Python JobSpec.metadata方法代码示例

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]
 def doBrokerage(self,inputList,vo,prodSourceLabel,workQueue):
     # list with a lock
     inputListWorld = ListWithLock([])
     # variables for submission
     maxBunchTask = 100
     # make logger
     tmpLog = MsgWrapper(logger)
     tmpLog.debug('start doBrokerage')
     # return for failure
     retFatal    = self.SC_FATAL
     retTmpError = self.SC_FAILED
     tmpLog.debug('vo={0} label={1} queue={2} nTasks={3}'.format(vo,prodSourceLabel,
                                                                 workQueue.queue_name,
                                                                 len(inputList)))
     # loop over all tasks
     allRwMap    = {}
     prioMap     = {}
     tt2Map      = {}
     expRWs      = {}
     jobSpecList = []
     for tmpJediTaskID,tmpInputList in inputList:
         for taskSpec,cloudName,inputChunk in tmpInputList:
             # collect tasks for WORLD
             if taskSpec.useWorldCloud():
                 inputListWorld.append((taskSpec,inputChunk))
                 continue
             # make JobSpec to be submitted for TaskAssigner
             jobSpec = JobSpec()
             jobSpec.taskID     = taskSpec.jediTaskID
             jobSpec.jediTaskID = taskSpec.jediTaskID
             # set managed to trigger TA
             jobSpec.prodSourceLabel  = 'managed'
             jobSpec.processingType   = taskSpec.processingType
             jobSpec.workingGroup     = taskSpec.workingGroup
             jobSpec.metadata         = taskSpec.processingType
             jobSpec.assignedPriority = taskSpec.taskPriority
             jobSpec.currentPriority  = taskSpec.currentPriority
             jobSpec.maxDiskCount     = (taskSpec.getOutDiskSize() + taskSpec.getWorkDiskSize()) / 1024 / 1024
             if taskSpec.useWorldCloud():
                 # use destinationSE to trigger task brokerage in WORLD cloud
                 jobSpec.destinationSE = taskSpec.cloud
             prodDBlock = None
             setProdDBlock = False
             for datasetSpec in inputChunk.getDatasets():
                 prodDBlock = datasetSpec.datasetName
                 if datasetSpec.isMaster():
                     jobSpec.prodDBlock = datasetSpec.datasetName
                     setProdDBlock = True
                 for fileSpec in datasetSpec.Files:
                     tmpInFileSpec = fileSpec.convertToJobFileSpec(datasetSpec)
                     jobSpec.addFile(tmpInFileSpec)
             # use secondary dataset name as prodDBlock
             if setProdDBlock == False and prodDBlock != None:
                 jobSpec.prodDBlock = prodDBlock
             # append
             jobSpecList.append(jobSpec)
             prioMap[jobSpec.taskID] = jobSpec.currentPriority
             tt2Map[jobSpec.taskID]  = jobSpec.processingType
             # get RW for a priority
             if not allRwMap.has_key(jobSpec.currentPriority):
                 tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(vo,prodSourceLabel,workQueue,
                                                                    jobSpec.currentPriority) 
                 if tmpRW == None:
                     tmpLog.error('failed to calculate RW with prio={0}'.format(jobSpec.currentPriority))
                     return retTmpError
                 allRwMap[jobSpec.currentPriority] = tmpRW
             # get expected RW
             expRW = self.taskBufferIF.calculateTaskRW_JEDI(jobSpec.jediTaskID)
             if expRW == None:
                 tmpLog.error('failed to calculate RW for jediTaskID={0}'.format(jobSpec.jediTaskID))
                 return retTmpError
             expRWs[jobSpec.taskID] = expRW
     # for old clouds
     if jobSpecList != []:
         # get fullRWs
         fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(vo,prodSourceLabel,None,None)
         if fullRWs == None:
             tmpLog.error('failed to calculate full RW')
             return retTmpError
         # set metadata
         for jobSpec in jobSpecList:
             rwValues = allRwMap[jobSpec.currentPriority]
             jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (jobSpec.metadata,
                                                       str(rwValues),str(expRWs),
                                                       str(prioMap),str(fullRWs),
                                                       str(tt2Map))
         tmpLog.debug('run task assigner for {0} tasks'.format(len(jobSpecList)))
         nBunchTask = 0
         while nBunchTask < len(jobSpecList):
             # get a bunch
             jobsBunch = jobSpecList[nBunchTask:nBunchTask+maxBunchTask]
             strIDs = 'jediTaskID='
             for tmpJobSpec in jobsBunch:
                 strIDs += '{0},'.format(tmpJobSpec.taskID)
             strIDs = strIDs[:-1]
             tmpLog.debug(strIDs)
             # increment index
             nBunchTask += maxBunchTask
             # run task brokerge
             stS,outSs = PandaClient.runTaskAssignment(jobsBunch)
#.........这里部分代码省略.........

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]

#.........这里部分代码省略.........
            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

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]
 def doBrokerage(self, inputList, vo, prodSourceLabel, workQueue):
     # variables for submission
     maxBunchTask = 100
     # make logger
     tmpLog = MsgWrapper(logger)
     tmpLog.debug("start doBrokerage")
     # return for failure
     retFatal = self.SC_FATAL
     retTmpError = self.SC_FAILED
     tmpLog.debug("vo={0} label={1} queue={2}".format(vo, prodSourceLabel, workQueue.queue_name))
     # loop over all tasks
     allRwMap = {}
     prioMap = {}
     tt2Map = {}
     expRWs = {}
     jobSpecList = []
     for tmpJediTaskID, tmpInputList in inputList:
         for taskSpec, cloudName, inputChunk in tmpInputList:
             # make JobSpec to be submitted for TaskAssigner
             jobSpec = JobSpec()
             jobSpec.taskID = taskSpec.jediTaskID
             jobSpec.jediTaskID = taskSpec.jediTaskID
             # set managed to trigger TA
             jobSpec.prodSourceLabel = "managed"
             jobSpec.processingType = taskSpec.processingType
             jobSpec.workingGroup = taskSpec.workingGroup
             jobSpec.metadata = taskSpec.processingType
             jobSpec.assignedPriority = taskSpec.taskPriority
             jobSpec.currentPriority = taskSpec.currentPriority
             jobSpec.maxDiskCount = (taskSpec.getOutDiskSize() + taskSpec.getWorkDiskSize()) / 1024 / 1024
             if taskSpec.useWorldCloud():
                 # use destinationSE to trigger task brokerage in WORLD cloud
                 jobSpec.destinationSE = taskSpec.cloud
             prodDBlock = None
             setProdDBlock = False
             for datasetSpec in inputChunk.getDatasets():
                 prodDBlock = datasetSpec.datasetName
                 if datasetSpec.isMaster():
                     jobSpec.prodDBlock = datasetSpec.datasetName
                     setProdDBlock = True
                 for fileSpec in datasetSpec.Files:
                     tmpInFileSpec = fileSpec.convertToJobFileSpec(datasetSpec)
                     jobSpec.addFile(tmpInFileSpec)
             # use secondary dataset name as prodDBlock
             if setProdDBlock == False and prodDBlock != None:
                 jobSpec.prodDBlock = prodDBlock
             # append
             jobSpecList.append(jobSpec)
             prioMap[jobSpec.taskID] = jobSpec.currentPriority
             tt2Map[jobSpec.taskID] = jobSpec.processingType
             # get RW for a priority
             if not allRwMap.has_key(jobSpec.currentPriority):
                 tmpRW = self.taskBufferIF.calculateRWwithPrio_JEDI(
                     vo, prodSourceLabel, workQueue, jobSpec.currentPriority
                 )
                 if tmpRW == None:
                     tmpLog.error("failed to calculate RW with prio={0}".format(jobSpec.currentPriority))
                     return retTmpError
                 allRwMap[jobSpec.currentPriority] = tmpRW
             # get expected RW
             expRW = self.taskBufferIF.calculateTaskRW_JEDI(jobSpec.jediTaskID)
             if expRW == None:
                 tmpLog.error("failed to calculate RW for jediTaskID={0}".format(jobSpec.jediTaskID))
                 return retTmpError
             expRWs[jobSpec.taskID] = expRW
     # get fullRWs
     fullRWs = self.taskBufferIF.calculateRWwithPrio_JEDI(vo, prodSourceLabel, None, None)
     if fullRWs == None:
         tmpLog.error("failed to calculate full RW")
         return retTmpError
     # set metadata
     for jobSpec in jobSpecList:
         rwValues = allRwMap[jobSpec.currentPriority]
         jobSpec.metadata = "%s;%s;%s;%s;%s;%s" % (
             jobSpec.metadata,
             str(rwValues),
             str(expRWs),
             str(prioMap),
             str(fullRWs),
             str(tt2Map),
         )
     tmpLog.debug("run task assigner for {0} tasks".format(len(jobSpecList)))
     nBunchTask = 0
     while nBunchTask < len(jobSpecList):
         # get a bunch
         jobsBunch = jobSpecList[nBunchTask : nBunchTask + maxBunchTask]
         strIDs = "jediTaskID="
         for tmpJobSpec in jobsBunch:
             strIDs += "{0},".format(tmpJobSpec.taskID)
         strIDs = strIDs[:-1]
         tmpLog.debug(strIDs)
         # increment index
         nBunchTask += maxBunchTask
         # run task brokerge
         stS, outSs = PandaClient.runTaskAssignment(jobsBunch)
         tmpLog.debug("{0}:{1}".format(stS, str(outSs)))
     # return
     tmpLog.debug("done")
     return self.SC_SUCCEEDED

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]
    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 = ''
#.........这里部分代码省略.........

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]
for i in [999905,999906,999907]:
    job = JobSpec()
    job.jobDefinitionID   = int(time.time()) % 10000
    job.jobName           = "%s_%d" % (commands.getoutput('uuidgen'),i)
    job.AtlasRelease      = 'Atlas-14.1.0'
    job.homepackage       = 'AtlasProduction/12.0.6.2'
    job.transformation    = 'csc_evgen_trf.py'
    job.destinationDBlock = datasetName
    #job.destinationSE     = destName
    job.currentPriority   = 1000
    job.prodSourceLabel   = 'managed'
    #job.prodSourceLabel   = 'test'
    #job.computingSite     = site
    job.cmtConfig         = 'i686-slc4-gcc34-opt'
    job.metadata          = 'evgen;%s;%s;%s' % (str({'FR': 46, 'NL': 45, 'NDGF': 300, 'CERN': 19, 'TW': 44110, 'CA': 2922, 'DE': 9903, 'IT': 1168, 'US': 6226, 'UK': 1026, 'ES': 26619}),str({999907:100,999906:200,999905:300}),str({999905:100,999906:910,999907:500}))
    #job.metadata          = 'evgen;%s' % str({'FR': 46, 'NL': 45, 'NDGF': 300, 'CERN': 19, 'TW': 44110, 'CA': 2922, 'DE': 9903, 'IT': 1168, 'US': 6226, 'UK': 1026, 'ES': 26619})

    #job.cloud = "UK"
    job.taskID = i
    
    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()

# 需要导入模块: from taskbuffer.JobSpec import JobSpec [as 别名]
# 或者: from taskbuffer.JobSpec.JobSpec import metadata [as 别名]
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)
#.........这里部分代码省略.........