def _setup_bulk_subjobs(self, dirac_ids, dirac_script): """ This is the old bulk submit method which is used to construct the subjobs for a parametric job Args: dirac_ids (list): This is a list of the Dirac ids which have been created dirac_script (str): Name of the dirac script which contains the job jdl """ f = open(dirac_script, 'r') parametric_datasets = get_parametric_datasets(f.read().split('\n')) f.close() if len(parametric_datasets) != len(dirac_ids): raise BackendError( 'Dirac', 'Missmatch between number of datasets defines in dirac API script and those returned by DIRAC' ) master_job = self.getJobObject() master_job.subjobs = [] for i in range(len(dirac_ids)): j = Job() j.copyFrom(master_job) j.splitter = None j.backend.id = dirac_ids[i] j.id = i j.inputdata = self._setup_subjob_dataset(parametric_datasets[i]) j.status = 'submitted' j.time.timenow('submitted') master_job.subjobs.append(j) return True
def master_setup_bulk_subjobs(self, jobs, jdefids): from Ganga.GPIDev.Lib.Job.Job import Job master_job = self.getJobObject() for i in range(len(jdefids)): j = Job() j.copyFrom(master_job) j.splitter = None j.backend = Panda() j.backend.id = jdefids[i] j.id = i j.status = 'submitted' j.time.timenow('submitted') master_job.subjobs.append(j) return True
def createSubjob(self, job, additional_skip_args=None): """ Create a new subjob by copying the master job and setting all fields correctly. """ from Ganga.GPIDev.Lib.Job.Job import Job if additional_skip_args is None: additional_skip_args = [] j = Job() skipping_args = [ 'splitter', 'inputsandbox', 'inputfiles', 'inputdata', 'subjobs' ] for arg in additional_skip_args: skipping_args.append(arg) j.copyFrom(job, skipping_args) j.splitter = None j.inputsandbox = [] j.inputfiles = [] j.inputdata = None return j
def _setup_bulk_subjobs(self, dirac_ids, dirac_script): f = open(dirac_script, 'r') parametric_datasets = get_parametric_datasets(f.read().split('\n')) f.close() if len(parametric_datasets) != len(dirac_ids): raise BackendError('Dirac', 'Missmatch between number of datasets defines in dirac API script and those returned by DIRAC') from Ganga.GPIDev.Lib.Job.Job import Job master_job = self.getJobObject() master_job.subjobs = [] for i in range(len(dirac_ids)): j = Job() j.copyFrom(master_job) j.splitter = None j.backend.id = dirac_ids[i] j.id = i j.inputdata = self._setup_subjob_dataset(parametric_datasets[i]) j.status = 'submitted' j.time.timenow('submitted') master_job.subjobs.append(j) master_job._commit() return True
def submit(N, K): jobs = [] for i in range(K): j = Job() j._auto__init__() j.backend = LCG() j.backend.middleware = 'GLITE' j.splitter = GenericSplitter() j.splitter.attribute = 'application.args' j.splitter.values = [['x']] * N j.submit() jobs.append(j) import time def finished(): for j in jobs: if not j.status in ['failed', 'completed']: return False return True while not finished(): time.sleep(1) return jobs
def master_updateMonitoringInformation(jobs): """Updates the statuses of the list of jobs provided by issuing crab -status.""" logger.info('Updating the monitoring information of ' + str(len(jobs)) + ' jobs') try: from Ganga.GPIDev.Lib.Job.Job import Job import sys, traceback for j in jobs: server = CRABServer() logger.debug( 'Updating monitoring information for job %d (%s)' % (j.id, j.status)) try: dictresult, status, reason = server.status(j) logger.info( 'CRAB3 server call answer status: %s - reason: %s' % (status, reason)) joblist = sorted(dictresult['result'][0]['jobList'], key=lambda x: x[1]) except KeyError: logger.info( 'Get status for job %d didn\'t return job list, skipping job for now.' % j.id) continue except: logger.error('Get status for job %d failed, skipping.' % j.id) raise if joblist: logger.info('There are subjob statuses for job %s' % j.id) logger.info('j: %s' % dir(j)) if not j.subjobs: logger.warning('No subjob object for job %s' % j.id) j.subjobs = [] for i in xrange(len(joblist)): subjob = joblist[i] index = int(subjob[1]) logger.info('Processing subjob %d, %s' % (index, subjob)) sj = Job() sj.copyFrom(j) sj.backend.crabid = index sj.id = i sj.updateStatus('submitting') sj.backend.checkReport(subjob) sj.backend.checkStatus() j.subjobs.append(sj) #j.subjobs = sorted(j.subjobs, key=lambda x: x.backend.id) #j._commit() else: for subjob in joblist: index = int(subjob[1]) logger.debug( 'Found subjob %s searching with index %s' % (j.subjobs[index - 1].backend.crabid, index)) j.subjobs[index - 1].backend.checkReport(subjob) j.subjobs[index - 1].backend.checkStatus() j.updateMasterJobStatus() else: logger.info('There are no subjobs for job %s' % (j.id)) logger.info('checking task status from report: %s' % dictresult['result'][0]['status']) taskstatus = dictresult['result'][0]['status'] if taskstatus in ['FAILED']: logger.info('Job failed: %s' % dictresult) j.updateStatus('failed') except Exception as e: logger.error(e) traceback.print_exc(file=sys.stdout)
def createUnits(self): """Create new units if required given the inputdata""" # call parent for chaining super(CoreTransform, self).createUnits() # Use the given splitter to create the unit definitions if len(self.units) > 0: # already have units so return return if self.unit_splitter == None and len(self.inputdata) == 0: raise ApplicationConfigurationError( "No unit splitter or InputData provided for CoreTransform unit creation, Transform %d (%s)" % (self.getID(), self.name)) # ----------------------------------------------------------------- # split over unit_splitter by preference if self.unit_splitter: # create a dummy job, assign everything and then call the split j = Job() j.backend = self.backend.clone() j.application = self.application.clone() if self.inputdata: j.inputdata = self.inputdata.clone() subjobs = self.unit_splitter.split(j) if len(subjobs) == 0: raise ApplicationConfigurationError( "Unit splitter gave no subjobs after split for CoreTransform unit creation, Transform %d (%s)" % (self.getID(), self.name)) # only copy the appropriate elements fields = [] if len(self.fields_to_copy) > 0: fields = self.fields_to_copy elif isType(self.unit_splitter, GenericSplitter): if self.unit_splitter.attribute != "": fields = [self.unit_splitter.attribute.split(".")[0]] else: for attr in self.unit_splitter.multi_attrs.keys(): fields.append(attr.split(".")[0]) # now create the units from these jobs for sj in subjobs: unit = CoreUnit() for attr in fields: setattr(unit, attr, copy.deepcopy(getattr(sj, attr))) self.addUnitToTRF(unit) # ----------------------------------------------------------------- # otherwise split on inputdata elif len(self.inputdata) > 0: if self.files_per_unit > 0: # combine all files and split accorindgly filelist = [] for ds in self.inputdata: if isType(ds, GangaDataset): for f in ds.files: if f.containsWildcards(): # we have a wildcard so grab the subfiles for sf in f.getSubFiles( process_wildcards=True): filelist.append(sf) else: # no wildcards so just add the file filelist.append(f) else: logger.warning("Dataset '%s' doesn't support files" % getName(ds)) # create DSs and units for this list of files fid = 0 while fid < len(filelist): unit = CoreUnit() unit.name = "Unit %d" % len(self.units) unit.inputdata = GangaDataset( files=filelist[fid:fid + self.files_per_unit]) unit.inputdata.treat_as_inputfiles = self.inputdata[ 0].treat_as_inputfiles fid += self.files_per_unit self.addUnitToTRF(unit) else: # just produce one unit per dataset for ds in self.inputdata: # avoid splitting over chain inputs if isType(ds, TaskChainInput): continue unit = CoreUnit() unit.name = "Unit %d" % len(self.units) unit.inputdata = copy.deepcopy(ds) self.addUnitToTRF(unit)
def makeRegisteredJob(): """Makes a new Job and registers it with the Registry""" from Ganga.GPIDev.Lib.Job.Job import Job j = Job() j._auto__init__() return j