def __mt_job_prepare__(self, rjobs, subjobconfigs, masterjobconfig):
'''preparing jobs in multiple threads'''
logger.warning(
'preparing %d subjobs ... it may take a while' % len(rjobs))
# prepare the master job (i.e. create shared inputsandbox, etc.)
master_input_sandbox = IBackend.master_prepare(self, masterjobconfig)
# uploading the master job if it's over the WMS sandbox limitation
for f in master_input_sandbox:
master_input_idx = self.__check_and_prestage_inputfile__(f)
if not master_input_idx:
logger.error('master input sandbox perparation failed: %s' % f)
return None
# the algorithm for preparing a single bulk job
class MyAlgorithm(Algorithm):
def __init__(self):
Algorithm.__init__(self)
def process(self, sj_info):
my_sc = sj_info[0]
my_sj = sj_info[1]
try:
logger.debug("preparing job %s" % my_sj.getFQID('.'))
jdlpath = my_sj.backend.preparejob(
my_sc, master_input_sandbox)
if (not jdlpath) or (not os.path.exists(jdlpath)):
raise GangaException(
'job %s not properly prepared' % my_sj.getFQID('.'))
self.__appendResult__(my_sj.id, jdlpath)
return True
except Exception as x:
log_user_exception()
return False
mt_data = []
for sc, sj in zip(subjobconfigs, rjobs):
mt_data.append([sc, sj])
myAlg = MyAlgorithm()
myData = Data(collection=mt_data)
runner = MTRunner(
name='lcg_jprepare', algorithm=myAlg, data=myData, numThread=10)
runner.start()
runner.join(-1)
if len(runner.getDoneList()) < len(mt_data):
return None
else:
# return a JDL file dictionary with subjob ids as keys, JDL file
# paths as values
return runner.getResults()
def __mt_job_prepare__(self, rjobs, subjobconfigs, masterjobconfig):
'''preparing jobs in multiple threads'''
logger.warning(
'preparing %d subjobs ... it may take a while' % len(rjobs))
# prepare the master job (i.e. create shared inputsandbox, etc.)
master_input_sandbox = IBackend.master_prepare(self, masterjobconfig)
# uploading the master job if it's over the WMS sandbox limitation
for f in master_input_sandbox:
master_input_idx = self.__check_and_prestage_inputfile__(f)
if not master_input_idx:
logger.error('master input sandbox perparation failed: %s' % f)
return None
# the algorithm for preparing a single bulk job
class MyAlgorithm(Algorithm):
def __init__(self):
Algorithm.__init__(self)
def process(self, sj_info):
my_sc = sj_info[0]
my_sj = sj_info[1]
try:
logger.debug("preparing job %s" % my_sj.getFQID('.'))
jdlpath = my_sj.backend.preparejob(
my_sc, master_input_sandbox)
if (not jdlpath) or (not os.path.exists(jdlpath)):
raise GangaException(
'job %s not properly prepared' % my_sj.getFQID('.'))
self.__appendResult__(my_sj.id, jdlpath)
return True
except Exception as x:
log_user_exception()
return False
mt_data = []
for sc, sj in zip(subjobconfigs, rjobs):
mt_data.append([sc, sj])
myAlg = MyAlgorithm()
myData = Data(collection=mt_data)
runner = MTRunner(
name='lcg_jprepare', algorithm=myAlg, data=myData, numThread=10)
runner.start()
runner.join(-1)
if len(runner.getDoneList()) < len(mt_data):
return None
else:
# return a JDL file dictionary with subjob ids as keys, JDL file
# paths as values
return runner.getResults()
def __mt_bulk_submit__(self, node_jdls):
'''submitting jobs in multiple threads'''
job = self.getJobObject()
logger.warning('submitting %d subjobs ... it may take a while' %
len(node_jdls))
# the algorithm for submitting a single bulk job
class MyAlgorithm(Algorithm):
def __init__(self, cred_req, masterInputWorkspace, ce, arcverbose):
Algorithm.__init__(self)
self.inpw = masterInputWorkspace
self.cred_req = cred_req
self.ce = ce
self.arcverbose = arcverbose
def process(self, jdl_info):
my_sj_id = jdl_info[0]
my_sj_jdl = jdl_info[1]
my_sj_jid = Grid.arc_submit(my_sj_jdl, self.ce,
self.arcverbose, self.cred_req)
if not my_sj_jid:
return False
else:
self.__appendResult__(my_sj_id, my_sj_jid)
return True
mt_data = []
for id, jdl in node_jdls.items():
mt_data.append((id, jdl))
myAlg = MyAlgorithm(cred_req=self.credential_requirements,
masterInputWorkspace=job.getInputWorkspace(),
ce=self.CE,
arcverbose=self.verbose)
myData = Data(collection=mt_data)
runner = MTRunner(name='arc_jsubmit',
algorithm=myAlg,
data=myData,
numThread=config['SubmissionThread'])
runner.start()
runner.join(timeout=-1)
if len(runner.getDoneList()) < len(mt_data):
# not all bulk jobs are successfully submitted. canceling the
# submitted jobs on WMS immediately
logger.error(
'some bulk jobs not successfully (re)submitted, canceling submitted jobs on WMS'
)
Grid.arc_cancel_multiple(runner.getResults().values(),
self.credential_requirements)
return None
else:
return runner.getResults()
def __mt_bulk_submit__(self, node_jdls):
'''submitting jobs in multiple threads'''
job = self.getJobObject()
logger.warning(
'submitting %d subjobs ... it may take a while' % len(node_jdls))
# the algorithm for submitting a single bulk job
class MyAlgorithm(Algorithm):
def __init__(self, gridObj, masterInputWorkspace, ce, arcverbose):
Algorithm.__init__(self)
self.inpw = masterInputWorkspace
self.gridObj = gridObj
self.ce = ce
self.arcverbose = arcverbose
def process(self, jdl_info):
my_sj_id = jdl_info[0]
my_sj_jdl = jdl_info[1]
#my_sj_jid = self.gridObj.arc_submit(my_sj_jdl, self.ce, self.verbose)
my_sj_jid = self.gridObj.arc_submit(
my_sj_jdl, self.ce, self.arcverbose)
if not my_sj_jid:
return False
else:
self.__appendResult__(my_sj_id, my_sj_jid)
return True
mt_data = []
for id, jdl in node_jdls.items():
mt_data.append((id, jdl))
myAlg = MyAlgorithm(gridObj=grids['GLITE'], masterInputWorkspace=job.getInputWorkspace(
), ce=self.CE, arcverbose=self.verbose)
myData = Data(collection=mt_data)
runner = MTRunner(name='arc_jsubmit', algorithm=myAlg,
data=myData, numThread=config['SubmissionThread'])
runner.start()
runner.join(timeout=-1)
if len(runner.getDoneList()) < len(mt_data):
# not all bulk jobs are successfully submitted. canceling the
# submitted jobs on WMS immediately
logger.error(
'some bulk jobs not successfully (re)submitted, canceling submitted jobs on WMS')
grids['GLITE'].arc_cancelMultiple(runner.getResults().values())
return None
else:
return runner.getResults()
