def defineEvgen16Job(self, i): """Define an Evgen16 job based on predefined values and randomly generated names """ job = JobSpec() job.computingSite = self.__site job.cloud = self.__cloud job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (uuid.uuid1(), i) job.AtlasRelease = 'Atlas-16.6.2' job.homepackage = 'AtlasProduction/16.6.2.1' job.transformation = 'Evgen_trf.py' job.destinationDBlock = self.__datasetName job.destinationSE = self.__destName job.currentPriority = 10000 job.prodSourceLabel = 'test' job.cmtConfig = 'i686-slc5-gcc43-opt' #Output file fileO = FileSpec() fileO.lfn = "%s.evgen.pool.root" % job.jobName fileO.destinationDBlock = job.destinationDBlock fileO.destinationSE = job.destinationSE fileO.dataset = job.destinationDBlock fileO.destinationDBlockToken = 'ATLASDATADISK' fileO.type = 'output' job.addFile(fileO) #Log file fileL = FileSpec() fileL.lfn = "%s.job.log.tgz" % job.jobName fileL.destinationDBlock = job.destinationDBlock fileL.destinationSE = job.destinationSE fileL.dataset = job.destinationDBlock fileL.destinationDBlockToken = 'ATLASDATADISK' fileL.type = 'log' job.addFile(fileL) job.jobParameters = "2760 105048 19901 101 200 MC10.105048.PythiaB_ccmu3mu1X.py %s NONE NONE NONE MC10JobOpts-latest-test.tar.gz" % fileO.lfn return job
def run(self, data): datasetName = 'panda:panda.destDB.%s' % commands.getoutput('uuidgen') destName = 'ANALY_RRC-KI-HPC' site = 'ANALY_RRC-KI-HPC' scope = config['DEFAULT_SCOPE'] distributive = data['distributive'] release = data['release'] parameters = data['parameters'] input_type = data['input_type'] input_params = data['input_params'] input_files = data['input_files'] output_type = data['output_type'] output_params = data['output_params'] output_files = data['output_files'] jobid = data['jobid'] _logger.debug('Jobid: ' + str(jobid)) job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = commands.getoutput('uuidgen') job.transformation = config['DEFAULT_TRF'] job.destinationDBlock = datasetName job.destinationSE = destName job.currentPriority = 1000 job.prodSourceLabel = 'user' job.computingSite = site job.cloud = 'RU' job.prodDBlock = "%s:%s.%s" % (scope, scope, job.jobName) job.jobParameters = '%s %s "%s"' % (release, distributive, parameters) params = {} _logger.debug('MoveData') ec = 0 ec, uploaded_input_files = movedata( params=params, fileList=input_files, fromType=input_type, fromParams=input_params, toType='hpc', toParams={'dest': '/' + re.sub(':', '/', job.prodDBlock)}) if ec != 0: _logger.error('Move data error: ' + ec[1]) return for file in uploaded_input_files: fileIT = FileSpec() fileIT.lfn = file fileIT.dataset = job.prodDBlock fileIT.prodDBlock = job.prodDBlock fileIT.type = 'input' fileIT.scope = scope fileIT.status = 'ready' fileIT.GUID = commands.getoutput('uuidgen') job.addFile(fileIT) for file in output_files: fileOT = FileSpec() fileOT.lfn = file fileOT.destinationDBlock = job.prodDBlock fileOT.destinationSE = job.destinationSE fileOT.dataset = job.prodDBlock fileOT.type = 'output' fileOT.scope = scope fileOT.GUID = commands.getoutput('uuidgen') job.addFile(fileOT) fileOL = FileSpec() fileOL.lfn = "%s.log.tgz" % job.jobName fileOL.destinationDBlock = job.destinationDBlock fileOL.destinationSE = job.destinationSE fileOL.dataset = job.destinationDBlock fileOL.type = 'log' fileOL.scope = 'panda' job.addFile(fileOL) self.jobList.append(job) #submitJob o = self.submitJobs(self.jobList) x = o[0] #update PandaID conn = MySQLdb.connect( host=self.dbhost, db=self.dbname, # port=self.dbport, connect_timeout=self.dbtimeout, user=self.dbuser, passwd=self.dbpasswd) cur = conn.cursor() try: varDict = {} PandaID = int(x[0]) varDict['id'] = jobid varDict['pandaId'] = PandaID sql = "UPDATE %s SET %s.pandaId=%s WHERE %s.id=%s" % ( self.table_jobs, self.table_jobs, varDict['pandaId'], self.table_jobs, varDict['id']) cur.execute(sql, varDict) except: _logger.error('SENDJOB: Incorrect server response') try: conn.commit() return True except: _logger.error("commit error") return False
def 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 prepare(self, app, appsubconfig, appmasterconfig, jobmasterconfig): '''prepare the subjob specific configuration''' from pandatools import Client from taskbuffer.JobSpec import JobSpec from taskbuffer.FileSpec import FileSpec job = app._getParent() logger.debug('AthenaPandaRTHandler prepare called for %s', job.getFQID('.')) # in case of a simple job get the dataset content, otherwise subjobs are filled by the splitter if job.inputdata and not job._getRoot().subjobs: if not job.inputdata.names: contents = job.inputdata.get_contents(overlap=False, size=True) for ds in contents.keys(): for f in contents[ds]: job.inputdata.guids.append(f[0]) job.inputdata.names.append(f[1][0]) job.inputdata.sizes.append(f[1][1]) job.inputdata.checksums.append(f[1][2]) job.inputdata.scopes.append(f[1][3]) site = job._getRoot().backend.site job.backend.site = site job.backend.actualCE = site cloud = job._getRoot().backend.requirements.cloud job.backend.requirements.cloud = cloud # if no outputdata are given if not job.outputdata: job.outputdata = DQ2OutputDataset() job.outputdata.datasetname = job._getRoot().outputdata.datasetname #if not job.outputdata.datasetname: else: job.outputdata.datasetname = job._getRoot().outputdata.datasetname if not job.outputdata.datasetname: raise ApplicationConfigurationError( None, 'DQ2OutputDataset has no datasetname') jspec = JobSpec() jspec.jobDefinitionID = job._getRoot().id jspec.jobName = commands.getoutput('uuidgen 2> /dev/null') jspec.transformation = '%s/runGen-00-00-02' % Client.baseURLSUB if job.inputdata: jspec.prodDBlock = job.inputdata.dataset[0] else: jspec.prodDBlock = 'NULL' jspec.destinationDBlock = job.outputdata.datasetname if job.outputdata.location: if not job._getRoot().subjobs or job.id == 0: logger.warning( 'You have specified outputdata.location. Note that Panda may not support writing to a user-defined output location.' ) jspec.destinationSE = job.outputdata.location else: jspec.destinationSE = site jspec.prodSourceLabel = configPanda['prodSourceLabelRun'] jspec.processingType = configPanda['processingType'] jspec.assignedPriority = configPanda['assignedPriorityRun'] jspec.cloud = cloud # memory if job.backend.requirements.memory != -1: jspec.minRamCount = job.backend.requirements.memory # cputime if job.backend.requirements.cputime != -1: jspec.maxCpuCount = job.backend.requirements.cputime jspec.computingSite = site # library (source files) if job.backend.libds: flib = FileSpec() flib.lfn = self.fileBO.lfn flib.GUID = self.fileBO.GUID flib.type = 'input' flib.status = self.fileBO.status flib.dataset = self.fileBO.destinationDBlock flib.dispatchDBlock = self.fileBO.destinationDBlock jspec.addFile(flib) elif job.backend.bexec: flib = FileSpec() flib.lfn = self.library flib.type = 'input' flib.dataset = self.libDataset flib.dispatchDBlock = self.libDataset jspec.addFile(flib) # input files FIXME: many more input types if job.inputdata: for guid, lfn, size, checksum, scope in zip( job.inputdata.guids, job.inputdata.names, job.inputdata.sizes, job.inputdata.checksums, job.inputdata.scopes): finp = FileSpec() finp.lfn = lfn finp.GUID = guid finp.scope = scope # finp.fsize = # finp.md5sum = finp.dataset = job.inputdata.dataset[0] finp.prodDBlock = job.inputdata.dataset[0] finp.dispatchDBlock = job.inputdata.dataset[0] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # output files # outMap = {} #FIXME: if options.outMeta != []: self.rundirectory = "." # log files flog = FileSpec() flog.lfn = '%s._$PANDAID.log.tgz' % job.outputdata.datasetname flog.type = 'log' flog.dataset = job.outputdata.datasetname flog.destinationDBlock = job.outputdata.datasetname flog.destinationSE = job.backend.site jspec.addFile(flog) # job parameters param = '' # source URL matchURL = re.search("(http.*://[^/]+)/", Client.baseURLCSRVSSL) srcURL = "" if matchURL != None: srcURL = matchURL.group(1) param += " --sourceURL %s " % srcURL param += '-r "%s" ' % self.rundirectory exe_name = job.application.exe if job.backend.bexec == '': if hasattr(job.application.exe, "name"): exe_name = os.path.basename(job.application.exe.name) # set jobO parameter if job.application.args: param += ' -j "" -p "%s %s" ' % ( exe_name, urllib.quote(" ".join(job.application.args))) else: param += ' -j "" -p "%s" ' % exe_name if self.inputsandbox: param += ' -a %s ' % self.inputsandbox else: param += '-l %s ' % self.library param += '-j "" -p "%s %s" ' % ( exe_name, urllib.quote(" ".join(job.application.args))) if job.inputdata: param += '-i "%s" ' % job.inputdata.names # fill outfiles outfiles = {} for f in self.extOutFile: tarnum = 1 if f.find('*') != -1: # archive * outfiles[f] = "outputbox%i.%s.%s.tar.gz" % ( tarnum, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S")) tarnum += 1 else: outfiles[f] = "%s.%s.%s" % (f, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S")) fout = FileSpec() fout.lfn = outfiles[f] fout.type = 'output' fout.dataset = job.outputdata.datasetname fout.destinationDBlock = job.outputdata.datasetname fout.destinationSE = job.backend.site jspec.addFile(fout) param += '-o "%s" ' % ( outfiles ) # must be double quotes, because python prints strings in 'single quotes' for file in jspec.Files: if file.type in ['output', 'log'] and configPanda['chirpconfig']: file.dispatchDBlockToken = configPanda['chirpconfig'] logger.debug('chirp file %s', file) jspec.jobParameters = param return jspec
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
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
for lfn in files.keys(): index += 1 job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'), index) job.AtlasRelease = 'Atlas-13.0.40' job.homepackage = 'AtlasProduction/13.0.40.3' job.transformation = 'csc_simul_trf.py' job.destinationDBlock = datasetName job.destinationSE = destName job.computingSite = site job.prodDBlock = 'valid1.005001.pythia_minbias.evgen.EVNT.e306_tid019128' job.prodSourceLabel = 'test' job.currentPriority = 10000 job.cloud = 'IT' fileI = FileSpec() fileI.dataset = job.prodDBlock fileI.prodDBlock = job.prodDBlock fileI.lfn = lfn fileI.type = 'input' job.addFile(fileI) fileD = FileSpec() fileD.dataset = 'ddo.000001.Atlas.Ideal.DBRelease.v040701' fileD.prodDBlock = 'ddo.000001.Atlas.Ideal.DBRelease.v030101' fileD.lfn = 'DBRelease-4.7.1.tar.gz' fileD.type = 'input' job.addFile(fileD)
def 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
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
job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'), index) job.AtlasRelease = 'Atlas-14.1.0\nAtlas-14.1.0' job.homepackage = 'AtlasProduction/14.1.0.3\nAtlasProduction/14.1.0.3' job.transformation = 'csc_digi_trf.py\ncsc_reco_trf.py' job.destinationDBlock = datasetName job.computingSite = site job.prodDBlock = 'valid1.005200.T1_McAtNlo_Jimmy.simul.HITS.e322_s429_tid022081' job.prodSourceLabel = 'test' job.currentPriority = 10000 job.cloud = 'US' for lfn in ['HITS.022081._00001.pool.root', 'HITS.022081._00002.pool.root']: fileI = FileSpec() fileI.dataset = job.prodDBlock fileI.prodDBlock = job.prodDBlock fileI.lfn = lfn fileI.type = 'input' job.addFile(fileI) fileD1 = FileSpec() fileD1.dataset = 'ddo.000001.Atlas.Ideal.DBRelease.v050001' fileD1.prodDBlock = fileD1.dataset fileD1.lfn = 'DBRelease-5.0.1.tar.gz' fileD1.type = 'input' job.addFile(fileD1)
def createJobSpec(self, task, outdataset, job, jobset, jobdef, site, jobname, lfnhanger, allsites, jobid): """Create a spec for one job :arg TaskWorker.DataObject.Task task: the task to work on :arg str outdataset: the output dataset name where all the produced files will be placed :arg WMCore.DataStructs.Job job: the abstract job :arg int jobset: the PanDA jobset corresponding to the current task :arg int jobdef: the PanDA jobdef where to append the current jobs --- not used :arg str site: the borkered site where to run the jobs :arg str jobname: the job name :arg str lfnhanger: the random string to be added in the output file name :arg list str allsites: all possible sites where the job can potentially run :arg int jobid: incremental job number :return: the sepc object.""" pandajob = JobSpec() ## always setting a job definition ID pandajob.jobDefinitionID = jobdef if jobdef else -1 ## always setting a job set ID pandajob.jobsetID = jobset if jobset else -1 pandajob.jobName = jobname pandajob.prodUserID = task['tm_user_dn'] pandajob.destinationDBlock = outdataset pandajob.prodDBlock = task['tm_input_dataset'] pandajob.prodSourceLabel = 'user' pandajob.computingSite = site pandajob.cloud = getSite(pandajob.computingSite) pandajob.destinationSE = 'local' pandajob.transformation = task['tm_transformation'] ## need to initialize this pandajob.metadata = '' def outFileSpec(of=None, log=False): """Local routine to create an FileSpec for the an job output/log file :arg str of: output file base name :return: FileSpec object for the output file.""" outfile = FileSpec() if log: outfile.lfn = "job.log_%d_%s.tgz" % (jobid, lfnhanger) outfile.type = 'log' else: outfile.lfn = '%s_%d_%s%s' % (os.path.splitext(of)[0], jobid, lfnhanger, os.path.splitext(of)[1]) outfile.type = 'output' outfile.destinationDBlock = pandajob.destinationDBlock outfile.destinationSE = task['tm_asyncdest'] outfile.dataset = pandajob.destinationDBlock return outfile alloutfiles = [] outjobpar = {} outfilestring = '' for outputfile in task['tm_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn for outputfile in task['tm_tfile_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn for outputfile in task['tm_edm_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn outfilestring = outfilestring[:-1] infiles = [] for inputfile in job['input_files']: infiles.append(inputfile['lfn']) pandajob.jobParameters = '-a %s ' % task['tm_user_sandbox'] pandajob.jobParameters += '--sourceURL %s ' % task['tm_cache_url'] pandajob.jobParameters += '--jobNumber=%s ' % jobid pandajob.jobParameters += '--cmsswVersion=%s ' % task['tm_job_sw'] pandajob.jobParameters += '--scramArch=%s ' % task['tm_job_arch'] pandajob.jobParameters += '--inputFile=\'%s\' ' % json.dumps(infiles) self.jobParametersSetting(pandajob, job, self.jobtypeMapper[task['tm_job_type']]) pandajob.jobParameters += '-o "%s" ' % str(outjobpar) pandajob.jobParameters += '--dbs_url=%s ' % task['tm_dbs_url'] pandajob.jobParameters += '--publish_dbs_url=%s ' % task[ 'tm_publish_dbs_url'] pandajob.jobParameters += '--publishFiles=%s ' % ( 'True' if task['tm_publication'] == 'T' else 'False') pandajob.jobParameters += '--saveLogs=%s ' % ( 'True' if task['tm_save_logs'] == 'T' else 'False') pandajob.jobParameters += '--availableSites=\'%s\' ' % json.dumps( allsites) pandajob.jobParameters += '--group=%s ' % ( task['tm_user_group'] if task['tm_user_group'] else '') pandajob.jobParameters += '--role=%s ' % (task['tm_user_role'] if task['tm_user_role'] else '') self.logger.info(type(task['tm_user_infiles'])) self.logger.info(task['tm_user_infiles']) if task['tm_user_infiles']: addinfilestring = '' for addinfile in task['tm_user_infiles']: addinfilestring += '%s,' % addinfile pandajob.jobParameters += '--userFiles=%s ' % ( addinfilestring[:-1]) pandajob.jobName = '%s' % task[ 'tm_taskname'] #Needed by ASO and Dashboard if 'panda_oldjobid' in job and job['panda_oldjobid']: pandajob.parentID = job['panda_oldjobid'] pandajob.addFile(outFileSpec(log=True)) for filetoadd in alloutfiles: pandajob.addFile(filetoadd) return pandajob
for lfn in files.keys(): index += 1 job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'),index) job.AtlasRelease = 'Atlas-13.0.40' job.homepackage = 'AtlasProduction/13.0.40.3' job.transformation = 'csc_simul_trf.py' job.destinationDBlock = datasetName job.destinationSE = destName job.computingSite = site job.prodDBlock = 'valid1.005001.pythia_minbias.evgen.EVNT.e306_tid019128' job.prodSourceLabel = 'test' job.currentPriority = 10000 job.cloud = 'IT' fileI = FileSpec() fileI.dataset = job.prodDBlock fileI.prodDBlock = job.prodDBlock fileI.lfn = lfn fileI.type = 'input' job.addFile(fileI) fileD = FileSpec() fileD.dataset = 'ddo.000001.Atlas.Ideal.DBRelease.v040701' fileD.prodDBlock = 'ddo.000001.Atlas.Ideal.DBRelease.v030101' fileD.lfn = 'DBRelease-4.7.1.tar.gz' fileD.type = 'input' job.addFile(fileD)
def send_job(jobid, siteid): _logger.debug('Jobid: ' + str(jobid)) site = sites_.get(siteid) job = jobs_.get(int(jobid)) cont = job.container files_catalog = cont.files fscope = getScope(job.owner.username) datasetName = '{}:{}'.format(fscope, cont.guid) distributive = job.distr.name release = job.distr.release # Prepare runScript parameters = job.distr.command parameters = parameters.replace("$COMMAND$", job.params) parameters = parameters.replace("$USERNAME$", job.owner.username) parameters = parameters.replace("$WORKINGGROUP$", job.owner.working_group) # Prepare metadata metadata = dict(user=job.owner.username) # Prepare PanDA Object pandajob = JobSpec() pandajob.jobDefinitionID = int(time.time()) % 10000 pandajob.jobName = cont.guid pandajob.transformation = client_config.DEFAULT_TRF pandajob.destinationDBlock = datasetName pandajob.destinationSE = site.se pandajob.currentPriority = 1000 pandajob.prodSourceLabel = 'user' pandajob.computingSite = site.ce pandajob.cloud = 'RU' pandajob.VO = 'atlas' pandajob.prodDBlock = "%s:%s" % (fscope, pandajob.jobName) pandajob.coreCount = job.corecount pandajob.metadata = json.dumps(metadata) #pandajob.workingGroup = job.owner.working_group if site.encode_commands: # It requires script wrapper on cluster side pandajob.jobParameters = '%s %s %s "%s"' % (cont.guid, release, distributive, parameters) else: pandajob.jobParameters = parameters has_input = False for fcc in files_catalog: if fcc.type == 'input': f = fcc.file guid = f.guid fileIT = FileSpec() fileIT.lfn = f.lfn fileIT.dataset = pandajob.prodDBlock fileIT.prodDBlock = pandajob.prodDBlock fileIT.type = 'input' fileIT.scope = fscope fileIT.status = 'ready' fileIT.GUID = guid pandajob.addFile(fileIT) has_input = True if fcc.type == 'output': f = fcc.file fileOT = FileSpec() fileOT.lfn = f.lfn fileOT.destinationDBlock = pandajob.prodDBlock fileOT.destinationSE = pandajob.destinationSE fileOT.dataset = pandajob.prodDBlock fileOT.type = 'output' fileOT.scope = fscope fileOT.GUID = f.guid pandajob.addFile(fileOT) # Save replica meta fc.new_replica(f, site) if not has_input: # Add fake input fileIT = FileSpec() fileIT.lfn = "fake.input" fileIT.dataset = pandajob.prodDBlock fileIT.prodDBlock = pandajob.prodDBlock fileIT.type = 'input' fileIT.scope = fscope fileIT.status = 'ready' fileIT.GUID = "fake.guid" pandajob.addFile(fileIT) # Prepare lof file fileOL = FileSpec() fileOL.lfn = "%s.log.tgz" % pandajob.jobName fileOL.destinationDBlock = pandajob.destinationDBlock fileOL.destinationSE = pandajob.destinationSE fileOL.dataset = '{}:logs'.format(fscope) fileOL.type = 'log' fileOL.scope = 'panda' pandajob.addFile(fileOL) # Save log meta log = File() log.scope = fscope log.lfn = fileOL.lfn log.guid = getGUID(log.scope, log.lfn) log.type = 'log' log.status = 'defined' files_.save(log) # Save replica meta fc.new_replica(log, site) # Register file in container fc.reg_file_in_cont(log, cont, 'log') # Submit job o = submitJobs([pandajob]) x = o[0] try: #update PandaID PandaID = int(x[0]) job.pandaid = PandaID job.ce = site.ce except: job.status = 'submit_error' jobs_.save(job) return 0
index = 0 for lfn in files.keys(): index += 1 job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'),index) job.AtlasRelease = 'Atlas-12.0.6' job.homepackage = 'AtlasProduction/12.0.6.4' job.transformation = 'csc_reco_trf.py' job.destinationDBlock = datasetName job.destinationSE = destName job.computingSite = site #job.prodDBlock = 'misal1_mc12.005200.T1_McAtNlo_Jimmy.digit.RDO.v12000601_tid007554' job.prodDBlock = 'misal1_mc12.005802.JF17_pythia_jet_filter.digit.RDO.v12000601_tid008610' job.cloud = 'US' job.prodSourceLabel = 'test' job.currentPriority = 10000 job.cmtConfig = 'i686-slc4-gcc34-opt' fileI = FileSpec() fileI.dataset = job.prodDBlock fileI.prodDBlock = job.prodDBlock fileI.lfn = lfn fileI.type = 'input' job.addFile(fileI) fileD = FileSpec() fileD.dataset = 'ddo.000001.Atlas.Ideal.DBRelease.v030101' fileD.prodDBlock = 'ddo.000001.Atlas.Ideal.DBRelease.v030101'
def prepare(self,app,appconfig,appmasterconfig,jobmasterconfig): '''prepare the subjob specific configuration''' # PandaTools from pandatools import Client from pandatools import AthenaUtils from taskbuffer.JobSpec import JobSpec from taskbuffer.FileSpec import FileSpec job = app._getParent() logger.debug('AthenaMCPandaRTHandler prepare called for %s', job.getFQID('.')) try: assert self.outsite except: logger.error("outsite not set. Aborting") raise Exception() job.backend.site = self.outsite job.backend.actualCE = self.outsite cloud = job._getRoot().backend.requirements.cloud job.backend.requirements.cloud = cloud # now just filling the job from AthenaMC data jspec = JobSpec() jspec.jobDefinitionID = job._getRoot().id jspec.jobName = commands.getoutput('uuidgen 2> /dev/null') jspec.AtlasRelease = 'Atlas-%s' % app.atlas_rel if app.transform_archive: jspec.homepackage = 'AnalysisTransforms'+app.transform_archive elif app.prod_release: jspec.homepackage = 'AnalysisTransforms-AtlasProduction_'+str(app.prod_release) jspec.transformation = '%s/runAthena-00-00-11' % Client.baseURLSUB #---->???? prodDBlock and destinationDBlock when facing several input / output datasets? jspec.prodDBlock = 'NULL' if job.inputdata and len(app.inputfiles)>0 and app.inputfiles[0] in app.dsetmap: jspec.prodDBlock = app.dsetmap[app.inputfiles[0]] # How to specify jspec.destinationDBlock when more than one type of output is available? Panda prod jobs seem to specify only the last output dataset outdset="" for type in ["EVNT","RDO","HITS","AOD","ESD","NTUP"]: if type in app.outputpaths.keys(): outdset=string.replace(app.outputpaths[type],"/",".") outdset=outdset[1:-1] break if not outdset: try: assert len(app.outputpaths.keys())>0 except: logger.error("app.outputpaths is empty: check your output datasets") raise type=app.outputpaths.keys()[0] outdset=string.replace(app.outputpaths[type],"/",".") outdset=outdset[1:-1] jspec.destinationDBlock = outdset jspec.destinationSE = self.outsite jspec.prodSourceLabel = 'user' jspec.assignedPriority = 1000 jspec.cloud = cloud # memory if job.backend.requirements.memory != -1: jspec.minRamCount = job.backend.requirements.memory jspec.computingSite = self.outsite jspec.cmtConfig = AthenaUtils.getCmtConfig(athenaVer=app.atlas_rel) # library (source files) flib = FileSpec() flib.lfn = self.library # flib.GUID = flib.type = 'input' # flib.status = flib.dataset = self.libDataset flib.dispatchDBlock = self.libDataset jspec.addFile(flib) # input files FIXME: many more input types for lfn in app.inputfiles: useguid=app.turls[lfn].replace("guid:","") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # add dbfiles if any: for lfn in app.dbfiles: useguid=app.dbturls[lfn].replace("guid:","") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # then minbias files for lfn in app.mbfiles: useguid=app.minbias_turls[lfn].replace("guid:","") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # then cavern files for lfn in app.cavernfiles: useguid=app.cavern_turls[lfn].replace("guid:","") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # output files( this includes the logfiles) # Output files jidtag="" job = app._getParent() # Returns job or subjob object if job._getRoot().subjobs: jidtag = job._getRoot().id else: jidtag = "%d" % job.id outfiles=app.subjobsOutfiles[job.id] pandaOutfiles={} for type in outfiles.keys(): pandaOutfiles[type]=outfiles[type]+"."+str(jidtag) if type=="LOG": pandaOutfiles[type]+=".tgz" #print pandaOutfiles for outtype in pandaOutfiles.keys(): fout = FileSpec() dset=string.replace(app.outputpaths[outtype],"/",".") dset=dset[1:-1] fout.dataset=dset fout.lfn=pandaOutfiles[outtype] fout.type = 'output' # fout.destinationDBlock = jspec.destinationDBlock fout.destinationDBlock = fout.dataset fout.destinationSE = jspec.destinationSE if outtype=='LOG': fout.type='log' fout.destinationDBlock = fout.dataset fout.destinationSE = job.backend.site jspec.addFile(fout) # job parameters param = '-l %s ' % self.library # user tarball. # use corruption checker if job.backend.requirements.corCheck: param += '--corCheck ' # disable to skip missing files if job.backend.requirements.notSkipMissing: param += '--notSkipMissing ' # transform parameters # need to update arglist with final output file name... newArgs=[] if app.mode == "evgen": app.args[3]=app.args[3]+" -t " if app.verbosity: app.args[3]=app.args[3]+" -l %s " % app.verbosity for arg in app.args[3:]: for type in outfiles.keys(): if arg.find(outfiles[type])>-1: arg=arg.replace(outfiles[type],pandaOutfiles[type]) newArgs.append(arg) arglist=string.join(newArgs," ") # print "Arglist:",arglist param += ' -r ./ ' param += ' -j "%s"' % urllib.quote(arglist) allinfiles=app.inputfiles+app.dbfiles # Input files. param += ' -i "%s" ' % allinfiles if len(app.mbfiles)>0: param+= ' -m "%s" ' % app.mbfiles if len(app.cavernfiles)>0: param+= ' -n "%s" ' % app.cavernfiles # param += '-m "[]" ' #%minList FIXME # param += '-n "[]" ' #%cavList FIXME del pandaOutfiles["LOG"] # logfiles do not appear in IROOT block, and this one is not needed anymore... param += ' -o "{\'IROOT\':%s }"' % str(pandaOutfiles.items()) # source URL matchURL = re.search("(http.*://[^/]+)/",Client.baseURLSSL) if matchURL != None: param += " --sourceURL %s " % matchURL.group(1) param += " --trf" jspec.jobParameters = param jspec.metadata="--trf \"%s\"" % arglist #print "SUBJOB DETAILS:",jspec.values() if app.dryrun: print "job.application.dryrun activated, printing out job parameters" print jspec.values() return return jspec
def createJobSpec(self, task, outdataset, job, jobset, jobdef, site, jobname, lfnhanger, allsites, jobid): """Create a spec for one job :arg TaskWorker.DataObject.Task task: the task to work on :arg str outdataset: the output dataset name where all the produced files will be placed :arg WMCore.DataStructs.Job job: the abstract job :arg int jobset: the PanDA jobset corresponding to the current task :arg int jobdef: the PanDA jobdef where to append the current jobs --- not used :arg str site: the borkered site where to run the jobs :arg str jobname: the job name :arg str lfnhanger: the random string to be added in the output file name :arg list str allsites: all possible sites where the job can potentially run :arg int jobid: incremental job number :return: the sepc object.""" pandajob = JobSpec() ## always setting a job definition ID pandajob.jobDefinitionID = jobdef if jobdef else -1 ## always setting a job set ID pandajob.jobsetID = jobset if jobset else -1 pandajob.jobName = jobname pandajob.prodUserID = task['tm_user_dn'] pandajob.destinationDBlock = outdataset pandajob.prodDBlock = task['tm_input_dataset'] pandajob.prodSourceLabel = 'user' pandajob.computingSite = site pandajob.cloud = getSite(pandajob.computingSite) pandajob.destinationSE = 'local' pandajob.transformation = task['tm_transformation'] ## need to initialize this pandajob.metadata = '' def outFileSpec(of=None, log=False): """Local routine to create an FileSpec for the an job output/log file :arg str of: output file base name :return: FileSpec object for the output file.""" outfile = FileSpec() if log: outfile.lfn = "job.log_%d_%s.tgz" % (jobid, lfnhanger) outfile.type = 'log' else: outfile.lfn = '%s_%d_%s%s' %(os.path.splitext(of)[0], jobid, lfnhanger, os.path.splitext(of)[1]) outfile.type = 'output' outfile.destinationDBlock = pandajob.destinationDBlock outfile.destinationSE = task['tm_asyncdest'] outfile.dataset = pandajob.destinationDBlock return outfile alloutfiles = [] outjobpar = {} outfilestring = '' for outputfile in task['tm_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn for outputfile in task['tm_tfile_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn for outputfile in task['tm_edm_outfiles']: outfilestring += '%s,' % outputfile filespec = outFileSpec(outputfile) alloutfiles.append(filespec) #pandajob.addFile(filespec) outjobpar[outputfile] = filespec.lfn outfilestring = outfilestring[:-1] infiles = [] for inputfile in job['input_files']: infiles.append( inputfile['lfn'] ) pandajob.jobParameters = '-a %s ' % task['tm_user_sandbox'] pandajob.jobParameters += '--sourceURL %s ' % task['tm_cache_url'] pandajob.jobParameters += '--jobNumber=%s ' % jobid pandajob.jobParameters += '--cmsswVersion=%s ' % task['tm_job_sw'] pandajob.jobParameters += '--scramArch=%s ' % task['tm_job_arch'] pandajob.jobParameters += '--inputFile=\'%s\' ' % json.dumps(infiles) self.jobParametersSetting(pandajob, job, self.jobtypeMapper[task['tm_job_type']]) pandajob.jobParameters += '-o "%s" ' % str(outjobpar) pandajob.jobParameters += '--dbs_url=%s ' % task['tm_dbs_url'] pandajob.jobParameters += '--publish_dbs_url=%s ' % task['tm_publish_dbs_url'] pandajob.jobParameters += '--publishFiles=%s ' % ('True' if task['tm_publication'] == 'T' else 'False') pandajob.jobParameters += '--saveLogs=%s ' % ('True' if task['tm_save_logs'] == 'T' else 'False') pandajob.jobParameters += '--availableSites=\'%s\' ' %json.dumps(allsites) pandajob.jobParameters += '--group=%s ' % (task['tm_user_group'] if task['tm_user_group'] else '') pandajob.jobParameters += '--role=%s ' % (task['tm_user_role'] if task['tm_user_role'] else '') self.logger.info(type(task['tm_user_infiles'])) self.logger.info(task['tm_user_infiles']) if task['tm_user_infiles']: addinfilestring = '' for addinfile in task['tm_user_infiles']: addinfilestring += '%s,' % addinfile pandajob.jobParameters += '--userFiles=%s ' % ( addinfilestring[:-1] ) pandajob.jobName = '%s' % task['tm_taskname'] #Needed by ASO and Dashboard if 'panda_oldjobid' in job and job['panda_oldjobid']: pandajob.parentID = job['panda_oldjobid'] pandajob.addFile(outFileSpec(log=True)) for filetoadd in alloutfiles: pandajob.addFile(filetoadd) return pandajob
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
jobList = [] for i in range(1): job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'), i) job.AtlasRelease = 'Atlas-17.0.5' job.homepackage = 'AtlasProduction/17.0.5.6' job.transformation = 'Evgen_trf.py' job.destinationDBlock = datasetName job.destinationSE = destName job.currentPriority = 10000 job.prodSourceLabel = 'test' job.computingSite = site job.cloud = cloud job.cmtConfig = 'i686-slc5-gcc43-opt' file = FileSpec() file.lfn = "%s.evgen.pool.root" % job.jobName file.destinationDBlock = job.destinationDBlock file.destinationSE = job.destinationSE file.dataset = job.destinationDBlock file.destinationDBlockToken = 'ATLASDATADISK' file.type = 'output' job.addFile(file) fileOL = FileSpec() fileOL.lfn = "%s.job.log.tgz" % job.jobName fileOL.destinationDBlock = job.destinationDBlock fileOL.destinationSE = job.destinationSE
def prepare(self,app,appsubconfig,appmasterconfig,jobmasterconfig): '''prepare the subjob specific configuration''' from pandatools import Client from taskbuffer.JobSpec import JobSpec from taskbuffer.FileSpec import FileSpec job = app._getParent() logger.debug('AthenaPandaRTHandler prepare called for %s', job.getFQID('.')) # in case of a simple job get the dataset content, otherwise subjobs are filled by the splitter if job.inputdata and not job._getRoot().subjobs: if not job.inputdata.names: contents = job.inputdata.get_contents(overlap=False, size=True) for ds in contents.keys(): for f in contents[ds]: job.inputdata.guids.append( f[0] ) job.inputdata.names.append( f[1][0] ) job.inputdata.sizes.append( f[1][1] ) job.inputdata.checksums.append( f[1][2] ) job.inputdata.scopes.append( f[1][3] ) site = job._getRoot().backend.site job.backend.site = site job.backend.actualCE = site cloud = job._getRoot().backend.requirements.cloud job.backend.requirements.cloud = cloud # if no outputdata are given if not job.outputdata: job.outputdata = DQ2OutputDataset() job.outputdata.datasetname = job._getRoot().outputdata.datasetname #if not job.outputdata.datasetname: else: job.outputdata.datasetname = job._getRoot().outputdata.datasetname if not job.outputdata.datasetname: raise ApplicationConfigurationError('DQ2OutputDataset has no datasetname') jspec = JobSpec() jspec.jobDefinitionID = job._getRoot().id jspec.jobName = commands.getoutput('uuidgen 2> /dev/null') jspec.transformation = '%s/runGen-00-00-02' % Client.baseURLSUB if job.inputdata: jspec.prodDBlock = job.inputdata.dataset[0] else: jspec.prodDBlock = 'NULL' jspec.destinationDBlock = job.outputdata.datasetname if job.outputdata.location: if not job._getRoot().subjobs or job.id == 0: logger.warning('You have specified outputdata.location. Note that Panda may not support writing to a user-defined output location.') jspec.destinationSE = job.outputdata.location else: jspec.destinationSE = site jspec.prodSourceLabel = configPanda['prodSourceLabelRun'] jspec.processingType = configPanda['processingType'] jspec.assignedPriority = configPanda['assignedPriorityRun'] jspec.cloud = cloud # memory if job.backend.requirements.memory != -1: jspec.minRamCount = job.backend.requirements.memory # cputime if job.backend.requirements.cputime != -1: jspec.maxCpuCount = job.backend.requirements.cputime jspec.computingSite = site # library (source files) if job.backend.libds: flib = FileSpec() flib.lfn = self.fileBO.lfn flib.GUID = self.fileBO.GUID flib.type = 'input' flib.status = self.fileBO.status flib.dataset = self.fileBO.destinationDBlock flib.dispatchDBlock = self.fileBO.destinationDBlock jspec.addFile(flib) elif job.backend.bexec: flib = FileSpec() flib.lfn = self.library flib.type = 'input' flib.dataset = self.libDataset flib.dispatchDBlock = self.libDataset jspec.addFile(flib) # input files FIXME: many more input types if job.inputdata: for guid, lfn, size, checksum, scope in zip(job.inputdata.guids,job.inputdata.names,job.inputdata.sizes, job.inputdata.checksums, job.inputdata.scopes): finp = FileSpec() finp.lfn = lfn finp.GUID = guid finp.scope = scope # finp.fsize = # finp.md5sum = finp.dataset = job.inputdata.dataset[0] finp.prodDBlock = job.inputdata.dataset[0] finp.dispatchDBlock = job.inputdata.dataset[0] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # output files # outMap = {} #FIXME: if options.outMeta != []: self.rundirectory = "." # log files flog = FileSpec() flog.lfn = '%s._$PANDAID.log.tgz' % job.outputdata.datasetname flog.type = 'log' flog.dataset = job.outputdata.datasetname flog.destinationDBlock = job.outputdata.datasetname flog.destinationSE = job.backend.site jspec.addFile(flog) # job parameters param = '' # source URL matchURL = re.search("(http.*://[^/]+)/",Client.baseURLCSRVSSL) srcURL = "" if matchURL != None: srcURL = matchURL.group(1) param += " --sourceURL %s " % srcURL param += '-r "%s" ' % self.rundirectory exe_name = job.application.exe if job.backend.bexec == '': if hasattr(job.application.exe, "name"): exe_name = os.path.basename(job.application.exe.name) # set jobO parameter if job.application.args: param += ' -j "" -p "%s %s" '%(exe_name,urllib.quote(" ".join(job.application.args))) else: param += ' -j "" -p "%s" '%exe_name if self.inputsandbox: param += ' -a %s '%self.inputsandbox else: param += '-l %s ' % self.library param += '-j "" -p "%s %s" ' % ( exe_name,urllib.quote(" ".join(job.application.args))) if job.inputdata: param += '-i "%s" ' % job.inputdata.names # fill outfiles outfiles = {} for f in self.extOutFile: tarnum = 1 if f.find('*') != -1: # archive * outfiles[f] = "outputbox%i.%s.%s.tar.gz" % (tarnum, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S") ) tarnum += 1 else: outfiles[f] = "%s.%s.%s" %(f, job.getFQID('.'), time.strftime("%Y%m%d%H%M%S")) fout = FileSpec() fout.lfn = outfiles[f] fout.type = 'output' fout.dataset = job.outputdata.datasetname fout.destinationDBlock = job.outputdata.datasetname fout.destinationSE = job.backend.site jspec.addFile(fout) param += '-o "%s" ' % (outfiles) # must be double quotes, because python prints strings in 'single quotes' for file in jspec.Files: if file.type in [ 'output', 'log'] and configPanda['chirpconfig']: file.dispatchDBlockToken = configPanda['chirpconfig'] logger.debug('chirp file %s',file) jspec.jobParameters = param return jspec
jobList = [] for i in range(1): job = JobSpec() job.jobDefinitionID = int(time.time()) % 10000 job.jobName = "%s_%d" % (commands.getoutput('uuidgen'),i) # job.AtlasRelease = 'Atlas-12.0.6' # job.homepackage = 'AtlasProduction/12.0.6.5' job.AtlasRelease = 'Atlas-12.0.7' job.homepackage = 'AtlasProduction/12.0.7.1' job.transformation = 'csc_evgen_trf.py' job.destinationDBlock = datasetName # job.destinationSE = destName # job.cloud = 'CA' job.cloud = cloud job.taskID = taskid job.currentPriority = 1000 job.prodSourceLabel = 'test' # job.prodSourceLabel = 'cloudtest' job.computingSite = site file = FileSpec() file.lfn = "%s.evgen.pool.root" % job.jobName file.destinationDBlock = job.destinationDBlock file.destinationSE = job.destinationSE file.dataset = job.destinationDBlock file.type = 'output' job.addFile(file) fileOL = FileSpec()
def prepare(self, app, appconfig, appmasterconfig, jobmasterconfig): '''prepare the subjob specific configuration''' # PandaTools from pandatools import Client from pandatools import AthenaUtils from taskbuffer.JobSpec import JobSpec from taskbuffer.FileSpec import FileSpec job = app._getParent() logger.debug('AthenaMCPandaRTHandler prepare called for %s', job.getFQID('.')) try: assert self.outsite except: logger.error("outsite not set. Aborting") raise Exception() job.backend.site = self.outsite job.backend.actualCE = self.outsite cloud = job._getRoot().backend.requirements.cloud job.backend.requirements.cloud = cloud # now just filling the job from AthenaMC data jspec = JobSpec() jspec.jobDefinitionID = job._getRoot().id jspec.jobName = commands.getoutput('uuidgen 2> /dev/null') jspec.AtlasRelease = 'Atlas-%s' % app.atlas_rel if app.transform_archive: jspec.homepackage = 'AnalysisTransforms' + app.transform_archive elif app.prod_release: jspec.homepackage = 'AnalysisTransforms-AtlasProduction_' + str( app.prod_release) jspec.transformation = '%s/runAthena-00-00-11' % Client.baseURLSUB #---->???? prodDBlock and destinationDBlock when facing several input / output datasets? jspec.prodDBlock = 'NULL' if job.inputdata and len( app.inputfiles) > 0 and app.inputfiles[0] in app.dsetmap: jspec.prodDBlock = app.dsetmap[app.inputfiles[0]] # How to specify jspec.destinationDBlock when more than one type of output is available? Panda prod jobs seem to specify only the last output dataset outdset = "" for type in ["EVNT", "RDO", "HITS", "AOD", "ESD", "NTUP"]: if type in app.outputpaths.keys(): outdset = string.replace(app.outputpaths[type], "/", ".") outdset = outdset[1:-1] break if not outdset: try: assert len(app.outputpaths.keys()) > 0 except: logger.error( "app.outputpaths is empty: check your output datasets") raise type = app.outputpaths.keys()[0] outdset = string.replace(app.outputpaths[type], "/", ".") outdset = outdset[1:-1] jspec.destinationDBlock = outdset jspec.destinationSE = self.outsite jspec.prodSourceLabel = 'user' jspec.assignedPriority = 1000 jspec.cloud = cloud # memory if job.backend.requirements.memory != -1: jspec.minRamCount = job.backend.requirements.memory jspec.computingSite = self.outsite jspec.cmtConfig = AthenaUtils.getCmtConfig(athenaVer=app.atlas_rel) # library (source files) flib = FileSpec() flib.lfn = self.library # flib.GUID = flib.type = 'input' # flib.status = flib.dataset = self.libDataset flib.dispatchDBlock = self.libDataset jspec.addFile(flib) # input files FIXME: many more input types for lfn in app.inputfiles: useguid = app.turls[lfn].replace("guid:", "") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # add dbfiles if any: for lfn in app.dbfiles: useguid = app.dbturls[lfn].replace("guid:", "") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # then minbias files for lfn in app.mbfiles: useguid = app.minbias_turls[lfn].replace("guid:", "") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # then cavern files for lfn in app.cavernfiles: useguid = app.cavern_turls[lfn].replace("guid:", "") finp = FileSpec() finp.lfn = lfn finp.GUID = useguid finp.dataset = app.dsetmap[lfn] finp.prodDBlock = app.dsetmap[lfn] finp.prodDBlockToken = 'local' finp.dispatchDBlock = app.dsetmap[lfn] finp.type = 'input' finp.status = 'ready' jspec.addFile(finp) # output files( this includes the logfiles) # Output files jidtag = "" job = app._getParent() # Returns job or subjob object if job._getRoot().subjobs: jidtag = job._getRoot().id else: jidtag = "%d" % job.id outfiles = app.subjobsOutfiles[job.id] pandaOutfiles = {} for type in outfiles.keys(): pandaOutfiles[type] = outfiles[type] + "." + str(jidtag) if type == "LOG": pandaOutfiles[type] += ".tgz" #print pandaOutfiles for outtype in pandaOutfiles.keys(): fout = FileSpec() dset = string.replace(app.outputpaths[outtype], "/", ".") dset = dset[1:-1] fout.dataset = dset fout.lfn = pandaOutfiles[outtype] fout.type = 'output' # fout.destinationDBlock = jspec.destinationDBlock fout.destinationDBlock = fout.dataset fout.destinationSE = jspec.destinationSE if outtype == 'LOG': fout.type = 'log' fout.destinationDBlock = fout.dataset fout.destinationSE = job.backend.site jspec.addFile(fout) # job parameters param = '-l %s ' % self.library # user tarball. # use corruption checker if job.backend.requirements.corCheck: param += '--corCheck ' # disable to skip missing files if job.backend.requirements.notSkipMissing: param += '--notSkipMissing ' # transform parameters # need to update arglist with final output file name... newArgs = [] if app.mode == "evgen": app.args[3] = app.args[3] + " -t " if app.verbosity: app.args[3] = app.args[3] + " -l %s " % app.verbosity for arg in app.args[3:]: for type in outfiles.keys(): if arg.find(outfiles[type]) > -1: arg = arg.replace(outfiles[type], pandaOutfiles[type]) newArgs.append(arg) arglist = string.join(newArgs, " ") # print "Arglist:",arglist param += ' -r ./ ' param += ' -j "%s"' % urllib.quote(arglist) allinfiles = app.inputfiles + app.dbfiles # Input files. param += ' -i "%s" ' % allinfiles if len(app.mbfiles) > 0: param += ' -m "%s" ' % app.mbfiles if len(app.cavernfiles) > 0: param += ' -n "%s" ' % app.cavernfiles # param += '-m "[]" ' #%minList FIXME # param += '-n "[]" ' #%cavList FIXME del pandaOutfiles[ "LOG"] # logfiles do not appear in IROOT block, and this one is not needed anymore... param += ' -o "{\'IROOT\':%s }"' % str(pandaOutfiles.items()) # source URL matchURL = re.search("(http.*://[^/]+)/", Client.baseURLSSL) if matchURL != None: param += " --sourceURL %s " % matchURL.group(1) param += " --trf" jspec.jobParameters = param jspec.metadata = "--trf \"%s\"" % arglist #print "SUBJOB DETAILS:",jspec.values() if app.dryrun: print "job.application.dryrun activated, printing out job parameters" print jspec.values() return return jspec