDir = os.getcwd()
tmpDir = os.path.join(curDir, MiscUtils.wrappedUuidGen())
os.makedirs(tmpDir)
# exit action
def _onExit(dir, del_command):
del_command('rm -rf %s' % dir)
atexit.register(_onExit, tmpDir, MiscUtils.commands_get_output)
# sandbox
def master_prepare(self,app,appmasterconfig):
# PandaTools
from pandatools import Client
from pandatools import AthenaUtils
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('AthenaMCPandaRTHandler master_prepare called for %s', job.getFQID('.'))
usertag = configDQ2['usertag']
#usertag='user09'
nickname = getNickname(allowMissingNickname=True)
self.libDataset = '%s.%s.ganga.%s_%d.lib._%06d' % (usertag,nickname,commands.getoutput('hostname').split('.')[0],int(time.time()),job.id)
# self.userprefix='%s.%s.ganga' % (usertag,gridProxy.identity())
sources = 'sources.%s.tar.gz' % commands.getoutput('uuidgen 2> /dev/null')
self.library = '%s.lib.tgz' % self.libDataset
# check DBRelease
# if job.backend.dbRelease != '' and job.backend.dbRelease.find(':') == -1:
# raise ApplicationConfigurationError(None,"ERROR : invalid argument for backend.dbRelease. Must be 'DatasetName:FileName'")
# unpack library
logger.debug('Creating source tarball ...')
tmpdir = '/tmp/%s' % commands.getoutput('uuidgen 2> /dev/null')
os.mkdir(tmpdir)
inputbox=[]
if os.path.exists(app.transform_archive):
# must add a condition on size.
inputbox += [ File(app.transform_archive) ]
if app.evgen_job_option:
self.evgen_job_option=app.evgen_job_option
if os.path.exists(app.evgen_job_option):
# locally modified job option file to add to the input sand box
inputbox += [ File(app.evgen_job_option) ]
self.evgen_job_option=app.evgen_job_option.split("/")[-1]
# add input sandbox files
if (job.inputsandbox):
for file in job.inputsandbox:
inputbox += [ file ]
# add option files
for extFile in job.backend.extOutFile:
try:
shutil.copy(extFile,tmpdir)
except IOError:
os.makedirs(tmpdir)
shutil.copy(extFile,tmpdir)
# fill the archive
for opt_file in inputbox:
try:
shutil.copy(opt_file.name,tmpdir)
except IOError:
os.makedirs(tmpdir)
shutil.copy(opt_file.name,tmpdir)
# now tar it up again
inpw = job.getInputWorkspace()
rc, output = commands.getstatusoutput('tar czf %s -C %s .' % (inpw.getPath(sources),tmpdir))
if rc:
logger.error('Packing sources failed with status %d',rc)
logger.error(output)
raise ApplicationConfigurationError(None,'Packing sources failed.')
shutil.rmtree(tmpdir)
# upload sources
logger.debug('Uploading source tarball ...')
try:
cwd = os.getcwd()
os.chdir(inpw.getPath())
rc, output = Client.putFile(sources)
if output != 'True':
logger.error('Uploading sources %s failed. Status = %d', sources, rc)
logger.error(output)
raise ApplicationConfigurationError(None,'Uploading archive failed')
finally:
os.chdir(cwd)
# Use Panda's brokerage
## if job.inputdata and len(app.sites)>0:
## # update cloud, use inputdata's
## from dq2.info.TiersOfATLAS import whichCloud,ToACache
## inclouds=[]
## for site in app.sites:
## cloudSite=whichCloud(app.sites[0])
## if cloudSite not in inclouds:
## inclouds.append(cloudSite)
## # now converting inclouds content into proper brokering stuff.
## outclouds=[]
## for cloudSite in inclouds:
## for cloudID, eachCloud in ToACache.dbcloud.iteritems():
## if cloudSite==eachCloud:
## cloud=cloudID
## outclouds.append(cloud)
## break
## print outclouds
## # finally, matching with user's wishes
## if len(outclouds)>0:
## if not job.backend.requirements.cloud: # no user wish, update
## job.backend.requirements.cloud=outclouds[0]
## else:
## try:
## assert job.backend.requirements.cloud in outclouds
## except:
## raise ApplicationConfigurationError(None,'Input dataset not available in target cloud %s. Please try any of the following %s' % (job.backend.requirements.cloud, str(outclouds)))
from GangaPanda.Lib.Panda.Panda import runPandaBrokerage
runPandaBrokerage(job)
if job.backend.site == 'AUTO':
raise ApplicationConfigurationError(None,'site is still AUTO after brokerage!')
# output dataset preparation and registration
try:
outDsLocation = Client.PandaSites[job.backend.site]['ddm']
except:
raise ApplicationConfigurationError(None,"Could not extract output dataset location from job.backend.site value: %s. Aborting" % job.backend.site)
if not app.dryrun:
for outtype in app.outputpaths.keys():
dset=string.replace(app.outputpaths[outtype],"/",".")
dset=dset[1:]
# dataset registration must be done only once.
print "registering output dataset %s at %s" % (dset,outDsLocation)
try:
Client.addDataset(dset,False,location=outDsLocation)
dq2_set_dataset_lifetime(dset, location=outDsLocation)
except:
raise ApplicationConfigurationError(None,"Fail to create output dataset %s. Aborting" % dset)
# extend registration to build job lib dataset:
print "registering output dataset %s at %s" % (self.libDataset,outDsLocation)
try:
Client.addDataset(self.libDataset,False,location=outDsLocation)
dq2_set_dataset_lifetime(self.libDataset, outDsLocation)
except:
raise ApplicationConfigurationError(None,"Fail to create output dataset %s. Aborting" % self.libDataset)
###
cacheVer = "-AtlasProduction_" + str(app.prod_release)
logger.debug("master job submit?")
self.outsite=job.backend.site
if app.se_name and app.se_name != "none" and not self.outsite:
self.outsite=app.se_name
# create build job
jspec = JobSpec()
jspec.jobDefinitionID = job.id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.AtlasRelease = 'Atlas-%s' % app.atlas_rel
jspec.homepackage = 'AnalysisTransforms'+cacheVer#+nightVer
jspec.transformation = '%s/buildJob-00-00-03' % Client.baseURLSUB # common base to Athena and AthenaMC jobs: buildJob is a pilot job which takes care of all inputs for the real jobs (in prepare()
jspec.destinationDBlock = self.libDataset
jspec.destinationSE = job.backend.site
jspec.prodSourceLabel = 'panda'
jspec.assignedPriority = 2000
jspec.computingSite = job.backend.site
jspec.cloud = job.backend.requirements.cloud
# jspec.jobParameters = self.args not known yet
jspec.jobParameters = '-o %s' % (self.library)
if app.userarea:
print app.userarea
jspec.jobParameters += ' -i %s' % (os.path.basename(app.userarea))
else:
jspec.jobParameters += ' -i %s' % (sources)
jspec.cmtConfig = AthenaUtils.getCmtConfig(athenaVer=app.atlas_rel)
matchURL = re.search('(http.*://[^/]+)/',Client.baseURLSSL)
if matchURL:
jspec.jobParameters += ' --sourceURL %s' % matchURL.group(1)
fout = FileSpec()
fout.lfn = self.library
fout.type = 'output'
fout.dataset = self.libDataset
fout.destinationDBlock = self.libDataset
jspec.addFile(fout)
flog = FileSpec()
flog.lfn = '%s.log.tgz' % self.libDataset
flog.type = 'log'
flog.dataset = self.libDataset
flog.destinationDBlock = self.libDataset
jspec.addFile(flog)
#print "MASTER JOB DETAILS:",jspec.jobParameters
return jspec
for r in res:
ids.append(r[0])
print 'got', len(ids), 'new pandaIDs'
print
except cx_Oracle.DatabaseError, exc:
error, = exc.args
print "updatePandaWAN.py - problem in establishing connection to db"
print "updatePandaWAN.py Oracle-Error-Code:", error.code
print "updatePandaWAN.py Oracle-Error-Message:", error.message
print 'leaving only first 200 in the list.'
del ids[2000:]
from pandatools import Client
status,jobSpec = Client.getFullJobStatus(ids,False)
print 'got back from panda:', len(jobSpec)
try:
connection = cx_Oracle.Connection(line)
cursor = cx_Oracle.Cursor(connection)
print 'Connection established.'
for i in jobSpec:
if i is None:
print "problem..."
continue
ti=i.pilotTiming.split('|')
if len(ti)!=5:
print i.PandaID,i.pilotTiming,ti
#cursor.execute("DELETE FROM ATLAS_WANHCTEST.result where pandaid=:pid",{'pid':i.PandaID});
continue
def expandExcludedSiteList(job):
'''Expand a site list taking wildcards into account'''
# first, check if there's anything to be done
check_ddm = False
wildcard = False
excl_sites = []
for s in job.backend.requirements.excluded_sites:
if s.find("ANALY_") == -1:
check_ddm = True
if s.find("*") != -1:
wildcard = True
if s.find("ANALY_") != -1 and s.find("*") == -1:
excl_sites.append(s)
if not check_ddm and not wildcard:
return excl_sites
# we have either wildcards or DDM sites listed
# First, find the allowed sites for this job and ensure no duplicates anywhere
from pandatools import Client
logger.info(
"Excluding DDM and wildcarded sites from Jedi job. Please wait....")
orig_ddm_list = []
new_ddm_list = []
for s in job.inputdata.get_locations():
if not s in orig_ddm_list:
orig_ddm_list.append(s)
new_ddm_list.append(s)
orig_panda_list = []
for s in [Client.convertDQ2toPandaID(x) for x in new_ddm_list]:
for s2 in Client.PandaSites.keys():
if s2.find(s) != -1 and not s2 in orig_panda_list:
orig_panda_list.append(s2)
if check_ddm:
# remove any DDM sites that are referenced, including wildcards
for s in job.backend.requirements.excluded_sites:
if s in orig_ddm_list:
new_ddm_list.remove(s)
if s.find("*") != -1:
for s2 in orig_ddm_list:
if fnmatch.fnmatch(s2, s):
new_ddm_list.remove(s2)
# now recreate the panda list and see if any have been dropped
new_panda_list = []
for s in [Client.convertDQ2toPandaID(x) for x in new_ddm_list]:
for s2 in Client.PandaSites.keys():
if s2.find(s) != -1 and not s2 in new_panda_list:
new_panda_list.append(s)
for s in orig_panda_list:
if not s in new_panda_list and not s in excl_sites:
excl_sites.append(s)
if wildcard:
# find wilcarded ANALY_* sites and exclude any that match good sites
for s in job.backend.requirements.excluded_sites:
if s.find("*") != -1:
for s2 in orig_panda_list:
if fnmatch.fnmatch(s2, s) and not s2 in excl_sites:
excl_sites.append(s2)
return excl_sites
def retry(self,JobsetID,newSite=False,newOpts={},noSubmit=False,ignoreDuplication=False,useJobsetID=False,retryBuild=False,reproduceFiles=[],unsetRetryID=False):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# force update just in case
self.status(JobsetID,True)
# set an empty map since mutable default value is used
if newOpts == {}:
newOpts = {}
# get jobset
newJobsetID = -1
jobList = self.getJobIDsWithSetID(JobsetID)
if jobList == None:
# works only for jobsetID
if useJobsetID:
return
# works with jobID
isJobset = False
jobList = [JobsetID]
else:
isJobset = True
tmpMsg = "ID=%s is composed of JobID=" % JobsetID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for JobID in jobList:
# get job info from local repository
localJob = self.getJobInfo(JobID)
if localJob == None:
tmpLog.warning("JobID=%s not found in local repository. Synchronization may be needed" % JobID)
return None
# for JEDI
if localJob.isJEDI():
status,out = Client.retryTask(
localJob.jediTaskID,
verbose=self.verbose,
properErrorCode=True,
newParams=newOpts)
if status != 0:
tmpLog.error(status)
tmpLog.error(out)
tmpLog.error("Failed to retry TaskID=%s" % localJob.jediTaskID)
return False
tmpStat,tmpDiag = out
if (not tmpStat in [0,True] and newOpts == {}) or (newOpts != {} and tmpStat != 3):
tmpLog.error(tmpDiag)
tmpLog.error("Failed to retry TaskID=%s" % localJob.jediTaskID)
return False
tmpLog.info(tmpDiag)
continue
# skip running job
if localJob.dbStatus != 'frozen':
tmpLog.info('Retry failed subjobs in running jobId=%s' % JobID)
status,out = Client.retryFailedJobsInActive(JobID,verbose=self.verbose)
if status != 0:
tmpLog.error(status)
tmpLog.error(out)
tmpLog.error("Failed to retry JobID=%s" % JobID)
else:
job = self.status(JobID)
if isJobset:
continue
else:
return
# skip already retried
if localJob.retryID != '0':
if isJobset:
tmpLog.info('Skip JobID=%s since already retried by JobID=%s JobsetID=%s' % \
(JobID,localJob.retryID,localJob.retryJobsetID))
continue
else:
tmpLog.warning('This job was already retried by JobID=%s' % localJob.retryID)
return
# check status of buildJob
if not retryBuild and not localJob.buildStatus in ['','finished']:
tmpMsgStr = 'Cannot retry since status of buildJob %s is %s (!= finished). ' \
% (localJob.PandaID.split(',')[0],localJob.buildStatus)
tmpMsgStr += 'Please execute %s with the same input/output datasets (or containers). ' % localJob.jobType
tmpMsgStr += 'It will run only on failed/cancelled/unused input files '
tmpMsgStr += 'and append output files to the output dataset container. '
tmpMsgStr += 'Or you may set retryBuild=True in pbook.retry() '
tmpLog.warning(tmpMsgStr)
if isJobset:
continue
else:
return
# check opts for newSite
if newSite or newOpts != {}:
if not localJob.outDS.endswith('/') and not newOpts.has_key('outDS') and not newOpts.has_key('--outDS'):
tmpLog.warning('You need to specify --outDS in newOpts to retry at new site unless container is used as output')
return
# get list of failed jobs
pandaIDs = localJob.PandaID.split(',')
statusList= localJob.jobStatus.split(',')
jobList = []
for idx in range(len(pandaIDs)):
# check status unless reproduce files
if reproduceFiles == [] and not statusList[idx] in ['failed','cancelled']:
continue
jobList.append(pandaIDs[idx])
# no failed job
if jobList == []:
if isJobset:
tmpLog.info('Skip JobID=%s since no failed jobs' % JobID)
continue
else:
tmpLog.info('No failed jobs to be retried for JobID=%s' % JobID)
return
# get full job spec
tmpLog.info("Retrying JobID=%s ..." % JobID)
tmpLog.info("Getting job info")
idxJL = 0
nQuery = 500
pandaJobs = []
while idxJL < len(jobList):
# avoid burst query
tmpLog.info(" %5s/%s" % (idxJL,len(jobList)))
status,oTmp = Client.getFullJobStatus(
jobList[idxJL:idxJL+nQuery],
verbose=self.verbose)
if status != 0:
tmpLog.error(status)
tmpLog.error(oTmp)
tmpLog.error("Cannot get job info from Panda server")
return
pandaJobs += oTmp
idxJL += nQuery
time.sleep(1)
tmpLog.info(" %5s/%s" % (len(jobList),len(jobList)))
# get PandaIDs to reproduce files
if reproduceFiles != []:
# change wildcard to .* for regexp
reproduceFilePatt = []
for tmpReproduceFile in reproduceFiles:
if '*' in tmpReproduceFile:
tmpReproduceFile = tmpReproduceFile.replace('*','.*')
reproduceFilePatt.append(tmpReproduceFile)
# get list of jobs which produced interesting files
tmpJobList = []
tmpPandaJobs = []
for tmpPandaJob in pandaJobs:
# check names
tmpMatchFlag = False
for tmpFile in tmpPandaJob.Files:
if tmpFile.type == 'output' and tmpFile.status == 'ready':
for tmpReproduceFile in reproduceFilePatt:
# normal matching
if tmpReproduceFile == tmpFile.lfn:
tmpMatchFlag = True
break
# wild card
if '*' in tmpReproduceFile and \
re.search('^'+tmpReproduceFile,tmpFile.lfn) != None:
tmpMatchFlag = True
break
if tmpMatchFlag:
break
# append
if tmpMatchFlag:
tmpJobList.append(tmpPandaJob.PandaID)
tmpPandaJobs.append(tmpPandaJob)
# use new list
jobList = tmpJobList
pandaJobs = tmpPandaJobs
if jobList == []:
tmpLog.info("No jobs to reproduce files : Jobs in JobID=%s didn't produce lost files" % JobID)
continue
# jobdefID
newJobdefID = PsubUtils.readJobDefID()
# reset some parameters
retryJobs = []
retrySite = None
retryElement = None
retryDestSE = None
outDsName = None
shadowList = []
oldLibDS = None
newLibDS = None
newLibTgz = None
rebroMap = {}
for idx in range(len(jobList)):
job = pandaJobs[idx]
# skip exired
if job == None:
tmpLog.warning("Could not retry jobs older than 30 days : JobID=%s (PandaID=%s) expired" \
% (JobID,jobList[idxJob]))
return
# skip jobs reassigned by rebrokerage
if (job.jobStatus == 'cancelled' and job.taskBufferErrorCode in [105,'105']) or \
(job.jobStatus == 'failed' and job.taskBufferErrorCode in [106,'106']):
# extract JobIDs of reassigned jobs
tmpM = re.search('JobsetID=(\d+) JobID=(\d+)',job.taskBufferErrorDiag)
if tmpM != None:
tmpRebKey = (tmpM.group(1),tmpM.group(2))
if not rebroMap.has_key(tmpRebKey):
rebroMap[tmpRebKey] = 0
# count # of reassigned jobs
rebroMap[tmpRebKey] += 1
continue
# get shadow list
if (not ignoreDuplication) and outDsName == None and job.prodSourceLabel == 'user':
# look for dataset for log since it doesn't have suffix even when --individualOutDS is used
for tmpFile in job.Files:
if tmpFile.type == 'log':
outDsName = tmpFile.dataset
break
# output dataset was not found
if outDsName == None:
tmpLog.error("Could not get output dataset name for JobID=%s (PandaID=%s)" \
% (JobID,job.PandaID))
return
# get files in shadow
if outDsName.endswith('/'):
shadowList = Client.getFilesInShadowDataset(
outDsName,
Client.suffixShadow,
self.verbose)
else:
# disable duplication check mainly for old overlay jobs since non-signal files are wrongly skipped
#shadowList = Client.getFilesInShadowDatasetOld(outDsName,Client.suffixShadow,self.verbose)
pass
# unify sitename
if retrySite == None:
retrySite = job.computingSite
retryElement = job.computingElement
retryDestSE = job.destinationSE
# reset
job.jobStatus = None
job.commandToPilot = None
job.startTime = None
job.endTime = None
job.attemptNr = 1+job.attemptNr
for attr in job._attributes:
if attr.endswith('ErrorCode') or attr.endswith('ErrorDiag'):
setattr(job,attr,None)
job.transExitCode = None
job.computingSite = retrySite
job.computingElement = retryElement
job.destinationSE = retryDestSE
job.dispatchDBlock = None
if not unsetRetryID:
job.jobExecutionID = JobID
job.jobDefinitionID = newJobdefID
job.parentID = job.PandaID
if job.jobsetID != ['NULL',None,-1]:
if not unsetRetryID:
job.sourceSite = job.jobsetID
job.jobsetID = newJobsetID
skipInputList = []
numUsedFiles = 0
# loop over all files
for file in job.Files:
file.rowID = None
if file.type == 'input':
# protection against wrong sync which doesn't update buildStatus correctly
if not retryBuild and file.lfn.endswith('.lib.tgz') and file.GUID == 'NULL':
tmpLog.warning('GUID for %s is unknown. Cannot retry when corresponding buildJob failed' \
% file.lfn)
return
if not retryBuild or not file.lfn.endswith('.lib.tgz'):
file.status = 'ready'
# set new lib dataset
if retryBuild and file.lfn.endswith('.lib.tgz'):
if newLibTgz != None:
file.lfn = newLibTgz
file.dataset = newLibDS
file.dispatchDBlock = newLibDS
# check with shadow for non lib.tgz/DBR
tmpDbrMatch = re.search('^DBRelease-.*\.tar\.gz$',file.lfn)
if tmpDbrMatch == None and not file.lfn.endswith('.lib.tgz'):
if file.lfn in shadowList:
skipInputList.append(file)
else:
numUsedFiles += 1
elif file.type in ('output','log'):
file.destinationSE = retryDestSE
file.destinationDBlock = re.sub('_sub\d+$','',file.destinationDBlock)
# add retry num
if file.dataset.endswith('/') or job.prodSourceLabel == 'panda':
oldOutDsName = file.destinationDBlock
retryDsPatt = '_r'
if reproduceFiles != []:
retryDsPatt = '_rp'
retryMatch = re.search(retryDsPatt+'(\d+)$',file.destinationDBlock)
if retryMatch == None:
file.destinationDBlock += (retryDsPatt+'1')
else:
tmpDestinationDBlock = re.sub(retryDsPatt+'(\d+)$','',file.destinationDBlock)
file.destinationDBlock = tmpDestinationDBlock + retryDsPatt + '%d' % (1+int(retryMatch.group(1)))
if job.processingType == 'usermerge':
job.jobParameters = job.jobParameters.replace(' %s ' % oldOutDsName,
' %s ' % file.destinationDBlock)
# use new dataset name for buildXYZ
if job.prodSourceLabel == 'panda':
if file.lfn.endswith('.lib.tgz'):
# get new libDS and lib.tgz names
oldLibDS = file.dataset
file.dataset = file.destinationDBlock
newLibDS = file.dataset
file.lfn = re.sub(oldLibDS,newLibDS,file.lfn)
newLibTgz = file.lfn
else:
file.dataset = file.destinationDBlock
# add attempt nr
oldName = file.lfn
if job.prodSourceLabel == 'panda' and file.lfn.endswith('.lib.tgz'):
continue
else:
# append attempt number at the tail
file.lfn = re.sub("\.\d+$","",file.lfn)
file.lfn = "%s.%d" % (file.lfn,job.attemptNr)
newName = file.lfn
# modify jobParameters
job.jobParameters = re.sub("'%s'" % oldName ,"'%s'" % newName,
job.jobParameters)
# look for output in trf
oldGenelicName = re.sub('\.\d+$','',oldName)
match = re.search(oldGenelicName+'(\.\d+)*(%20|")',job.jobParameters)
if match != None:
job.jobParameters = job.jobParameters.replace(match.group(0),newName+match.group(2))
# change lib.tgz name
if retryBuild and newLibDS != None:
job.jobParameters = re.sub(oldLibDS,newLibDS,job.jobParameters)
# change destinationDBlock
if job.prodSourceLabel == 'panda':
job.destinationDBlock = newLibDS
# all files are used by others
if numUsedFiles == 0 and skipInputList != []:
continue
# remove skipped files
strSkipped = ''
for tmpFile in skipInputList:
strSkipped += '%s,' % tmpFile.lfn
job.Files.remove(tmpFile)
strSkipped = strSkipped[:-1]
# modify jobpar
if strSkipped != '':
optionToSkipFiles = '--skipInputByRetry'
if not optionToSkipFiles in job.jobParameters:
# just append
job.jobParameters += "%s=%s " % (optionToSkipFiles,strSkipped)
else:
# extract already skipped files
tmpMatch = re.search("(%s=[^ ]+)",job.jobParameters)
if tmpMatch == None:
tmpLog.error("Failed to extract arg of %s for PandaID=%s" \
% (optionToSkipFiles,job.PandaID))
return
# replace
job.jobParameters = re.sub(tmpMatch.group(1),"%s,%s" % (tmpMatch.group(1),optionToSkipFiles),
job.jobParameters)
if self.verbose:
tmpLog.debug(job.jobParameters)
# append
retryJobs.append(job)
# info on rebrokeage
if rebroMap != {}:
for tmpRebKey,tmpRebNumJobs in rebroMap.iteritems():
tmpRebSetID,tmpRebJobID = tmpRebKey
tmpLog.info('Skip %s jobs since JobID=%s JobsetID=%s already reassigned them to another site' % \
(tmpRebNumJobs,tmpRebJobID,tmpRebSetID))
if retryJobs == []:
tmpLog.info("No more jobs to be retried for JobID=%s" % JobID)
if isJobset:
continue
else:
return
# all input files were or are being used by other jobs
if retryJobs == []:
tmpLog.info('All input files were or are being used by other jobs for the same output. No jobs to be retried. If you need to ignore duplication check (e.g., using the same EVNT file for multiple simulation subjobs), set ignoreDuplication=True. i.e. retry(123,ignoreDuplication=True)')
if isJobset:
continue
else:
return
# check voms role
if not retryJobs[0].workingGroup in ['NULL',None,'']:
# VOMS role was used
if not "--workingGroup" in job.metadata:
# extract voms roles from metadata
match = re.search("--voms( |=)[ \"]*([^ \"]+)",job.metadata)
if match != None:
vomsRoles = match.group(2)
else:
vomsRoles = "atlas:/atlas/%s/Role=production" % retryJobs[0].workingGroup
# regenerate proxy with VOMS roles
try:
tmpLog.info("Checking proxy role to resubmit %s jobs" % retryJobs[0].workingGroup)
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,vomsRoles,
useCache=True)
except:
tmpLog.error("Failed to generate a proxy with %s" % vomsRoles)
return
# check runtime env for new site submission
if (newSite or newOpts != {}):
if retryJobs[0].processingType == 'pathena' or '--useAthenaPackages' in retryJobs[0].metadata:
from pandatools import AthenaUtils
stA,retA = AthenaUtils.getAthenaVer()
if not stA:
tmpLog.error("Failed to get Athena rel/cache version in current runtime env")
return
athenaVer = retA['athenaVer']
cacheVer = retA['cacheVer']
nightVer = retA['nightVer']
wrongSetup = False
if retryJobs[0].AtlasRelease != 'Atlas-%s' % athenaVer:
wrongSetup = True
errMsg = "Current Athena version Atlas-%s is inconsitent with the previous submission %s. " % (athenaVer,retryJobs[0].AtlasRelease)
elif retryJobs[0].homepackage != 'AnalysisTransforms'+cacheVer+nightVer:
wrongSetup = True
errMsg = "Current cache version %s is inconsitent with the previous submission. " % cacheVer.replace('-','').replace('_','-')
if wrongSetup:
errMsg += 'You need to have the same runtime env as before since all job spec need to be re-created to send jobs to a new site. '
errMsg += 'Please setup Athena correctly and restart pbook'
tmpLog.error(errMsg)
return
# test mode
if noSubmit:
continue
# invoke pathena/prun to send job to new site
if (newSite or newOpts != {}) and retryJobs[0].processingType != 'usermerge':
# set parent jobID and jobsetID
newOpts['provenanceID'] = retryJobs[0].jobExecutionID
newOpts['panda_parentJobsetID'] = retryJobs[0].sourceSite
tmpLog.info("Constructing job spec again to be sent to another site ...")
comStat= PsubUtils.execWithModifiedParams(retryJobs,newOpts,self.verbose,newSite)
if comStat == 0:
# update database
time.sleep(2)
self.sync()
else:
tmpLog.error("Failed to submit jobs to Panda server")
return
# register datasets
tmpOutDsLocation = Client.PandaSites[retryJobs[-1].computingSite]['ddm']
addedDataset = []
shadowDSname = None
for tmpFile in retryJobs[-1].Files:
if tmpFile.type in ['output','log'] and tmpFile.dataset.endswith('/'):
# add shadow
"""
removed shadow
if shadowDSname == None and tmpFile.type == 'log':
shadowDSname = "%s%s" % (tmpFile.destinationDBlock,Client.suffixShadow)
Client.addDataset(shadowDSname,self.verbose)
"""
# add datasets
if not tmpFile.destinationDBlock in addedDataset:
# create dataset
Client.addDataset(
tmpFile.destinationDBlock,
self.verbose,
location=tmpOutDsLocation,
dsCheck=False)
# add to container
Client.addDatasetsToContainer(
tmpFile.dataset,
[tmpFile.destinationDBlock],
self.verbose)
# append
addedDataset.append(tmpFile.destinationDBlock)
# register libDS
if retryBuild and newLibDS != None:
Client.addDataset(
newLibDS,
self.verbose,
location=tmpOutDsLocation,
dsCheck=False)
# submit
tmpLog.info("Submitting job ...")
status,out = Client.submitJobs(retryJobs,verbose=self.verbose)
if out == None or status != 0:
tmpLog.error(status)
tmpLog.error(out)
tmpLog.error("Failed to submit jobs to Panda server")
return
# update database
pandaIDstatus = {}
newJobID = None
for items in out:
# get newJobID
if newJobID == None:
newJobID = items[1]
# check PandaID
PandaID = items[0]
if PandaID == 'NULL':
tmpLog.error("Panda server returned wrong IDs. It may have a temporary problem")
return
# set newJobsetID
if newJobsetID in [None,-1]:
newJobsetID = items[2]['jobsetID']
# dummy statuso
pandaIDstatus[PandaID] = ('defined','NULL')
# set retry ID
if not unsetRetryID:
localJob.retryID = newJobID
if not newJobsetID in [None,-1,'NULL']:
localJob.retryJobsetID = newJobsetID
try:
PdbUtils.updateJobDB(localJob,self.verbose)
except:
tmpLog.error("Failed to set retryID for JobID=%s" % JobID)
return
# set new paramers
newLocalJob = PdbUtils.convertPtoD(retryJobs,pandaIDstatus)
newLocalJob.JobID = newJobID
if not newJobsetID in [None,-1,'NULL']:
newLocalJob.groupID = newJobsetID
newLocalJob.creationTime = datetime.datetime.utcnow()
# insert to DB
try:
PdbUtils.insertJobDB(newLocalJob,self.verbose)
except:
tmpLog.error("Failed to insert JobID=%s to local repository" % newJobID)
return
# write new jobdefID
PsubUtils.writeJobDefID(newJobID)
# done
tmpMsg = 'Done. New JobID=%s' % newJobID
if not newJobsetID in [None,-1,'NULL']:
tmpMsg += " JobsetID=%s" % newJobsetID
tmpLog.info(tmpMsg)
def master_prepare(self,app,appconfig):
'''Prepare the master job'''
from pandatools import Client
from taskbuffer.JobSpec import JobSpec
from taskbuffer.FileSpec import FileSpec
job = app._getParent()
logger.debug('ExecutablePandaRTHandler master_prepare called for %s', job.getFQID('.'))
# set chirp variables
if configPanda['chirpconfig'] or configPanda['chirpserver']:
setChirpVariables()
# Pack inputsandbox
inputsandbox = 'sources.%s.tar' % commands.getoutput('uuidgen 2> /dev/null')
inpw = job.getInputWorkspace()
# add user script to inputsandbox
if hasattr(job.application.exe, "name"):
if not job.application.exe in job.inputsandbox:
job.inputsandbox.append(job.application.exe)
for fname in [f.name for f in job.inputsandbox]:
fname.rstrip(os.sep)
path = fname[:fname.rfind(os.sep)]
f = fname[fname.rfind(os.sep)+1:]
rc, output = commands.getstatusoutput('tar rf %s -C %s %s' % (inpw.getPath(inputsandbox), path, f))
if rc:
logger.error('Packing inputsandbox failed with status %d',rc)
logger.error(output)
raise ApplicationConfigurationError('Packing inputsandbox failed.')
if len(job.inputsandbox) > 0:
rc, output = commands.getstatusoutput('gzip %s' % (inpw.getPath(inputsandbox)))
if rc:
logger.error('Packing inputsandbox failed with status %d',rc)
logger.error(output)
raise ApplicationConfigurationError('Packing inputsandbox failed.')
inputsandbox += ".gz"
else:
inputsandbox = None
# Upload Inputsandbox
if inputsandbox:
logger.debug('Uploading source tarball ...')
uploadSources(inpw.getPath(),os.path.basename(inputsandbox))
self.inputsandbox = inputsandbox
else:
self.inputsandbox = None
# input dataset
if job.inputdata:
if job.inputdata._name != 'DQ2Dataset':
raise ApplicationConfigurationError('PANDA application supports only DQ2Datasets')
# run brokerage here if not splitting
if not job.splitter:
from GangaPanda.Lib.Panda.Panda import runPandaBrokerage
runPandaBrokerage(job)
elif job.splitter._name not in ['DQ2JobSplitter', 'ArgSplitter', 'ArgSplitterTask']:
raise ApplicationConfigurationError('Panda splitter must be DQ2JobSplitter or ArgSplitter')
if job.backend.site == 'AUTO':
raise ApplicationConfigurationError('site is still AUTO after brokerage!')
# output dataset
if job.outputdata:
if job.outputdata._name != 'DQ2OutputDataset':
raise ApplicationConfigurationError('Panda backend supports only DQ2OutputDataset')
else:
logger.info('Adding missing DQ2OutputDataset')
job.outputdata = DQ2OutputDataset()
job.outputdata.datasetname,outlfn = dq2outputdatasetname(job.outputdata.datasetname, job.id, job.outputdata.isGroupDS, job.outputdata.groupname)
self.outDsLocation = Client.PandaSites[job.backend.site]['ddm']
try:
Client.addDataset(job.outputdata.datasetname,False,location=self.outDsLocation)
logger.info('Output dataset %s registered at %s'%(job.outputdata.datasetname,self.outDsLocation))
dq2_set_dataset_lifetime(job.outputdata.datasetname, location=self.outDsLocation)
except exceptions.SystemExit:
raise BackendError('Panda','Exception in Client.addDataset %s: %s %s'%(job.outputdata.datasetname,sys.exc_info()[0],sys.exc_info()[1]))
# handle the libds
if job.backend.libds:
self.libDataset = job.backend.libds
self.fileBO = getLibFileSpecFromLibDS(self.libDataset)
self.library = self.fileBO.lfn
elif job.backend.bexec:
self.libDataset = job.outputdata.datasetname+'.lib'
self.library = '%s.tgz' % self.libDataset
try:
Client.addDataset(self.libDataset,False,location=self.outDsLocation)
dq2_set_dataset_lifetime(self.libDataset, location=self.outDsLocation)
logger.info('Lib dataset %s registered at %s'%(self.libDataset,self.outDsLocation))
except exceptions.SystemExit:
raise BackendError('Panda','Exception in Client.addDataset %s: %s %s'%(self.libDataset,sys.exc_info()[0],sys.exc_info()[1]))
# collect extOutFiles
self.extOutFile = []
for tmpName in job.outputdata.outputdata:
if tmpName != '':
self.extOutFile.append(tmpName)
for tmpName in job.outputsandbox:
if tmpName != '':
self.extOutFile.append(tmpName)
for tmpName in job.backend.extOutFile:
if tmpName != '':
self.extOutFile.append(tmpName)
# create build job
if job.backend.bexec != '':
jspec = JobSpec()
jspec.jobDefinitionID = job.id
jspec.jobName = commands.getoutput('uuidgen 2> /dev/null')
jspec.transformation = '%s/buildGen-00-00-01' % Client.baseURLSUB
if Client.isDQ2free(job.backend.site):
jspec.destinationDBlock = '%s/%s' % (job.outputdata.datasetname,self.libDataset)
jspec.destinationSE = 'local'
else:
jspec.destinationDBlock = self.libDataset
jspec.destinationSE = job.backend.site
jspec.prodSourceLabel = configPanda['prodSourceLabelBuild']
jspec.processingType = configPanda['processingType']
jspec.assignedPriority = configPanda['assignedPriorityBuild']
jspec.computingSite = job.backend.site
jspec.cloud = job.backend.requirements.cloud
jspec.jobParameters = '-o %s' % (self.library)
if self.inputsandbox:
jspec.jobParameters += ' -i %s' % (self.inputsandbox)
else:
raise ApplicationConfigurationError('Executable on Panda with build job defined, but inputsandbox is emtpy !')
matchURL = re.search('(http.*://[^/]+)/',Client.baseURLCSRVSSL)
if matchURL:
jspec.jobParameters += ' --sourceURL %s ' % matchURL.group(1)
if job.backend.bexec != '':
jspec.jobParameters += ' --bexec "%s" ' % urllib.quote(job.backend.bexec)
jspec.jobParameters += ' -r %s ' % '.'
fout = FileSpec()
fout.lfn = self.library
fout.type = 'output'
fout.dataset = self.libDataset
fout.destinationDBlock = self.libDataset
jspec.addFile(fout)
flog = FileSpec()
flog.lfn = '%s.log.tgz' % self.libDataset
flog.type = 'log'
flog.dataset = self.libDataset
flog.destinationDBlock = self.libDataset
jspec.addFile(flog)
return jspec
else:
return None
def kill(self,JobID,useJobsetID=False):
# get logger
tmpLog = PLogger.getPandaLogger()
# check proxy
self.gridPassPhrase,self.vomsFQAN = PsubUtils.checkGridProxy(
self.gridPassPhrase,
False,
self.verbose,
useCache=True)
# force update just in case
self.status(JobID,True)
# get jobset
jobList = self.getJobIDsWithSetID(JobID)
if jobList == None:
# works only for jobsetID
if useJobsetID:
return
# works with jobID
jobList = [JobID]
else:
tmpMsg = "ID=%s is composed of JobID=" % JobID
for tmpJobID in jobList:
tmpMsg += '%s,' % tmpJobID
tmpMsg = tmpMsg[:-1]
tmpLog.info(tmpMsg)
for tmpJobID in jobList:
# get job info from local repository
job = self.getJobInfo(tmpJobID)
if job == None:
tmpLog.warning("JobID=%s not found in local repository. Synchronization may be needed" % tmpJobID)
continue
# skip frozen job
if job.dbStatus == 'frozen':
tmpLog.info('All subJobs in JobID=%s already finished/failed' % tmpJobID)
continue
if not job.isJEDI():
# get PandaID list
killJobs = job.PandaID.split(',')
# kill
tmpLog.info('Sending kill command ...')
status,output = Client.killJobs(killJobs,self.verbose)
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to kill JobID=%s" % tmpJobID)
return False
# update database
job.commandToPilot = 'tobekilled'
# update DB
try:
PdbUtils.updateJobDB(job,self.verbose)
except:
tmpLog.error("Failed to update local repository for JobID=%s" % tmpJobID)
return False
else:
# kill JEDI task
tmpLog.info('Sending killTask command ...')
status,output = Client.killTask(job.jediTaskID,self.verbose)
# communication error
if status != 0:
tmpLog.error(output)
tmpLog.error("Failed to kill JobID=%s" % tmpJobID)
return False
tmpStat,tmpDiag = output
if not tmpStat:
tmpLog.error(tmpDiag)
tmpLog.error("Failed to kill JobID=%s" % tmpJobID)
return False
tmpLog.info(tmpDiag)
# done
if job.isJEDI():
tmpLog.info('Done. TaskID=%s will be killed in 30min' % job.jediTaskID)
else:
tmpLog.info('Done. JobID=%s will be killed in 30min' % tmpJobID)
return True
