Experiment.py

# Class definition:
#   Experiment
#   This class is the main experiment class; ATLAS etc will inherit from this class
#   Instances are generated with ExperimentFactory
#   Subclasses should implement all needed methods prototyped in this class
#   Note: not compatible with Singleton Design Pattern due to the subclassing

import os
import re
import time
import commands
from subprocess import Popen, PIPE

from PilotErrors import PilotErrors
from pUtil import tolog                    # Dump to pilot log
from pUtil import readpar                  # Used to read values from the schedconfig DB (queuedata)
from pUtil import getCmtconfig             # cmtconfig (move to subclass)
from pUtil import getDirectAccessDic       # Get the direct access dictionary
from pUtil import isBuildJob               # Is the current job a build job?
from pUtil import remove                   # Used to remove redundant file before log file creation
from pUtil import getPilotlogFilename      # Used in the subprocess arguments method
from pUtil import extractHPCInfo           # Used by getSubprocessName() to determine HPC plug-in if necessary

class Experiment(object):

#    experiment = "generic"               # String defining the experiment

    # private data members
    __experiment = "generic"               # String defining the experiment
    __instance = None                      # Boolean used by subclasses to become a Singleton
    __error = PilotErrors()                # PilotErrors object
    __doFileLookups = False                # True for LFC based file lookups (basically a dummy data member here since singleton object is static)
    __cache = ""                           # Cache URL used e.g. by LSST

    # Required methods

    def __init__(self, *args, **kwargs):
        """ Default initialization """

        # e.g. self.__errorLabel = errorLabel
#        self.experiment = kwargs.get('experiment')
        pass

    def getExperiment(self):
        """ Return a string with the experiment name """

#        return self.experiment
        return self.__experiment

    def getJobExecutionCommand(self):
        """ Define and test the command(s) that will be used to execute the payload """
        # E.g. cmd = "source <path>/setup.sh; <path>/python <script>"

        cmd = ""

        return cmd

    def getFileLookups(self):
        """ Return the file lookup boolean """

        return self.__doFileLookups

    def doFileLookups(self, doFileLookups):
        """ Update the file lookups boolean """

        # Only implement this method if class really wants to update the __doFileLookups boolean
        # ATLAS wants to implement this, but not CMS
        # Method is used by Mover
        # self.__doFileLookups = doFileLookups
        pass

    def willDoAlternativeFileLookups(self):
        """ Should file lookups be done using alternative methods? """

        # E.g. in the migration period where LFC lookups are halted in favour of other methods in the Rucio API
        # (for ATLAS), this method could be useful. See the usage in Mover::getReplicaDictionary() which is called
        # after Experiment::willDoFileLookups() defined above. The motivation is that direct LFC calls are not to be
        # used any longer by the pilot, and in the migration period the actual LFC calls will be done in the Rucio
        # API. Eventually this API will switch to alternative file lookups.

        return False

    def willDoFileLookups(self):
        """ Should (LFC) file lookups be done by the pilot or not? """

        return self.__doFileLookups

    def willDoFileRegistration(self):
        """ Should (LFC) file registration be done by the pilot or not? """

        return False

    def getFileCatalog(self):
        """ Return the default file catalog to use (e.g. for replica lookups) """
        # See usage in Mover.py

        # e.g. 'lfc://prod-lfc-atlas.cern.ch:/grid/atlas'
        return ""

    # Additional optional methods
    # These methods are optional and can be left as they are here, or modified according to special needs

    def verifyProxy(self, envsetup="", limit=None):
        """ Check for a valid voms/grid proxy longer than N hours """
        # Use 'limit' to set required length

        tolog("(verifyProxy() is not implemented)")

        exitcode = 0
        pilotErrorDiag = ""

        return exitcode, pilotErrorDiag

    def removeRedundantFiles(self, workdir):
        """ Remove redundant files and directories """

        # List of files and directories to be removed from work directory prior to log file creation
        # Make sure that any large files or directories that are not wanted in the log file are included in this list
        dir_list = [
                    "buildJob*",
                    "external",
                    "fort.*",
                    "home",
                    "python",
                    "share",
                    "workdir",
                    "*.py",
                    "*.pyc",
                    "*.root*",
                    "JEM",
                    "tmp*",
                    "*.tmp",
                    "*.TMP",
                    "scratch",
                    ]

        for _dir in dir_list:
            files = glob(os.path.join(workdir, _dir))
            rc = remove(files)
            if not rc:
                tolog("IGNORE: Failed to remove redundant file(s): %s" % (files))

    def getPayloadName(self, job):
        """ Set a suitable name for the payload stdout """

        # The payload <name> gets translated into <name>_stdout.txt
        # which is the name of the stdout file produced by the payload execution
        # (essentially commands.getoutput("<setup>; <payload executable> [options] > <name>_stdout.txt"))

        # The job object can be used to create more precise stdout names (see e.g. the ATLASExperiment implementation)

        return "payload"

    def isOutOfMemory(self, **kwargs):
        """ Try to identify out of memory errors in the stderr/out """

        return False

    def getNumberOfEvents(self, **kwargs):
        """ Return the number of events """

        return 0

    def specialChecks(self, **kwargs):
        """ Implement special checks here """

        # Return False if fatal failure, otherwise return True
        # The pilot will abort if this method returns a False

        # On an HPC system, it might be good to skip certain checks (e.g. CVMFS, LFC, etc). Refer to schedconfig.resourcetype, set to 'hpc' on an HPC queue

        status = False

        tolog("No special checks for \'%s\'" % (self.experiment))

        return True # obviously change this to 'status' once implemented

    def checkSpecialEnvVars(self, sitename):
        """ Check special environment variables """

        ec = 0
        tolog("No special env var checks for site %s" % (sitename))

        return ec

    def setINDS(self, realDatasetsIn):
        """ Extract the dataset as set by pathena option --inDS and set the INDS environmental variable """
        # Needed by pathena (move to ATLASExperiment later)

        inDS = ""
        for ds in realDatasetsIn:
            if "DBRelease" not in ds and ".lib." not in ds:
                inDS = ds
                break
        if inDS != "":
            tolog("Setting INDS env variable to: %s" % (inDS))
            os.environ['INDS'] = inDS
        else:
            tolog("INDS unknown")

    def getValidBaseURLs(self, order=None):
        """ Return list of valid base URLs """
        # if order is defined, return given item first
        # e.g. order=http://atlpan.web.cern.ch/atlpan -> ['http://atlpan.web.cern.ch/atlpan', ...]

        validBaseURLs = []
        _validBaseURLs = ["http://www.usatlas.bnl.gov",\
                          "https://www.usatlas.bnl.gov",\
                          "http://pandaserver.cern.ch",\
                          "http://atlpan.web.cern.ch/atlpan",\
                          "https://atlpan.web.cern.ch/atlpan",\
                          "http://classis01.roma1.infn.it",\
                          "http://atlas-install.roma1.infn.it"]

        if order:
            validBaseURLs.append(order)
            for url in _validBaseURLs:
                if url != order:
                    validBaseURLs.append(url)
        else:
            validBaseURLs = _validBaseURLs

        tolog("getValidBaseURLs will return: %s" % str(validBaseURLs))
        return validBaseURLs

    def downloadTrf(self, wgetCommand, jobTrf):
        """ Download the trf """

        status = False
        pilotErrorDiag = ""
        cmd = "%s %s" % (wgetCommand, jobTrf)
        trial = 1
        max_trials = 3

        # try to download the trf a maximum of 3 times
        while trial <= max_trials:
            tolog("Executing command [Trial %d/%d]: %s" % (trial, max_trials, cmd))
            ec, rets = commands.getstatusoutput(cmd)
            if not rets:
                rets = "(None)"
            if ec != 0:
                # Analyze exit code / output
                from futil import check_syserr
                check_syserr(ec, rets)
                pilotErrorDiag = "wget command failed: %d, %s" % (ec, rets)
                tolog("!!WARNING!!3000!! %s" % (pilotErrorDiag))
                if trial == max_trials:
                    tolog("!!FAILED!!3000!! Could not download trf: %s" % (rets))
                    status = False
                    break
                else:
                    tolog("Will try again after 60s..")
                    from time import sleep
                    sleep(60)
            else:
                pilotErrorDiag = ""
                tolog("wget command returned: %s" % (rets))
                status = True
                break
            trial += 1

        return status, pilotErrorDiag

    def getAnalysisTrf(self, wgetCommand, origTRF, pilot_initdir):
        """ Get the trf to be used for analysis jobs """

        pilotErrorDiag = ""
        trfName = origTRF.split('/')[-1]
        tolog("trfName = %s" % (trfName))
        origBaseURL = ""

        # Copy trf from pilot init dir if distributed with pilot code
        fname = os.path.join(pilot_initdir, trfName)
        status = False
        if os.path.exists(fname):
            from shutil import copy2
            try:
                copy2(fname, os.getcwd())
            except Exception, e:
                tolog("!!WARNING!!2999!! Could not copy trf from pilot init dir: %s" % str(e))
            else:
                tolog("Copied trf (%s) from pilot init dir" % (fname))
                status = True

        # Download trf
        if not status:
            # verify the base URL
            for baseURL in self.getValidBaseURLs():
                if origTRF.startswith(baseURL):
                    origBaseURL = baseURL
                    break

            if origBaseURL == "":
                pilotErrorDiag = "Invalid base URL: %s" % (origTRF)
                return self.__error.ERR_TRFDOWNLOAD, pilotErrorDiag, ""
            else:
                tolog("Verified the trf base URL: %s" % (origBaseURL))

            # try to download from the required location, if not - switch to backup
            for baseURL in self.getValidBaseURLs(order=origBaseURL):
                trf = re.sub(origBaseURL, baseURL, origTRF)
                tolog("Attempting to download trf: %s" % (trf))
                status, pilotErrorDiag = self.downloadTrf(wgetCommand, trf)
                if status:
                    break

            if not status:
                return self.__error.ERR_TRFDOWNLOAD, pilotErrorDiag, ""

            tolog("Successfully downloaded trf")

        tolog("Changing permission of %s to 0755" % (trfName))
        try:
            os.chmod(trfName, 0755)
        except Exception, e:
            pilotErrorDiag = "Failed to chmod %s: %s" % (trfName, str(e))
            return self.__error.ERR_CHMODTRF, pilotErrorDiag, ""

        return 0, pilotErrorDiag, trfName

    def getAnalysisRunCommand(self, job, jobSite, trfName):
        """ Get the run command for analysis jobs """
        # The run command is used to setup up the user job transform

        ec = 0
        pilotErrorDiag = ""
        run_command = ""

        return ec, pilotErrorDiag, run_command

    def getFileTransferInfo(self, transferType, buildJob):
        """ Get all relevant fields related to file transfer """

        copysetup = readpar('copysetupin')

        # create the direct access dictionary
        fileTransferInfo = getDirectAccessDic(copysetup)

        # if copysetupin did not contain direct access info, try the copysetup instead
        if not fileTransferInfo:
            copysetup = readpar('copysetup')
            fileTransferInfo = getDirectAccessDic(copysetup)

        # should the copytool be used?
        useCopyTool = False
        useFileStager = False
        useDirectAccess = False
        oldPrefix = ""
        newPrefix = ""
        dInfo = None
        if fileTransferInfo:
            dInfo = True
            # no direct access / remote I/O, use standard copytool (copy-to-scratch)
            if fileTransferInfo['useCopyTool']:
                useCopyTool = True
            # do not set the LFC host for file stager
            if fileTransferInfo['useFileStager']:
                useFileStager = True
            if fileTransferInfo['directIn']:
                useDirectAccess = True

            oldPrefix = fileTransferInfo['oldPrefix']
            newPrefix = fileTransferInfo['newPrefix']

        # override settings for transferType direct
        if transferType == 'direct':
            useCopyTool = False
            useFileStager = False
            useDirectAccess = True

        # should pilot create TURL based PFC? (not done here, but setup needs to be aware of it)
        # if dInfo and useDirectAccess and oldPrefix == "" and newPrefix == "":
        if (transferType == 'direct' or (useFileStager and useDirectAccess)) and (oldPrefix == "" and newPrefix == "") and not buildJob:
#        if (transferType == 'direct' or (not useFileStager and useDirectAccess)) and (oldPrefix == "" and newPrefix == ""):
            usePFCTurl = True
        else:
            usePFCTurl = False

        # force usePFCTurl for all jobs
        if not buildJob and useDirectAccess:
            tolog("Forced usePFCTurl (reset old/newPrefix)")
            usePFCTurl = True
            oldPrefix = ""
            newPrefix = ""

        if os.environ.get("TestXRootD", 'False') == 'True':
            import re
            re.sub(r'\/xrootdsetup\.sh', '/xrootdsetup-dev.sh', copysetup)

        return dInfo, useCopyTool, useDirectAccess, useFileStager, oldPrefix, newPrefix, copysetup, usePFCTurl

    def getGuidsFromJobPars(self, jobPars, inputFiles, inFilesGuids):
        """ Extract the correct guid from the input file list """

        # the guids list is used for direct reading in an LFC environment
        # 1. extract input file list for direct reading from jobPars
        # 2. for each input file in this list, find the corresponding guid from the input file guid list
        # since jobPars is entered by a human, the order of the input files might not be the same

        guidList = []

        jobPars = jobPars.replace("'","")
        jobPars = jobPars.replace(", ",",")

        pattern = re.compile(r'\-i \"\[([A-Za-z0-9.,_-]+)\]\"')
        directReadingInputFiles = re.findall(pattern, jobPars)
        inFiles = []
        if directReadingInputFiles != []:
            inFiles = directReadingInputFiles[0].split(",")
        else:
            match = re.search("-i ([A-Za-z0-9.\[\],_-]+) ", jobPars)
            if match != None:
                compactInFiles = match.group(1)
                match = re.search('(.*)\[(.+)\](.*)\[(.+)\]', compactInFiles)
                if match != None:
                    inputFiles = []
                    head = match.group(1)
                    tail = match.group(3)
                    body = match.group(2).split(',')
                    attr = match.group(4).split(',')
                    for idx in range(len(body)):
                        lfn = '%s%s%s%s' % (head, body[idx], tail, attr[idx])
                        inputFiles.append(lfn)
                else:
                    inputFiles = [compactInFiles]

        if inFiles != []:
            for inFile in inFiles:
                # get the corresponding index from the inputFiles list, which has the same order as inFilesGuids
                try:
                    index = inputFiles.index(inFile)
                except Exception, e:
                    tolog("!!WARNING!!2999!! Exception caught: %s (direct reading will fail)" % str(e))
                else:
                    # add the corresponding guid to the list
                    guidList.append(inFilesGuids[index])

        return guidList

    def getMetadataForRegistration(self, guid):
        """ Return metadata for [LFC] file registration """

        # This method can insert special metadata into the metadata.xml file
        # E.g. it can add preliminary XML tags for info that will only be known
        # at a later time, such as "<metadata att_name="surl" att_value="%s-surltobeset"/>\n' % (guid)"
        # The <guid>-surltobeset will be replaced by the pilot by the appropriate value once it is known
        # Inputs:
        #   guid = file guid
        # Returns:
        #   metadata string
        # See e.g. the CMSExperiment implementation

        # The method is called from pUtil::PFCxml() during metadata file creation

        return ""

    def getAttrForRegistration(self):
        """ Return the attribute of the metadata XML to be updated with surl value """

        # Used in combination with Experiment::getMetadataForRegistration()
        # The attribute (default 'surl') will be copied into the metadata string used for pattern matching
        # E.g. re.compile('\<metadata att\_name\=\"%s\" att\_value\=\"([a-zA-Z0-9-]+)\-surltobeset\"\/\>' % (attribute))

        return 'surl'

    def getExpSpecificMetadata(self, job, workdir):
        """ Return experiment specific metadata """

        # Inputs:
        #   job = PanDA pilot job object (see Job class)
        #   workdir = relevant work directory where the metadata is located
        # Returns:
        #   metadata xml string
        # See e.g. implementation in CMSExperiment

        return ""

    def getFileCatalogHosts(self):
        """ Return a list of file catalog hosts """

        # The method is used in combination with federated xrootd (FAX).
        # In case FAX is allowed on a given site, the pilot might need to lookup
        # replica information in more than one LFC catalog. Normally a site has only
        # one LFC (as set in schedconfig.lfchost). Providing a list of hosts will increase
        # the probability that FAX will succeed
        # See e.g. ATLASExperiment implementation

        return []

    def verifySwbase(self, appdir):
        """ Confirm existence of appdir/swbase """

        # appdir/swbase is a queuedata parameter specifying the base location of physics analysis / release software
        # This method will simply verify that the corresponding directory exists
        #
        # Input:
        #   appdir = application/software/release directory (e.g. /cvmfs/atlas.cern.ch/repo/sw)
        # Return:
        #   error code (0 for success)

        return 0

    def interpretPayloadStdout(self, job, res, getstatusoutput_was_interrupted, current_job_number, runCommandList, failureCode):
        """ Payload error interpretation and handling """

        # NOTE: TODO, hide argument complexity with kwargs**

        # This method can be used to interpret special errors that only occur in actual payload stdout, e.g. memory errors that have
        # caused the payload to crash
        #
        # Inputs:
        #   job = PanDA pilot job object (see Job class)
        #   res =
        #   getstatusoutput_was_interrupted = True in case the payload execution command was aborted (e.g. keyboard CTRL-C)
        #   current_job_number = current job number, in case of multi-trf (ATLAS)
        #   runCommandList = list of payload execution commands (e.g. used by ATLAS to get to a setup file)
        #   failureCode = signal error code
        # Returns:
        #   Updated PanDA pilot job objectwith proper payload error information, if needed
        #
        # The following Job attributes can be updated here
        #   result = tuple of size 3 that contain the standard error info: result[0] = current job status (e.g. failed, finished, holding),
        #            result[1] = payload error code, result[2] = PanDA pilot error code
        #   pilotErrorDiag = error diagnostics (string of up to 256 characters that will appear on the PanDA monitor job web page for a failed job)
        #   exeError

        return job

    def getSubprocessName(self, eventService):
        """ Select which subprocess is to be run by the Monitor """

        # The default subprocess is RunJob (name='Normal', which performs payload setup, stage-in, payload execution and stage-out).
        # An alternative subprocess is the runEvent module which downloads events from an Event Server, executes a payload
        # and stages ou output files asynchronously as they are ready.
        # Note: send the entire job object to this method since there might be other subprocesses created at a later time which
        # will be identified by this method using some other job data member

        # Default subprocess name
        name = "RunJob"

        # Select alternative subprocess names for HPCs
        isHPC, _name = extractHPCInfo(readpar('catchall'))
        if isHPC:
            name = "RunJob" + _name # e.g. "RunJobTitan" is the proper subprocess name for the Titan plug-in

        # for es merge jobs
        if _name and _name.startswith("Hpc"):
            name = "RunJob"

        # Are we going to run an event service job?
        if eventService:
            tolog("Encountered an event service job")
            if isHPC:
                name = "RunJob%sEvent" % (_name)
            else:
                name = "RunJobEvent"

        tolog("Selected subprocess: %s" % (name))

        return name

    def getSubprocessArguments(self, env, port, subprocessName="RunJob"):
        """ Argument list needed to launch the subprocess by the pilot/Monitor """

        # The pilot/Monitor is forking a subprocess which will be monitored for work dir size, hanging processes etc
        # This method returns the arguments needed to execute the subprocess (python <subprocess name> <arguments>)
        # By default the pilot has implementations for RunJob.py (standard job) and RunJobEvent.py (event server job)
        # If a new subprocess module is added, it startup arguments need to be specified here

        jobargs = None
        tolog("Will set up subprocess arguments for type: %s" % (subprocessName))

        url = '%s:%s/server/panda' % (env['pshttpurl'], str(env['psport']))

        if subprocessName == "RunJobEvent":
            jobargs = [env['pyexe'], "RunJobEvent.py",
                       "-a", env['thisSite'].appdir,
                       "-b", env['queuename'],
                       "-d", env['jobDic']["prod"][1].workdir,
                       "-g", env['inputDir'],
                       "-i", env['jobDic']["prod"][1].tarFileGuid,
                       "-k", getPilotlogFilename(),
                       "-l", env['pilot_initdir'],
                       "-m", env['outputDir'],
                       "-o", env['thisSite'].workdir,
                       "-p", str(port),
                       "-s", env['thisSite'].sitename,
                       "-t", str(env['proxycheckFlag']),
                       "-x", str(env['stageinretry']),
                       "-E", str(env['stageoutretry']),
                       "-F", env['experiment'],
                       "-H", env['cache'],
                       "-W", url]
        else:
            jobargs = [env['pyexe'], "%s.py" % (subprocessName),
                       "-a", env['thisSite'].appdir,
                       "-b", env['queuename'],
                       "-d", env['jobDic']["prod"][1].workdir,
                       "-g", env['inputDir'],
                       "-i", env['jobDic']["prod"][1].tarFileGuid,
                       "-k", getPilotlogFilename(),
                       "-l", env['pilot_initdir'],
                       "-m", env['outputDir'],
                       "-o", env['thisSite'].workdir,
                       "-p", str(port),
                       "-s", env['thisSite'].sitename,
                       "-t", str(env['proxycheckFlag']),
                       "-x", str(env['stageinretry']),
                       "-E", str(env['stageoutretry']),
                       "-F", env['experiment'],
                       "-H", env['cache'],
                       "-W", url]
            if 'yodaNodes' in env and subprocessName == "RunJobHpcEvent":
                jobargs.append("-N")
                jobargs.append(str(env['yodaNodes']))
            if 'yodaQueue' in env and subprocessName == "RunJobHpcEvent":
                jobargs.append("-Q")
                jobargs.append(str(env['yodaQueue']))

        tolog("Will use arguments: %s" % str(jobargs))
        return jobargs

    # Optional
    def doSpecialLogFileTransfer(self, **argdict):
        """ Should the log file be transfered to a special SE? """

        # The log file can at the end of the job be stored in a special SE - in addition to the normal stage-out of the log file
        # If this method returns True, the JobLog class will attempt to store the log file in a secondary SE after the transfer of
        # the log to the primary/normal SE. Additional information about the secondary SE is required and can be specified in
        # another optional method defined in the *Experiment classes

        # eventService = argdict.get('eventService', False)

        return False

    # Optional
    def getSchedconfigURL(self, protocol="http://"):
        """ Define the URL for the schedconfig / PanDA server"""
        # This method gets called from SiteInformation in case the URL is not set (by the wrapper)

        return protocol + "pandaserver.cern.ch"

    # Optional
    def getSubprocess(self, cmd, stdout=None, stderr=None):
        """ Execute and return a subprocess """

        process = None
        try:
            tolog("Executing command: %s" % (cmd))
            if stdout and stderr:
                # use stdout/stdout file objects to redirect the stdout/stderr streams
                process = Popen(cmd, shell=True, stdout=stdout, stderr=stderr, preexec_fn=os.setsid)
            else:
                process = Popen(cmd, shell=True)
        except Exception, e:
            tolog("!!WARNING!!2344!! Caught exception: %s" % (e))
        else:
            tolog("Subprocess is running")

        return process

    # Optional
    def getJobExecutionCommand4EventService(self):
        """ Define and test the command(s) that will be used to execute the payload for the event service """
        # E.g. cmd = ["source <path>/setup.sh; <path>/python <script>"]
        # The command returned from this method is executed using subprocess.Popen() from the runEvent module

        # Note: this optional method only need to be defined in case the event service is to be used
        # As of March 2014, this is not yet functional or documented.

        # The actual command must be declared as a list since that is expected by Popen()
        cmd = [""]

        return cmd

    # Optional
    def postGetJobActions(self, job):
        """ Perform any special post-job definition download actions here """

        # This method is called after the getJob() method has successfully downloaded a new job (job definition) from
        # the server. If the job definition e.g. contains information that contradicts WN specifics, this method can
        # be used to fail the job

        # Return any error code using ec, and any error message using pilotErrorDiag

        ec = 0
        pilotErrorDiag = ""

        return ec, pilotErrorDiag

    # Optional
    def useTracingService(self):

        return False

    # Optional
    def updateJobSetupScript(self, workdir, create=False, to_script=None):
        """ Create or update the job setup script (used to recreate the job locally if needed) """

        # If create=True, this step will only create the file with the script header (bash info)

        if create:
            filename = os.path.basename(self.getJobSetupScriptName(workdir))
            tolog("Creating job setup script with stage-in and payload execution commands: %s" % (filename))
            to_script = "#!/bin/bash\n# %s %s\n\n" % (filename, time.strftime("%d %b %Y %H:%M:%S", time.gmtime(time.time())))

        # Add the string to the setup script
        if to_script:
            self.addToJobSetupScript(to_script, workdir)

    # Optional
    def getJobSetupScriptName(self, workdir):
        """ return the name of the job setup file """

        return os.path.join(workdir, "job_setup.sh")

    # Optional
    def addToJobSetupScript(self, cmd, workdir):
        """ add/append command to job setup file """

        filename = self.getJobSetupScriptName(workdir)
        if not os.path.exists(filename):
            try:
                fp = open(filename, "w")
            except OSError, e:
                tolog("!!WARNING!!1880!! Could not open job setup file for writing: %s" % str(e))
        else:
            try:
                fp = open(filename, "a")
            except OSError, e:
                tolog("!!WARNING!!1880!! Could not open job setup file for appending: %s" % str(e))

        if fp:
            fp.write(cmd)
            fp.write("\n\n")
            fp.close()
            tolog("Updated %s: %s" % (filename, cmd))

    # Optional
    def getRelease(self, release):
        """ Return a list of the software release id's """
        # Assuming 'release' is a string that separates release id's with '\n'
        # Used in the case of payload using multiple steps with different release versions
        # E.g. release = "19.0.0\n19.1.0" -> ['19.0.0', '19.1.0']

        return release.split("\n")

    # Optional
    def formatReleaseString(release):
        """ Return a special formatted release string """
        # E.g. release = "Atlas-19.0.0" -> "19.0.0"
        # This method is required for ATLAS but is probably of no interest for any other PanDA user

        return release

    # Optional
    def setCache(self, cache):
        """ Cache URL """
        # Used e.g. by LSST

        self.__cache = cache

    # Optional
    def getCache(self):
        """ Return the cache URL """
        # Used e.g. by LSST

        return self.__cache

    # Optional
    def useTracingService(self):
        """ Use the Rucio Tracing Service """
        # A service provided by the Rucio system that allows for file transfer tracking; all file transfers
        # are reported by the pilot to the Rucio Tracing Service if this method returns True

        return False

    # Optional
    def updateJobDefinition(self, job, filename):
        """ Update the job definition file and object before using it in RunJob """

        # This method is called from Monitor, before RunJob is launched, which allows to make changes to the job object after it was downloaded from the job dispatcher
        # (used within Monitor) and the job definition file (which is used from RunJob to recreate the same job object as is used in Monitor).
        # 'job' is the job object, defined in Job.py, while 'filename' is the name of the file containing the job definition information.

        return job

    # Optional
    def shouldExecuteUtility(self):
        """ Determine whether a special utility should be executed """

        # The RunJob class has the possibility to execute a special utility, e.g. a memory monitor, that runs in parallel
        # to the payload (launched after the main payload process).
        # The utility is executed if this method returns True. The utility is currently expected to produce
        # a summary JSON file whose name is defined by the getUtilityJSONFilename() method. The contents of
        # this file (ie. the full JSON dictionary) will be added to the job update.
        #
        # Example of summary JSON file (ATLAS case):
        #   {"Max":{"maxVMEM":40058624,"maxPSS":10340177,"maxRSS":16342012,"maxSwap":16235568},
        #    "Avg":{"avgVMEM":19384236,"avgPSS":5023500,"avgRSS":6501489,"avgSwap":5964997}}
        #
        # While running, the MemoryMonitor also produces a regularly updated text file with the following format: (tab separated)
        #   Time          VMEM        PSS        RSS        Swap         (first line in file)
        #   1447960494    16099644    3971809    6578312    1978060

        return False

    # Optional
    def getUtilityOutputFilename(self):
        """ Return the filename of a utility output file """

        # For explanation, see shouldExecuteUtility()
        return "memory_monitor_output.txt"

    # Optional
    def getUtilityJSONFilename(self):
        """ Return the filename of a utility JSON file """

        # For explanation, see shouldExecuteUtility()
        return "utility_summary.json"

    # Optional
    def getUtilityInfo(self, workdir, pilot_initdir, allowTxtFile=False):
        """ Add the utility info to the node structure if available """

        # Extract the relevant information from the utility tool output and add it to the dictionary
        # returned by this method. The dictionary will be merged with the node dictionary in 
        # PandaServerClient::getNodeStructure() and sent to the PanDA server

        return {}

    # Optional
    def getUtilityCommand(self, **argdict):
        """ Prepare a utility command string """

        # This method can be used to prepare a setup string for an optional utility tool, e.g. a memory monitor,
        # that will be executed by the pilot in parallel with the payload.
        # The pilot will look for an output JSON file (summary.json) and will extract pre-determined fields
        # from it and report them with the job updates. Currently the pilot expects to find fields related
        # to memory information.

        # pid = argdict.get('pid', 0)

        return ""

    # Optional
    def getGUIDSourceFilename(self):
        """ Return the filename of the file containing the GUIDs for the output files """

        # In the case of ATLAS, Athena produces an XML file containing the GUIDs of the output files. The name of this
        # file is PoolFileCatalog.xml. If this method returns an empty string (ie the default), the GUID generation will
        # be done by the pilot in RunJobUtilities::getOutFilesGuids()

        return ""

    # Optional
    def buildFAXPath(self, **argdict):
        """ Build a proper FAX path """

        # This method builds proper FAX paths and is used in pure FAX mode (i.e. when FAX is used in forced mode),
        # particularly when the PoolFileCatalog.xml is built prior to stage-in
        # Only needed if FAX mechanism is used in forced mode (i.e. when copytoolin='fax')

        lfn = argdict.get('lfn', 'default_lfn')
        scope = argdict.get('scope', 'default_scope')
        subpath = argdict.get('subpath', 'atlas/rucio/')
        pandaID = argdict.get('pandaID', '')
        sourceSite = argdict.get('sourceSite', 'default_sourcesite')
        computingSite = argdict.get('computingSite', 'default_computingsite')

        # Get the proper FAX redirector (default ATLAS implementation)
        from FAXTools import getFAXRedirectors
        # First get the global redirectors (several, since the lib file might not be at the same place for overflow jobs)
        fax_redirectors_dictionary = getFAXRedirectors(computingSite, sourceSite, pandaID)
        tolog("fax_redirectors_dictionary=%s"%str(fax_redirectors_dictionary))
        # select the proper fax redirector
        if ".lib." in lfn:
            redirector = fax_redirectors_dictionary['computingsite']
        else:
            redirector = fax_redirectors_dictionary['sourcesite']

        # Make sure the redirector ends with a double slash
        if not redirector.endswith('//'):
            if redirector.endswith('/'):
                redirector += "/"
            else:
                redirector += "//"

        # Make sure that the subpath does not begin with a slash
        if subpath.startswith('/') and len(subpath) > 1:
            subpath = subpath[1:]

        tolog("redirector=%s"%(redirector))
        tolog("subpath=%s"%(subpath))
        tolog("scope=%s"%(scope))
        tolog("lfn=%s"%(lfn))

        return redirector + subpath + scope + ":" + lfn

if __name__ == "__main__":

    a=Experiment()
    print a.getSubprocessName(False)