recoverMissingLumis.py

#!/usr/bin/env python

from WMCore.ACDC.CouchCollection import CouchCollection
from WMCore.ACDC.CouchFileset import CouchFileset
from WMCore.Database.CMSCouch import Database
from WMCore.Services.DBS.DBSReader import DBSReader
from WMCore.Services.PhEDEx.PhEDEx import PhEDEx
from WMCore.Services.RequestDB.RequestDBReader import RequestDBReader
from WMCore.Services.SiteDB.SiteDB import SiteDBJSON
from WMCore.WMSpec.WMWorkload import WMWorkloadHelper
from pprint import pprint

import json
import logging
import pickle
import sys
import uuid
import time
import threading

from optparse import OptionParser

jsonBlob = {
    "createRequest":
        {
            "RequestString": "RequestString-OVERRIDE-ME",
            "RequestPriority": 9e5,
            "TimePerEvent": 10,
            "PrepID": None,
            "Campaign": None,
            "OriginalRequestName": "",
            "InitialTaskPath": "",
            "ACDCServer": "",
            "ACDCDatabase": "",
            "RequestType": "Resubmission",
            "Requestor": "",
            "Group": "",
            # "Memory": 2300,
            "SizePerEvent": 200,
            "CollectionName": "",
            "IgnoredOutputModules": []
        },
    "changeSplitting":
        {
            "DataProcessing":
                {
                    "SplittingAlgo": "LumiBased",
                    "lumis_per_job": 1,
                    "halt_job_on_file_boundaries": "True"
                }
        },
    "assignRequest":
        {
            "SiteWhitelist": "SiteWhitelist-OVERRIDE-ME",
            "MergedLFNBase": "/store/data",
            "UnmergedLFNBase": "/store/unmerged",
            "AcquisitionEra": "AcquisitionEra-OVERRIDE-ME",
            "ProcessingVersion": 1,
            "MaxRSS": 2411724,
            "MaxVSize": 20411724,
            "Dashboard": "reprocessing",
            "Team": "Team--OVERRIDE-ME",
            "OpenRunningTimeout": 0
        }
}


def buildDatasetTree(workload):
    """
    _buildDatasetTree_

    Build a tree with named edges that represents the connections between
    datasets in the workload. Assuming a standard ReReco with two skims, the
    graph would look like this:

    root = 'RAW'
    graph = {'RAW' : ['RECO', 'AOD', 'DQM'],
             'RECO' : ['USERSkim1', 'UserSkim2']}
    edges = {('RAW', 'RECO') : {'task' : '/request/DataProcessing', 'outMod' : ['RECOoutput']},
             ('RAW', 'AOD') : {'task' : '/request/DataProcessing', 'outMod' : ['AODoutput']},
             ('RAW', 'DQM') : {'task' : '/request/DataProcessing', 'outMod' : ['DQMoutput']},
             ('RECO', 'USERSkim1') : {'task' : '/request/DataProcessing/Merge/Skim1',
                                      'outMod' : ['Skim1']}
             ('RECO', 'USERSkim2') : {'task' : '/request/DataProcessing/Merge/Skim2',
                                      'outMod' : ['Skim2']}}

    Only non-transient outputs are valid nodes in the graph, and the name on
    the edge represents the task and list of output modules
    to be reproduced to get between input and output.
    """
    root = workload.listInputDatasets()[0]
    graph, edges = traverseWorkloadForDatasets(workload)
    graph, edges = reduceGraph(graph, root, edges)
    return graph, edges


def reduceGraph(graph, root, edges):
    """
    _reduceGraph_

    Take a graph and edge definition, starting from the root
    remove the transient nodes and ammend the connections
    between the persistent nodes. This is returned in a new graph
    and edges object.
    """
    newGraph = {}
    newEdges = {}
    if root not in graph:
        return newGraph, newEdges
    for child in graph[root]:
        tempGraph, tempEdges = reduceGraph(graph, child, edges)
        newGraph.update(tempGraph)
        newEdges.update(tempEdges)
        if child.endswith('transient'):
            if root not in newGraph:
                newGraph[root] = []
            newGraph[root].extend(newGraph[child])
            nodes = newGraph[child]
            for node in nodes:
                newEdges[(root, node)] = {'task': edges[(root, child)]['task'],
                                          'outMod': []}
                newEdges[(root, node)]['outMod'].extend(edges[(root, child)]['outMod'])
                newEdges[(root, node)]['outMod'].extend(newEdges[(child, node)]['outMod'])
                del newEdges[(child, node)]
            del newGraph[child]
        else:
            if root not in newGraph:
                newGraph[root] = []
            newGraph[root].append(child)
            newEdges[(root, child)] = edges[(root, child)]
            if newEdges[(root, child)]['outMod'] == ['Merged']:
                newEdges[(root, child)]['outMod'] = []
    return newGraph, newEdges


def traverseWorkloadForDatasets(workload, initialTask=None,
                                roots=None):
    """
    _traverseWorkloadForDatasets_

    Go through the workload recursively and build a graph that links
    the different tasks with their outputs in a graph-like structure.
    """
    graph = {}
    edges = {}

    if initialTask:
        taskIterator = initialTask.childTaskIterator()
    else:
        taskIterator = workload.taskIterator()

    for task in taskIterator:
        taskChildren = {}
        inputRef = task.inputReference()
        for stepName in task.listAllStepNames():
            stepHelper = task.getStepHelper(stepName)
            if not getattr(stepHelper.data.output, "keep", True):
                continue

            if stepHelper.stepType() == "CMSSW" or \
                            stepHelper.stepType() == "MulticoreCMSSW":
                for outputModuleName in stepHelper.listOutputModules():
                    outputModule = stepHelper.getOutputModule(outputModuleName)
                    transient = getattr(outputModule, "transient", False)
                    outputDataset = "/%s/%s/%s" % (outputModule.primaryDataset,
                                                   outputModule.processedDataset,
                                                   outputModule.dataTier)
                    if transient or task.taskType() != "Merge":
                        outputDataset += "transient"
                    if roots is None:
                        root = task.getInputDatasetPath()
                    else:
                        root = roots[inputRef.outputModule]
                    if root not in graph:
                        graph[root] = []
                    graph[root].append(outputDataset)
                    edges[(root, outputDataset)] = {'task': task.getPathName(),
                                                    'outMod': [outputModuleName]}
                    taskChildren[outputModuleName] = outputDataset
        otherGraph, otherEdges = traverseWorkloadForDatasets(workload,
                                                             task,
                                                             taskChildren)
        graph.update(otherGraph)
        edges.update(otherEdges)
    return graph, edges


def getOutputModules(workload, initialTask=None):
    """
    _getOutputModules_

    Retrieve the list of all output modules in the workload.
    """
    outputModules = []
    if initialTask:
        taskIterator = initialTask.childTaskIterator()
    else:
        taskIterator = workload.taskIterator()

    for task in taskIterator:
        for stepName in task.listAllStepNames():
            stepHelper = task.getStepHelper(stepName)
            if not getattr(stepHelper.data.output, "keep", True):
                continue
            if stepHelper.stepType() == "CMSSW" or \
                            stepHelper.stepType() == "MulticoreCMSSW":
                for outputModuleName in stepHelper.listOutputModules():
                    if task.taskType() != "Merge":
                        outputModules.append(outputModuleName)
        otherOutputModules = getOutputModules(workload,
                                              task)
        outputModules.extend(otherOutputModules)
    return outputModules

def ThreadBuster( threads, n_threads, sleepy, verbose=False):

    ntotal=len(threads)
    print "Processing",ntotal,"threads with",n_threads,"max concurrent"
    start_now = time.mktime(time.gmtime())
    r_threads = []
    bug_every=max(len(threads) / 10., 100.) ## 10 steps of eta verbosity
    next_ping = int(len(threads)/bug_every)
    while threads:
        running = sum([t.is_alive() for t in r_threads])
        #if verbose: print running,"/",n_threads,"running threads"
        if n_threads==None or running < n_threads:
            startme = n_threads-running if n_threads else len(threads)
            if verbose or int(len(threads)/bug_every)<next_ping:
                next_ping =int(len(threads)/bug_every)
                now= time.mktime(time.gmtime())
                spend = (now - start_now)
                n_done = ntotal-len(threads)
                print "Starting",startme,"new threads",len(threads),"remaining" 
                if n_done:
                    eta = (spend / n_done) * len(threads)
                    print "Will finish in ~%.2f [s]"%(eta)
            if startme > n_threads/5.:
                sleepy/=2.
            for it in range(startme):
                if threads:
                    r_threads.append( threads.pop(-1))
                    r_threads[-1].start()
        time.sleep(sleepy)
    ##then wait for completion
    while sum([t.is_alive() for t in r_threads]):
        time.sleep(1)
        
    ## and swap list back
    return r_threads


def getFiles(datasetName, runBlacklist, runWhitelist, blockBlacklist,
             blockWhitelist, dbsUrl, fakeLocation=False):
    """
    _getFiles_

    Get the full information of a dataset including files, blocks, runs and lumis.
    Filter it using run and block white/black lists.

    It can receive and optional DBSUrl.
    """
    dbsReader = DBSReader(endpoint=dbsUrl)
    phedexReader = PhEDEx()
    siteDB = SiteDBJSON()


    class BlockBuster(threading.Thread):
        def __init__(self, **args):
            threading.Thread.__init__(self)
            for k,v in args.items():
                setattr(self,k,v)
            self.major_failure = False

        def run(self):
            self.files = {}
            logging = self.l
            has_parent = self.hp
            fakeLocation = self.fl
            blockName = self.bn
            blockBlacklist = self.bbl
            blockWhitelist = self.bwl


            if blockBlacklist and blockName in blockBlacklist:
                return
            if blockWhitelist and blockName not in blockWhitelist:
                return

            phedexReader = PhEDEx()
            siteDB = SiteDBJSON()
            dbsReader = DBSReader(endpoint=self.dbs)
            replicaInfo = phedexReader.getReplicaInfoForBlocks(block=blockName,
                                                                    subscribed='y')
            blockFiles = dbsReader.listFilesInBlock(blockName, lumis=True)
            if has_parent:
                try:
                    blockFileParents = dbsReader.listFilesInBlockWithParents(blockName)
                except:
                    print blockName, "does not appear to have a parent, even though it should. Very suspicious"
                    blockFileParents = dbsReader.listFilesInBlock(blockName)
            else:
                blockFileParents = dbsReader.listFilesInBlock(blockName)

            blockLocations = set()
            # load block locations
            if len(replicaInfo["phedex"]["block"]) > 0:
                for replica in replicaInfo["phedex"]["block"][0]["replica"]:
                    PNN = replica["node"]
                    PSNs = siteDB.PNNtoPSN(PNN)
                    blockLocations.add(PNN)
                    #logging.debug("PhEDEx Node Name: %s\tPSNs: %s", PNN, PSNs)

            # We cannot upload docs without location, so force it in case it's empty
            if not blockLocations:
                if fakeLocation:
                    #logging.info("\t\t %s\tno location", blockName)
                    blockLocations.update([u'T1_US_FNAL_Disk', u'T2_CH_CERN'])
                elif not has_parent:  ## this should be the source
                    #logging.info("Blockname: %s\tno location, ABORT", blockName)
                    self.major_failure = True
                    #sys.exit(1)
                
            #logging.info("Blockname: %s\tLocations: %s", blockName, blockLocations)

            # for each file on the block
            for blockFile in blockFiles:
                parentLFNs = []
                # populate parent information
                if blockFileParents and "ParentList" in blockFileParents[0]:
                    for fileParent in blockFileParents[0]["ParentList"]:
                        parentLFNs.append(fileParent["LogicalFileName"])
                runInfo = {}
                # Lumis not included in file
                for lumiSection in blockFile["LumiList"]:
                    if runBlacklist and lumiSection["RunNumber"] in runBlacklist:
                        continue
                    if runWhitelist and lumiSection["RunNumber"] not in runWhitelist:
                        continue

                    if lumiSection["RunNumber"] not in runInfo.keys():
                        runInfo[lumiSection["RunNumber"]] = []

                    runInfo[lumiSection["RunNumber"]].append(lumiSection["LumiSectionNumber"])
                if len(runInfo.keys()) > 0:
                    self.files[blockFile["LogicalFileName"]] = {"runs": runInfo,
                                                                "events": blockFile["NumberOfEvents"],
                                                                "size": blockFile["FileSize"],
                                                                "locations": list(blockLocations),
                                                                "parents": parentLFNs}
            return
            

    files = {}
    outputDatasetParts = datasetName.split("/")
    print "dataset", datasetName, "parts", outputDatasetParts
    try:
        # retrieve list of blocks from dataset
        blockNames = dbsReader.listFileBlocks(datasetName)
    except:
        raise RuntimeError("Dataset %s doesn't exist in given DBS instance" % datasetName)

    has_parent = False
    try:
        parents = dbsReader.listDatasetParents(datasetName)
        if parents: has_parent = True
    except:
        print "Dataset with no parent"
        pass

    bthreads=[]
    # traverse each block
    for blockName in blockNames:
        bthreads.append( BlockBuster( bn = blockName,
                                      hp=has_parent, 
                                      fl = fakeLocation,
                                      bbl = blockBlacklist,
                                      bwl = blockWhitelist,
                                      l = logging,
                                      dbs=dbsUrl))

    print len(bthreads),"block query created"
    bthreads = ThreadBuster( bthreads, 40, 2., verbose=False)

    for t in bthreads:
        if t.major_failure: sys.exit(1)
        files.update(t.files)

    return files


def diffDatasets(inputDataset, outputDataset):
    """
    _diffDatasets_

    Compare two datasets in terms of lumi section content.
    Return a dictionary containing the run/lumis that are in the input
    but not in the output.
    """

    inputRunInfo = {}
    # make a set of input run-lumis.
    for inputLFN in inputDataset:
        for inputRun in inputDataset[inputLFN]["runs"]:
            if inputRun not in inputRunInfo:
                inputRunInfo[inputRun] = set()
                # if type is list or [list]
            if type(inputDataset[inputLFN]["runs"][inputRun][0]) is list:
                inputRunInfo[inputRun].update(inputDataset[inputLFN]["runs"][inputRun][0])
            elif type(inputDataset[inputLFN]["runs"][inputRun]) is list:
                inputRunInfo[inputRun].update(inputDataset[inputLFN]["runs"][inputRun])
            else:
                raise RuntimeError("Don't know what this is:" + str(inputDataset[inputLFN]["runs"][inputRun]))

    outputRunInfo = {}
    # make a set of input run-lumis.
    for outputLFN in outputDataset:
        for outputRun in outputDataset[outputLFN]["runs"]:
            if outputRun not in outputRunInfo:
                outputRunInfo[outputRun] = set()
            # if type is list or [list]
            if type(outputDataset[outputLFN]["runs"][outputRun][0]) is list:
                outputRunInfo[outputRun].update(outputDataset[outputLFN]["runs"][outputRun][0])
            elif type(outputDataset[outputLFN]["runs"][outputRun]) is list:
                outputRunInfo[outputRun].update(outputDataset[outputLFN]["runs"][outputRun])
            else:
                raise RuntimeError("Don't know what this is:" + str(outputDataset[outputLFN]["runs"][outputRun]))

    diffRunInfo = {}
    # make set difference for each run
    for inputRun in inputRunInfo:
        if inputRun not in outputRunInfo:
            diffRunInfo[inputRun] = inputRunInfo[inputRun]
        else:
            diffLumis = inputRunInfo[inputRun] - outputRunInfo[inputRun]
            if diffLumis:
                diffRunInfo[inputRun] = diffLumis

    return diffRunInfo


def buildDifferenceMap(workload, datasetInformation):
    """
    _buildDifferenceMap_

    Check the output datasets and find which lumis are differing from the input
    so we may construct the appriopiate requests. It returns
    a dictionary structure with the output datasets as the key
    and the difference in lumis.
    """
    differences = {}
    inputDataset = workload.listInputDatasets()[0]
    logging.info("InputDataset : %s", inputDataset)
    logging.info("OutputDataset: %s", workload.listOutputDatasets())
    for dataset in workload.listOutputDatasets():
        difference = diffDatasets(datasetInformation[inputDataset], datasetInformation[dataset])
        if difference:
            differences[dataset] = difference
    return differences


def getRequestInformationAndWorkload(requestName, reqmgrUrl, centralRequestDBURL):
    """
    _getRequestInformationAndWorkload_

    Retrieve the request information for assignment
    and the full pickled workload.
    """
    wfDBReader = RequestDBReader(centralRequestDBURL, couchapp="ReqMgr")
    result = wfDBReader.getRequestByNames(requestName, True)
    workloadDB = Database(result[requestName]['CouchWorkloadDBName'], result[requestName]['CouchURL'])
    workloadPickle = workloadDB.getAttachment(requestName, 'spec')
    spec = pickle.loads(workloadPickle)
    workload = WMWorkloadHelper(spec)
    return workload, result[requestName]


def defineRequests(workload, requestInfo,
                   acdcCouchUrl, acdcCouchDb,
                   requestor, group,
                   dbsUrl,
                   fakeLocation,
                   datasetInformation=None):
    """
    _defineRequests_

    This is the ultimate function,
    it will create JSONs for the appropiate resubmission requests
    that can be feed into the reqmgr.py script and it will assemble
    acdc records that can be uploaded to the database.
    """
    main_now = time.mktime(time.gmtime())
    # First retrieve the run and block lists and load
    # the information of all datasets
    logging.debug("Original request info:\n%s", requestInfo)
    topTask = workload.getTopLevelTask()[0]
    runWhitelist = topTask.inputRunWhitelist()
    runBlacklist = topTask.inputRunBlacklist()
    blockWhitelist = topTask.inputBlockWhitelist()
    blockBlacklist = topTask.inputBlockBlacklist()
    inputDataset = workload.listInputDatasets()[0]
    outputModules = getOutputModules(workload)
    if datasetInformation is None:
        datasetInformation = {}
        logging.info("Loading DBS information for the datasets...")
        now = time.mktime(time.gmtime())
        datasetInformation[inputDataset] = getFiles(inputDataset, runBlacklist, runWhitelist,
                                                    blockBlacklist, blockWhitelist, dbsUrl, fakeLocation=fakeLocation)
        print time.mktime(time.gmtime())-now,"[s] for a call to getFiles",inputDataset
        for dataset in workload.listOutputDatasets():
            now = time.mktime(time.gmtime())
            datasetInformation[dataset] = getFiles(dataset, runBlacklist, runWhitelist, blockBlacklist, blockWhitelist,
                                                   dbsUrl)
            print time.mktime(time.gmtime())-now,"[s] for a call to getFiles",dataset
        logging.info("Finished loading DBS information for the datasets...")

    # Now get the information about the datasets and tasks
    nodes, edges = buildDatasetTree(workload)
    logging.info("Dataset tree built...")
    for k, v in nodes.items():
        logging.debug("%s : %s" % (k, v))
    for k, v in edges.items():
        logging.debug("%s : %s" % (k, v))
    # Load the difference information between input and outputs
    differenceInformation = buildDifferenceMap(workload, datasetInformation)
    logging.info("Difference map processed...")
    logging.debug("%s" % str(differenceInformation))
    # Define an object that will hold the potential requests
    requests = []
    logging.info("Now definining the required requests...")
    # First generate requests for the datasets with children, that way we can
    # shoot the requests with skims in single requests
    for dataset in differenceInformation.keys():
        if dataset not in nodes:
            continue
        datasetsToRecover = [dataset]
        diffedLumis = differenceInformation[dataset]
        taskToRecover = edges[(inputDataset, dataset)]['task']
        outputModulesToRecover = edges[(inputDataset, dataset)]['outMod']
        intersectionDiff = {}
        for childDataset in nodes[dataset]:
            childDiffLumis = differenceInformation[childDataset]
            matchAvailable = False
            for run in diffedLumis:
                if run in childDiffLumis:
                    for lumi in diffedLumis[run]:
                        if lumi in childDiffLumis[run]:
                            matchAvailable = True
                            break
            if matchAvailable:
                outputModulesToRecover.extend(edges[(dataset, childDataset)]['outMod'])
                datasetsToRecover.append(childDataset)
            for run in diffedLumis:
                if run in childDiffLumis:
                    if run not in intersectionDiff:
                        intersectionDiff[run] = set()
                        intersectionDiff[run] = diffedLumis[run] & childDiffLumis[run]
                    else:
                        intersectionDiff[run] &= diffedLumis[run] & childDiffLumis[run]
                else:
                    intersectionDiff[run] = set()
        for run in intersectionDiff:
            if not intersectionDiff[run]:
                del intersectionDiff[run]
        if not intersectionDiff:
            # Can't create request for this dataset + children
            continue
        for run in intersectionDiff:
            for childDataset in nodes[dataset]:
                childDiffLumis = differenceInformation[childDataset]
                if run in childDiffLumis:
                    childDiffLumis[run] -= intersectionDiff[run]
                    if not childDiffLumis[run]:
                        del childDiffLumis[run]
            diffedLumis[run] -= intersectionDiff[run]
            if not diffedLumis[run]:
                del diffedLumis[run]
        if not diffedLumis:
            del differenceInformation[dataset]
        for childDataset in nodes[dataset]:
            if not differenceInformation[childDataset]:
                del differenceInformation[childDataset]

        requestObject = {'task': taskToRecover,
                         'input': inputDataset,
                         'lumis': intersectionDiff,
                         'outMod': outputModulesToRecover,
                         'outputs': datasetsToRecover}
        requests.append(requestObject)
    # Now go through all the output datasets, creating a single request for
    # each
    for dataset in differenceInformation:
        datasetsToRecover = [dataset]
        diffedLumis = differenceInformation[dataset]
        if (inputDataset, dataset) in edges:
            taskToRecover = edges[(inputDataset, dataset)]['task']
            outputModulesToRecover = edges[(inputDataset, dataset)]['outMod']
            parentDataset = inputDataset
        else:
            for parentDataset in nodes:
                if dataset in nodes[parentDataset]:
                    taskToRecover = edges[(parentDataset, dataset)]['task']
                    outputModulesToRecover = edges[(parentDataset, dataset)]['outMod']
                    break
        requestObject = {'task': taskToRecover,
                         'input': parentDataset,
                         'lumis': diffedLumis,
                         'outMod': outputModulesToRecover,
                         'outputs': datasetsToRecover}
        requests.append(requestObject)

    logging.info("About to upload ACDC records to: %s/%s" % (acdcCouchUrl, acdcCouchDb))
    ## this printout is making a lot of crap
    ##pprint(requests)

    # With the request objects we need to build ACDC records and
    # request JSONs

    class CouchBuster(threading.Thread):
        def __init__(self, **args):
            threading.Thread.__init__(self)
            import copy
            for k,v in args.items():
                #if not k in ['c']:
                #    setattr(self,k,copy.deepcopy(v))
                #else:
                #    setattr(self,k,v)
                setattr(self,k,v)

        def run(self):

            lfn = self.lfn
            if self.v:
                print "Starting for",lfn
            now = time.mktime(time.gmtime())
            fileInfo = self.fi
            requestObject = self.ro
            fileRuns = {}
            acdcCouchUrl = self.ac
            acdcCouchDb = self.acd
            filesetName = self.fsn
            collection = self.c
            self.lumis = 0
            self.files = 0
            for run in fileInfo['runs']:
                if run in requestObject['lumis']:
                    for lumi in fileInfo['runs'][run][0]:
                        if lumi in requestObject['lumis'][run]:
                            if run not in fileRuns:
                                fileRuns[run] = []
                            fileRuns[run].append(lumi)
                            self.lumis += 1
            if fileRuns:
                self.files += 1
                fileset = CouchFileset(**{"url": acdcCouchUrl,
                                          "database": acdcCouchDb,
                                          "name": filesetName})
                fileset.setCollection(collection)
                acdcRuns = []
                for run in fileRuns:
                    runObject = {}
                    runObject['run_number'] = int(run)
                    runObject['lumis'] = fileRuns[run]
                    acdcRuns.append(runObject)
                acdcFile = {"lfn": lfn,
                            "first_event": 0,
                            "last_event": 0,
                            "checksums": {},
                            "size": fileInfo["size"],
                            "events": fileInfo["events"],
                            "merged": 1,
                            "parents": fileInfo["parents"],
                            "locations": fileInfo["locations"],
                            "runs": acdcRuns
                           }
                #
                fileset.makeFilelist({lfn: acdcFile})
            if self.v:
                print time.mktime(time.gmtime()) - now,"[s] for makeFilelist",lfn
                

    for idx, requestObject in enumerate(requests):
        now = time.mktime(time.gmtime())
        collectionName = '%s_%s' % (workload.name(), str(uuid.uuid1()))
        print time.mktime(time.gmtime()) - now,"[s]","starting",idx,"in collection name",collectionName
        filesetName = requestObject['task']
        collection = CouchCollection(**{"url": acdcCouchUrl,
                                        "database": acdcCouchDb,
                                        "name": collectionName})
        print time.mktime(time.gmtime()) - now,"[s]","collection created"
        files = 0
        lumis = 0
        cthreads=[]
        for lfn in datasetInformation[requestObject['input']]:
            cthreads.append( CouchBuster( lfn = lfn,
                                          fi = datasetInformation[requestObject['input']][lfn],
                                          ro = requestObject,
                                          ac = acdcCouchUrl,
                                          acd = acdcCouchDb,
                                          fsn = filesetName,
                                          c = collection,
                                          v = False
                                          ))
            
        print len(cthreads),"CouchBuster created"            
        cthreads = ThreadBuster( cthreads, 40, 2., verbose=False)

        for t in cthreads:
            files += t.files
            lumis += t.lumis
        

        print time.mktime(time.gmtime()) - now,"[s]","ending loop"
        # Put the creation parameters
        creationDict = jsonBlob["createRequest"]
        creationDict["OriginalRequestName"] = str(workload.name())
        creationDict["InitialTaskPath"] = requestObject['task']
        creationDict["CollectionName"] = collectionName
        creationDict["IgnoredOutputModules"] = list(set(outputModules) - set(requestObject['outMod']))
        creationDict["ACDCServer"] = acdcCouchUrl
        creationDict["ACDCDatabase"] = acdcCouchDb
        #creationDict["RequestString"] = "r-%d-%s" % (idx, workload.name())[:50]
        creationDict["RequestString"] = "r-%d-%s" % (idx, requestInfo['RequestString'])
        creationDict["Requestor"] = requestor
        creationDict["RequestPriority"] = min(500000, requestInfo.get('RequestPriority',60000)*2 )
        creationDict["Group"] = group
        creationDict["TimePerEvent"] = requestInfo['TimePerEvent']
        creationDict["Memory"] = requestInfo['Memory']
        creationDict["SizePerEvent"] = requestInfo['SizePerEvent']
        creationDict["PrepID"] = requestInfo.get('PrepID')
        creationDict["Campaign"] = requestInfo.get('Campaign')

        # Assign parameters
        assignDict = jsonBlob["assignRequest"]
        team = requestInfo['Team']
        processingString = requestInfo['ProcessingString']
        processingVersion = requestInfo['ProcessingVersion']
        acqEra = requestInfo['AcquisitionEra']
        mergedLFNBase = requestInfo['MergedLFNBase']
        unmergedLFNBase = requestInfo['UnmergedLFNBase']
        # processingString = workload.getProcessingString()
        # processingVersion = workload.getProcessingVersion()
        # acqEra = workload.getAcquisitionEra()
        # mergedLFNBase = workload.getMergedLFNBase()
        # unmergedLFNBase = workload.getUnmergedLFNBase()
        topTask = workload.getTopLevelTask()[0]
        siteWhitelist = topTask.siteWhitelist()
        assignDict["SiteWhitelist"] = siteWhitelist
        assignDict["MergedLFNBase"] = mergedLFNBase
        assignDict["UnmergedLFNBase"] = unmergedLFNBase
        assignDict["AcquisitionEra"] = acqEra
        assignDict["Team"] = team
        try:
            int(processingVersion)
            assignDict["ProcessingVersion"] = int(processingVersion)
            if processingString is not None and processingString != 'None':
                assignDict["ProcessingString"] = processingString
        except Exception:
            tokens = processingVersion.split('-')
            assignDict["ProcessingVersion"] = int(tokens[-1][1:])
            assignDict["ProcessingString"] = ('-').join(tokens[:-1])
        print time.mktime(time.gmtime()) - now,"[s]","data prepared"
        fileHandle = open('%s.json' % creationDict["RequestString"], 'w')
        json.dump(jsonBlob, fileHandle)
        fileHandle.close()
        print time.mktime(time.gmtime()) - now,"[s]","json made"
        logging.info("Created JSON %s for recovery of %s" % ('%s.json' % creationDict["RequestString"],
                                                             requestObject['outputs']))
        logging.info("This will recover %d lumis in %d files" % (lumis, files))
    print time.mktime(time.gmtime()) - main_now,"[s]","to complete"

def main():
    """
    _main_

    Define the CLI options and perform the main script function
    """
    myOptParser = OptionParser()
    myOptParser.add_option("-r", "--requestName", dest="requestName",
                           help="Name of the request to recover")
    myOptParser.add_option("-q", "--requestor", dest="requestor",
                           help="Requestor for the Resubmission requests")
    myOptParser.add_option("-g", "--group", dest="group",
                           help="Group for the Resubmission requests")
    myOptParser.add_option("-d", "--dbsUrl", dest="dbsUrl",
                           default="https://cmsweb.cern.ch/dbs/prod/global/DBSReader",
                           help="URL for the DBS instance to get the dataset info (both DBS2 and DBS3 work)")
    myOptParser.add_option("-a", "--acdcUrl", dest="acdcUrl",
                           default="https://cmsweb.cern.ch/couchdb",
                           help="URL for the ACDC database where the records can be uploaded")
    myOptParser.add_option("-s", "--acdcServer", dest="acdcServer",
                           default="acdcserver",
                           help="Database name for the ACDC records")
    myOptParser.add_option("-u", "--reqMgrUrl", dest="reqMgrUrl",
                           default="https://cmsweb.cern.ch/reqmgr2/data",
                           help="URL to the request manager API")
    myOptParser.add_option("-c", "--centralRequestDBURL", dest="centralRequestDBURL",
                           default="https://cmsweb.cern.ch/couchdb/reqmgr_workload_cache",
                           help="URL to the centralRequestDBURL")
    myOptParser.add_option("-j", "--json", dest="jsonFile",
                           help="For testing only: JSON file with the dataset information already loaded")
    myOptParser.add_option("-v", "--verbose", dest="verbose",
                           default=False, action="store_true",
                           help="Increase the level of verbosity for debug")
    myOptParser.add_option("-f", "--fake", dest="fake",
                           default=False, action="store_true",
                           help="In case there is no block location, forces it to CERN and FNAL")
    val, _ = myOptParser.parse_args()

    logging.getLogger().addHandler(logging.StreamHandler(sys.stdout))
    if val.verbose:
        logging.getLogger().setLevel(logging.DEBUG)
    else:
        logging.getLogger().setLevel(logging.INFO)

    # First load the request
    if val.requestName is None:
        raise RuntimeError("Request name must be specified.")
    workload, requestInfo = getRequestInformationAndWorkload(val.requestName, val.reqMgrUrl, val.centralRequestDBURL)
    # If testing, then load the pre-fetched dataset info
    datasetInformation = None
    if val.jsonFile is not None:
        fileHandle = open(val.jsonFile, 'r')
        datasetInformation = json.load(fileHandle)
        fileHandle.close()
    if val.requestor is None or val.group is None:
        raise RuntimeError("Requestor and group must be specified.")
    # Define the requests
    defineRequests(workload, requestInfo, val.acdcUrl, val.acdcServer, val.requestor,
                   val.group, val.dbsUrl, val.fake, datasetInformation)


if __name__ == "__main__":
    sys.exit(main())