def loadData(self):
"""
_loadData_
Load all the files that belong to this fileset.
"""
existingTransaction = self.beginTransaction()
if self.name == None or self.id < 0:
self.load()
action = self.daofactory(classname = "Files.InFileset")
results = action.execute(fileset = self.id,
conn = self.getDBConn(),
transaction = self.existingTransaction())
self.files = set()
self.newfiles = set()
for result in results:
file = File(id = result["fileid"])
file.loadData(parentage = 1)
self.files.add(file)
self.commitTransaction(existingTransaction)
return
def createFileFromDataStructsFile(self, file, jobID):
"""
_createFileFromDataStructsFile_
This function will create a WMBS File given a DataStructs file
"""
wmbsFile = File()
wmbsFile.update(file)
if type(file["locations"]) == set:
seName = list(file["locations"])[0]
elif type(file["locations"]) == list:
if len(file['locations']) > 1:
logging.error("Have more then one location for a file in job %i" % (jobID))
logging.error("Choosing location %s" % (file['locations'][0]))
seName = file["locations"][0]
else:
seName = file["locations"]
wmbsFile["locations"] = set()
if seName != None:
wmbsFile.setLocation(se = seName, immediateSave = False)
wmbsFile['jid'] = jobID
self.wmbsFilesToBuild.append(wmbsFile)
return wmbsFile
def populateACDCCouch(self, numFiles=3, lumisPerJob=4, eventsPerJob=100, numberOfJobs=250):
"""
_populateACDCCouch_
Create production files in couchDB to test the creation
of ACDC jobs for the EventBased algorithm.
"""
# Define some constants
workflowName = "ACDC_TestEventBased"
filesetName = "/%s/Production" % workflowName
lumisPerFile = lumisPerJob * numberOfJobs
for i in range(numFiles):
lfn = "MCFakeFile-some-hash-%s" % str(i).zfill(5)
for j in range(numberOfJobs):
firstEvent = j * eventsPerJob + 1
acdcFile = File(lfn=lfn, size=1024, events=eventsPerJob, locations=self.validLocations,
merged=False, first_event=firstEvent)
run = Run(1, *range(1 + j * lumisPerJob + (i * lumisPerFile),
1 + (j + 1) * lumisPerJob + (i * lumisPerFile)))
acdcFile.addRun(run)
acdcDoc = {"collection_name": workflowName,
"collection_type": "ACDC.CollectionTypes.DataCollection",
"files": {lfn: acdcFile},
"fileset_name": filesetName}
self.couchDB.queue(acdcDoc)
self.couchDB.commit()
return
def testAddToFileset(self):
"""
_AddToFileset_
Test to see if we can add to a fileset using the DAO
"""
testFileset = Fileset(name = "inputFileset")
testFileset.create()
testFileA = File(lfn = "/this/is/a/lfnA", size = 1024, events = 10)
testFileA.addRun(Run( 1, *[45]))
testFileA.create()
testFileB = File(lfn = "/this/is/a/lfnB", size = 1024, events = 10)
testFileB.addRun(Run( 1, *[45]))
testFileB.create()
addToFileset = self.daofactory(classname = "Files.AddToFileset")
addToFileset.execute(file = [testFileA['lfn'], testFileB['lfn']],
fileset = testFileset.id)
testFileset2 = Fileset(name = "inputFileset")
testFileset2.loadData()
self.assertEqual(len(testFileset2.files), 2)
for file in testFileset2.files:
self.assertTrue(file in [testFileA, testFileB])
# Check that adding twice doesn't crash
addToFileset.execute(file = [testFileA['lfn'], testFileB['lfn']],
fileset = testFileset.id)
def createTestSubscription(self, nFiles, nSites=1, closeFileset=False):
"""
_createTestSubscription_
Create a set of test subscriptions for testing purposes.
"""
if nSites > self.nSites:
nSites = self.nSites
testFileset = Fileset(name="TestFileset")
testFileset.create()
# Create a testWorkflow
testWorkflow = Workflow(spec="spec.xml", owner="Steve", name="wf001", task="Test")
testWorkflow.create()
# Create the files for each site
for s in range(nSites):
for i in range(nFiles):
newFile = File(makeUUID(), size=1024, events=100, locations=set(["site%i.cern.ch" % s]))
newFile.create()
testFileset.addFile(newFile)
testFileset.commit()
testSubscription = Subscription(
fileset=testFileset, workflow=testWorkflow, split_algo="MinFileBased", type="Processing"
)
testSubscription.create()
# Close the fileset
if closeFileset:
testFileset.markOpen(isOpen=False)
return testSubscription
def sortedFilesFromMergeUnits(mergeUnits):
"""
_sortedFilesFromMergeUnits_
Given a list of merge units sort them and the files that they contain.
Return a list of sorted WMBS File structures.
"""
mergeUnits.sort(mergeUnitCompare)
sortedFiles = []
for mergeUnit in mergeUnits:
mergeUnit["files"].sort(fileCompare)
for file in mergeUnit["files"]:
newFile = File(id=file["file_id"], lfn=file["file_lfn"],
events=file["file_events"])
newFile.addRun(Run(file["file_run"], file["file_lumi"]))
# The WMBS data structure puts locations that are passed in through
# the constructor in the "newlocations" attribute. We want these to
# be in the "locations" attribute so that they get picked up by the
# job submitter.
newFile["locations"] = set([file["pnn"]])
newFile.addRun(Run(file["file_run"], file["file_lumi"]))
sortedFiles.append(newFile)
return sortedFiles
def createTestJob(subscriptionType="Merge"):
"""
_createTestJob_
Create a test job with two files as input. This will also create the
appropriate workflow, jobgroup and subscription.
"""
testWorkflow = Workflow(spec=makeUUID(), owner="Simon", name=makeUUID(), task="Test")
testWorkflow.create()
testWMBSFileset = Fileset(name="TestFileset")
testWMBSFileset.create()
testSubscription = Subscription(fileset=testWMBSFileset, workflow=testWorkflow, type=subscriptionType)
testSubscription.create()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testFileA = File(lfn="/this/is/a/lfnA", size=1024, events=10)
testFileA.addRun(Run(1, *[45]))
testFileB = File(lfn="/this/is/a/lfnB", size=1024, events=10)
testFileB.addRun(Run(1, *[46]))
testFileA.create()
testFileB.create()
testJob = Job(name=makeUUID(), files=[testFileA, testFileB])
testJob["couch_record"] = "somecouchrecord"
testJob["location"] = "test.site.ch"
testJob.create(group=testJobGroup)
testJob.associateFiles()
return testJob
def populateACDCCouch(self, numFiles = 2, lumisPerJob = 35,
eventsPerJob = 20000):
"""
_populateACDCCouch_
Create production files in couchDB to test the creation
of ACDC jobs for the EventBased algorithm
"""
# Define some constants
workflowName = "ACDC_TestEventBased"
filesetName = "/%s/Production" % workflowName
owner = "*****@*****.**"
group = "unknown"
lumisPerFile = lumisPerJob * 250
for i in range(numFiles):
for j in range(250):
lfn = "MCFakeFile-some-hash-%s" % str(i).zfill(5)
acdcFile = File(lfn = lfn, size = 100, events = eventsPerJob, locations = self.validLocations,
merged = False, first_event = 1)
run = Run(1, *range(1 + (i * lumisPerFile) + j * lumisPerJob,
(j + 1) * lumisPerJob + (i * lumisPerFile) + 2))
acdcFile.addRun(run)
acdcDoc = {"collection_name" : workflowName,
"collection_type" : "ACDC.CollectionTypes.DataCollection",
"files" : {lfn : acdcFile},
"fileset_name" : filesetName,
"owner" : {"user": owner,
"group" : group}}
self.couchDB.queue(acdcDoc)
self.couchDB.commit()
return
def loadData(self):
"""
_loadData_
Load all information about the job, including the mask and all input
files. Either the ID or the name must be specified before this is
called.
"""
existingTransaction = self.beginTransaction()
self.load()
self.getMask()
fileAction = self.daofactory(classname = "Jobs.LoadFiles")
files = fileAction.execute(self["id"], conn = self.getDBConn(),
transaction = self.existingTransaction())
self["input_files"] = []
for file in files:
newFile = File(id = file["id"])
newFile.loadData(parentage = 0)
self.addFile(newFile)
self.commitTransaction(existingTransaction)
return
def generateFakeMCFile(self, numEvents = 100, firstEvent = 1,
lastEvent = 100, firstLumi = 1, lastLumi = 10,
index = 1):
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCTestFileset %i" % index)
singleMCFileset.create()
newFile = File("MCFakeFileTest %i" % index, size = 1000,
events = numEvents,
locations = set(["somese.cern.ch"]))
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
newFile.create()
singleMCFileset.addFile(newFile)
singleMCFileset.commit()
testWorkflow = Workflow(spec = "spec.xml", owner = "Steve",
name = "wf001", task="Test")
testWorkflow.create()
singleMCFileSubscription = Subscription(fileset = singleMCFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Production")
singleMCFileSubscription.create()
return singleMCFileSubscription
def createFileFromDataStructsFile(self, file, jobID):
"""
_createFileFromDataStructsFile_
This function will create a WMBS File given a DataStructs file
"""
wmbsFile = File()
wmbsFile.update(file)
if isinstance(file["locations"], set):
pnn = list(file["locations"])[0]
elif isinstance(file["locations"], list):
if len(file['locations']) > 1:
logging.error("Have more then one location for a file in job %i" % (jobID))
logging.error("Choosing location %s" % (file['locations'][0]))
pnn = file["locations"][0]
else:
pnn = file["locations"]
wmbsFile["locations"] = set()
if pnn != None:
wmbsFile.setLocation(pnn = pnn, immediateSave = False)
wmbsFile['jid'] = jobID
return wmbsFile
def createJob(self, streamerList, jobEvents, jobSize, memoryRequirement, numberOfCores = 1):
"""
_createJob_
"""
if not self.createdGroup:
self.newGroup()
self.createdGroup = True
self.newJob(name = "%s-%s" % (self.jobNamePrefix, makeUUID()))
if numberOfCores > 1:
self.currentJob.addBaggageParameter("numberOfCores", numberOfCores)
for streamer in streamerList:
f = File(id = streamer['id'],
lfn = streamer['lfn'])
f.setLocation(streamer['location'], immediateSave = False)
self.currentJob.addFile(f)
# job time based on
# - 5 min initialization
# - 0.5MB/s repack speed
# - checksum calculation at 5MB/s
# - stageout at 5MB/s
# job disk based on
# - RAW on local disk (factor 1)
jobTime = 300 + jobSize/500000 + (jobSize*2)/5000000
self.currentJob.addResourceEstimates(jobTime = jobTime, disk = jobSize/1024, memory = memoryRequirement)
return
def injectFilesFromDBS(inputFileset, datasetPath, runsWhiteList=[]):
"""
_injectFilesFromDBS_
"""
print "injecting files from %s into %s, please wait..." % (datasetPath, inputFileset.name)
args={}
args["url"] = "http://cmsdbsprod.cern.ch/cms_dbs_prod_global/servlet/DBSServlet"
args["version"] = "DBS_2_1_1"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFiles(path = datasetPath, retriveList = ["retrive_lumi", "retrive_run"])
print " found %d files, inserting into wmbs..." % (len(dbsResults))
for dbsResult in dbsResults:
if runsWhiteList and str(dbsResult["LumiList"][0]["RunNumber"]) not in runsWhiteList:
continue
myFile = File(lfn = dbsResult["LogicalFileName"], size = dbsResult["FileSize"],
events = dbsResult["NumberOfEvents"], checksums = {"cksum": dbsResult["Checksum"]},
locations = "cmssrm.fnal.gov", merged = True)
myRun = Run(runNumber = dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.lumis.append(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
if len(inputFileset) < 1:
raise Exception, "No files were selected!"
inputFileset.commit()
inputFileset.markOpen(False)
return
def injectFilesFromDBS(inputFileset, datasetPath):
"""
_injectFilesFromDBS_
"""
print "injecting files from %s into %s, please wait..." % (datasetPath, inputFileset.name)
args={}
args["url"] = "https://cmsweb.cern.ch/dbs/prod/global/DBSReader"
args["version"] = "DBS_2_0_9"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFiles(path = datasetPath, retriveList = ["retrive_block","retrive_lumi", "retrive_run"])
# NOTE : this is to limit the number of jobs to create ... simply using first 10 files get for the needed dataset
dbsResults =dbsResults[0:2]
print " found %d files, inserting into wmbs..." % (len(dbsResults))
for dbsResult in dbsResults:
myFile = File(lfn = dbsResult["LogicalFileName"], size = dbsResult["FileSize"],
events = dbsResult["NumberOfEvents"], checksums = {"cksum": dbsResult["Checksum"]},
locations = set(['srm.ciemat.es','storm-se-01.ba.infn.it','storage01.lcg.cscs.ch']))
myRun = Run(runNumber = dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.lumis.append(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
inputFileset.commit()
inputFileset.markOpen(False)
return
def testFilesWithoutOtherSubscriptions(self):
"""
_testFilesWithoutOtherSubscriptions_
Test the case where files only in the delete subscription
can happen if cleanup of the other subscriptions is fast
"""
testWorkflowA = Workflow(spec="specA.xml", owner="Steve", name="wfA", task="Test")
testWorkflowA.create()
testFileset = Fileset(name="TestFileset")
testFileset.create()
allFiles = []
for i in range(500):
testFile = File(str(i), size=1000, events=100, locations=set(["somese.cern.ch"]))
testFile.create()
allFiles.append(testFile)
testFileset.addFile(testFile)
testFileset.commit()
testSubscriptionA = Subscription(
fileset=testFileset, workflow=testWorkflowA, split_algo="SiblingProcessingBased", type="Processing"
)
testSubscriptionA.create()
splitter = SplitterFactory()
deleteFactoryA = splitter(package="WMCore.WMBS", subscription=testSubscriptionA)
result = deleteFactoryA(files_per_job=50)
self.assertEqual(len(result), 1, "Error: Wrong number of job groups returned.")
self.assertEqual(len(result[0].jobs), 10, "Error: Wrong number of jobs returned.")
return
def createJob(self, streamerList, jobEvents, jobSize):
"""
_createJob_
create an express job processing
the passed in list of streamers
"""
if not self.createdGroup:
self.newGroup()
self.createdGroup = True
self.newJob(name = "%s-%s" % (self.jobNamePrefix, makeUUID()))
for streamer in streamerList:
f = File(id = streamer['id'],
lfn = streamer['lfn'])
f.setLocation(streamer['location'], immediateSave = False)
self.currentJob.addFile(f)
# job time based on
# - 5 min initialization
# - 0.5MB/s repack speed
# - 45s/evt reco speed
# - checksum calculation at 5MB/s (twice)
# - stageout at 5MB/s
# job disk based on
# - streamer on local disk (factor 1)
# - RAW on local disk (factor 1)
# - FEVT/ALCARECO/DQM on local disk (factor 4)
jobTime = 300 + jobSize/500000 + jobEvents*45 + (jobSize*4*3)/5000000
self.currentJob.addResourceEstimates(jobTime = jobTime, disk = (jobSize*6)/1024)
return
def _addACDCFileToWMBSFile(self, acdcFile, inFileset = True):
"""
"""
wmbsParents = []
#pass empty check sum since it won't be updated to dbs anyway
checksums = {}
wmbsFile = File(lfn = str(acdcFile["lfn"]),
size = acdcFile["size"],
events = acdcFile["events"],
checksums = checksums,
#TODO: need to get list of parent lfn
parents = acdcFile["parents"],
locations = acdcFile["locations"],
merged = acdcFile.get('merged', True))
## TODO need to get the lumi lists
for run in acdcFile['runs']:
wmbsFile.addRun(run)
dbsFile = self._convertACDCFileToDBSFile(acdcFile)
self._addToDBSBuffer(dbsFile, checksums, acdcFile["locations"])
logging.info("WMBS File: %s\n on Location: %s"
% (wmbsFile['lfn'], wmbsFile['locations']))
if inFileset:
wmbsFile['inFileset'] = True
else:
wmbsFile['inFileset'] = False
self.wmbsFilesToCreate.append(wmbsFile)
return wmbsFile
def createLargeFileBlock(self):
"""
_createLargeFileBlock_
Creates a large group of files for testing
"""
testFileset = Fileset(name = "TestFilesetX")
testFileset.create()
for i in range(5000):
newFile = File(makeUUID(), size = 1000, events = 100,
locations = set(["somese.cern.ch"]))
newFile.create()
testFileset.addFile(newFile)
testFileset.commit()
testWorkflow = Workflow(spec = "spec.xml", owner = "mnorman",
name = "wf003", task="Test" )
testWorkflow.create()
largeSubscription = Subscription(fileset = testFileset,
workflow = testWorkflow,
split_algo = "FileBased",
type = "Processing")
largeSubscription.create()
return largeSubscription
def createSingleJobWorkflow(self):
"""
Create a workflow with one jobs and two files and store the results in instance variables
"""
self.testWorkflow = Workflow(spec="spec.xml", owner="Simon", name="wf001", task="Test")
self.testWorkflow.create()
testWMBSFileset = Fileset(name="TestFileset")
testWMBSFileset.create()
testSubscription = Subscription(fileset=testWMBSFileset, workflow=self.testWorkflow)
testSubscription.create()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
self.testFileA = File(lfn="/this/is/a/lfnA", size=1024, events=10)
self.testFileA.addRun(Run(1, *[45]))
self.testFileB = File(lfn="/this/is/a/lfnB", size=1024, events=10)
self.testFileB.addRun(Run(1, *[46]))
self.testFileA.create()
self.testFileB.create()
self.testJob = Job(name="TestJob", files=[self.testFileA, self.testFileB])
self.testJob.create(group=testJobGroup)
self.testJob.associateFiles()
def createJob(self, streamerList, jobEvents, jobSize, timePerEvent, sizePerEvent, memoryRequirement):
"""
_createJob_
create an express job processing
the passed in list of streamers
"""
if not self.createdGroup:
self.newGroup()
self.createdGroup = True
self.newJob(name = "%s-%s" % (self.jobNamePrefix, makeUUID()))
for streamer in streamerList:
f = File(id = streamer['id'],
lfn = streamer['lfn'])
f.setLocation(streamer['location'], immediateSave = False)
self.currentJob.addFile(f)
# job time based on
# - 5 min initialization (twice)
# - 0.5MB/s repack speed
# - reco with timePerEvent
# - checksum calculation at 5MB/s
# - stageout at 5MB/s
# job disk based on
# - streamer or RAW on local disk (factor 1)
# - FEVT/ALCARECO/DQM on local disk (sizePerEvent)
jobTime = 600 + jobSize/500000 + jobEvents*timePerEvent + (jobEvents*sizePerEvent*2)/5000000
self.currentJob.addResourceEstimates(jobTime = min(jobTime, 47*3600),
disk = min(jobSize/1024 + jobEvents*sizePerEvent, 20000000),
memory = memoryRequirement)
return
def createJob(self, fileList, jobSize):
"""
_createJob_
create an express merge job for
the passed in list of files
"""
if not self.createdGroup:
self.newGroup()
self.createdGroup = True
self.newJob(name = "%s-%s" % (self.jobNamePrefix, makeUUID()))
largestFile = 0
for fileInfo in fileList:
largestFile = max(largestFile, fileInfo['filesize'])
f = File(id = fileInfo['id'],
lfn = fileInfo['lfn'])
f.setLocation(fileInfo['location'], immediateSave = False)
self.currentJob.addFile(f)
# job time based on
# - 5 min initialization
# - 5MB/s merge speed
# - checksum calculation at 5MB/s (twice)
# - stageout at 5MB/s
# job disk based on
# - input for largest file on local disk
# - output on local disk (factor 1)
jobTime = 300 + (jobSize*4)/5000000
self.currentJob.addResourceEstimates(jobTime = jobTime, disk = (jobSize+largestFile)/1024)
return
def testD_NonContinuousLumis(self):
"""
_NonContinuousLumis_
Test and see if LumiBased can work when the lumis are non continuous
"""
baseName = makeUUID()
nFiles = 10
testFileset = Fileset(name = baseName)
testFileset.create()
for i in range(nFiles):
newFile = File(lfn = '%s_%i' % (baseName, i), size = 1000,
events = 100, locations = "somese.cern.ch")
# Set to two non-continuous lumi numbers
lumis = [100 + i, 200 + i]
newFile.addRun(Run(i, *lumis))
newFile.create()
testFileset.addFile(newFile)
testFileset.commit()
testSubscription = Subscription(fileset = testFileset,
workflow = self.testWorkflow,
split_algo = "LumiBased",
type = "Processing")
testSubscription.create()
splitter = SplitterFactory()
jobFactory = splitter(package = "WMCore.WMBS",
subscription = testSubscription)
jobGroups = jobFactory(lumis_per_job = 2,
halt_job_on_file_boundaries = False,
splitOnRun = False,
performance = self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 10)
for j in jobs:
runs = j['mask'].getRunAndLumis()
for r in runs.keys():
self.assertEqual(len(runs[r]), 2)
for l in runs[r]:
# Each run should have two lumis
# Each lumi should be of form [x, x]
# meaning that the first and last lumis are the same
self.assertEqual(len(l), 2)
self.assertEqual(l[0], l[1])
self.assertEqual(j['estimatedJobTime'], 100 * 12)
self.assertEqual(j['estimatedDiskUsage'], 100 * 400)
self.assertEqual(j['estimatedMemoryUsage'], 2300)
return
def loadFiles(self, size=10):
"""
_loadFiles_
Grab some files from the resultProxy
Should handle multiple proxies. Not really sure about that
"""
if len(self.proxies) < 1:
# Well, you don't have any proxies.
# This is what happens when you ran out of files last time
logging.info("No additional files found; Ending.")
return set()
resultProxy = self.proxies[0]
rawResults = []
if type(resultProxy.keys) == list:
keys = resultProxy.keys
else:
keys = resultProxy.keys()
if type(keys) == set:
# If it's a set, handle it
keys = list(keys)
files = set()
while len(rawResults) < size and len(self.proxies) > 0:
length = size - len(rawResults)
newResults = resultProxy.fetchmany(size=length)
if len(newResults) < length:
# Assume we're all out
# Eliminate this proxy
self.proxies.remove(resultProxy)
rawResults.extend(newResults)
if rawResults == []:
# Nothing to do
return set()
fileList = self.formatDict(results=rawResults, keys=keys)
fileIDs = list(set([x["fileid"] for x in fileList]))
myThread = threading.currentThread()
fileInfoAct = self.daoFactory(classname="Files.GetForJobSplittingByID")
fileInfoDict = fileInfoAct.execute(file=fileIDs, conn=myThread.transaction.conn, transaction=True)
getLocAction = self.daoFactory(classname="Files.GetLocationBulk")
getLocDict = getLocAction.execute(files=fileIDs, conn=myThread.transaction.conn, transaction=True)
for fID in fileIDs:
fl = WMBSFile(id=fID)
fl.update(fileInfoDict[fID])
locations = getLocDict.get((fID), [])
for loc in locations:
fl.setLocation(loc, immediateSave=False)
files.add(fl)
return files
def testAddDupsToFileset(self):
"""
_AddToDupsFileset_
Verify the the dups version of the AddToFileset DAO will not add files
to a fileset if they're already associated to another fileset with the
same workflow.
"""
testWorkflowA = Workflow(spec = 'hello', owner = "mnorman",
name = "wf001", task="basicWorkload/Production")
testWorkflowA.create()
testWorkflowB = Workflow(spec = 'hello', owner = "mnorman",
name = "wf001", task="basicWorkload/Production2")
testWorkflowB.create()
testFilesetA = Fileset(name = "inputFilesetA")
testFilesetA.create()
testFilesetB = Fileset(name = "inputFilesetB")
testFilesetB.create()
testSubscriptionA = Subscription(workflow = testWorkflowA, fileset = testFilesetA)
testSubscriptionA.create()
testSubscriptionB = Subscription(workflow = testWorkflowB, fileset = testFilesetB)
testSubscriptionB.create()
testFileA = File(lfn = "/this/is/a/lfnA", size = 1024, events = 10)
testFileA.addRun(Run( 1, *[45]))
testFileA.create()
testFileB = File(lfn = "/this/is/a/lfnB", size = 1024, events = 10)
testFileB.addRun(Run( 1, *[45]))
testFileB.create()
addToFileset = self.daofactory(classname = "Files.AddDupsToFileset")
addToFileset.execute(file = [testFileA['lfn'], testFileB['lfn']],
fileset = testFilesetA.id, workflow = "wf001")
testFileset2 = Fileset(name = "inputFilesetA")
testFileset2.loadData()
self.assertEqual(len(testFileset2.files), 2)
for file in testFileset2.files:
self.assertTrue(file in [testFileA, testFileB])
# Check that adding twice doesn't crash
addToFileset.execute(file = [testFileA['lfn'], testFileB['lfn']],
fileset = testFilesetA.id, workflow = "wf001")
# Files should not get added to fileset B because fileset A is associated
# with wf001.
addToFileset.execute(file = [testFileA['lfn'], testFileB['lfn']],
fileset = testFilesetB.id, workflow = "wf001")
testFileset2 = Fileset(name = "inputFilesetB")
testFileset2.loadData()
self.assertEqual(len(testFileset2.files), 0)
return
def testMultipleLocations(self):
"""
_testMultipleLocations_
Verify that the sibling processing based algorithm doesn't create jobs
that run over files at multiple sites.
"""
testFile1 = File("testFile1", size=1000, events=100, locations=set(["somese2.cern.ch"]))
testFile1.create()
testFile2 = File("testFile2", size=1000, events=100, locations=set(["somese2.cern.ch"]))
testFile2.create()
testFile3 = File("testFile3", size=1000, events=100, locations=set(["somese2.cern.ch"]))
testFile3.create()
self.testFilesetA.addFile(testFile1)
self.testFilesetA.addFile(testFile2)
self.testFilesetA.addFile(testFile3)
self.testFilesetA.commit()
self.testFilesetA.markOpen(False)
self.testSubscriptionA.completeFiles([testFile1, testFile2, testFile3])
self.testSubscriptionA.completeFiles([self.testFileA, self.testFileB, self.testFileC])
splitter = SplitterFactory()
deleteFactoryA = splitter(package="WMCore.WMBS", subscription=self.deleteSubscriptionA)
result = deleteFactoryA(files_per_job=50)
assert len(result) == 2, "Error: Wrong number of jobgroups returned."
goldenFilesA = ["testFileA", "testFileB", "testFileC"]
goldenFilesB = ["testFile1", "testFile2", "testFile3"]
locations = {
"testFileA": "somese.cern.ch",
"testFileB": "somese.cern.ch",
"testFileC": "somese.cern.ch",
"testFile1": "somese2.cern.ch",
"testFile2": "somese2.cern.ch",
"testFile3": "somese2.cern.ch",
}
for jobGroup in result:
assert len(jobGroup.jobs) == 1, "Error: Wrong number of jobs in jobgroup."
assert len(jobGroup.jobs[0]["input_files"]) == 3, "Error: Wrong number of input files in job."
jobSite = list(jobGroup.jobs[0]["input_files"][0]["locations"])[0]
if jobSite == "somese.cern.ch":
goldenFiles = goldenFilesA
else:
goldenFiles = goldenFilesB
for jobFile in jobGroup.jobs[0]["input_files"]:
assert list(jobFile["locations"])[0] == locations[jobFile["lfn"]], "Error: Wrong site for file."
goldenFiles.remove(jobFile["lfn"])
assert len(goldenFiles) == 0, "Error: Files are missing."
return
def injectFile(self):
"""
_injectFile_
Inject a file into the periodic splitting input fileset.
"""
testFile = File(lfn = "/this/is/a/lfn%s" % time.time(), size = 1000,
events = 100, locations = set(["somese.cern.ch"]))
testFile.create()
self.testFileset.addFile(testFile)
self.testFileset.commit()
return
def createFile(self, lfn, events, run, lumis, location):
"""
_createFile_
Create a file for testing
"""
newFile = File(lfn=lfn, size=1000, events=events)
lumiList = []
for lumi in range(lumis):
lumiList.append((run * lumis) + lumi)
newFile.addRun(Run(run, *lumiList))
newFile.setLocation(location)
return newFile
def testCreateTransaction(self):
"""
_testCreateTransaction_
Begin a transaction and then create a file in the database. Afterwards,
rollback the transaction. Use the File class's exists() method to
to verify that the file doesn't exist before it was created, exists
after it was created and doesn't exist after the transaction was rolled
back.
"""
myThread = threading.currentThread()
myThread.transaction.begin()
testFile = File(lfn = "/this/is/a/lfn", size = 1024, events = 10, checksums={'cksum':1111})
assert testFile.exists() == False, \
"ERROR: File exists before it was created"
testFile.addRun(Run(1, *[45]))
testFile.create()
assert testFile.exists() > 0, \
"ERROR: File does not exist after it was created"
myThread.transaction.rollback()
assert testFile.exists() == False, \
"ERROR: File exists after transaction was rolled back."
return
def testLargeNumberOfFiles(self):
"""
_testLargeNumberOfFiles_
Setup a subscription with 500 files and verify that the splitting algo
works correctly.
"""
testWorkflowA = Workflow(spec = "specA.xml", owner = "Steve",
name = "wfA", task = "Test")
testWorkflowA.create()
testWorkflowB = Workflow(spec = "specB.xml", owner = "Steve",
name = "wfB", task = "Test")
testWorkflowB.create()
testFileset = Fileset(name = "TestFileset")
testFileset.create()
allFiles = []
for i in range(500):
testFile = File(str(i), size = 1000, events = 100,
locations = set(["T2_CH_CERN"]))
testFile.create()
allFiles.append(testFile)
testFileset.addFile(testFile)
testFileset.commit()
testSubscriptionA = Subscription(fileset = testFileset,
workflow = testWorkflowA,
split_algo = "FileBased",
type = "Processing")
testSubscriptionA.create()
testSubscriptionB = Subscription(fileset = testFileset,
workflow = testWorkflowB,
split_algo = "SiblingProcessingBased",
type = "Processing")
testSubscriptionB.create()
testSubscriptionA.completeFiles(allFiles)
splitter = SplitterFactory()
deleteFactoryA = splitter(package = "WMCore.WMBS",
subscription = testSubscriptionB)
result = deleteFactoryA(files_per_job = 50)
self.assertEqual(len(result), 1,
"Error: Wrong number of job groups returned.")
self.assertEqual(len(result[0].jobs), 10,
"Error: Wrong number of jobs returned.")
return
def testCreateWithLocation(self):
"""
_testCreateWithLocation_
Create a file and add a couple locations. Load the file from the
database to make sure that the locations were set correctly.
"""
testFileA = File(lfn = "/this/is/a/lfn", size = 1024, events = 10,
checksums = {'cksum':1},
locations = set(["se1.fnal.gov", "se1.cern.ch"]))
testFileA.addRun(Run( 1, *[45]))
testFileA.create()
testFileB = File(id = testFileA["id"])
testFileB.loadData()
goldenLocations = ["se1.fnal.gov", "se1.cern.ch"]
for location in testFileB["locations"]:
assert location in goldenLocations, \
"ERROR: Unknown file location"
goldenLocations.remove(location)
assert len(goldenLocations) == 0, \
"ERROR: Some locations are missing"
return
def testLoadOutputID(self):
"""
_testLoadOutputID_
Test whether we can load an output ID for a job
"""
testWorkflow = Workflow(spec="spec.xml",
owner="Steve",
name="wf001",
task="Test")
testWorkflow.create()
testFileset = Fileset(name="TestFileset")
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow)
testSubscription.create()
testFileA = File(lfn=makeUUID(), locations="test.site.ch")
testFileB = File(lfn=makeUUID(), locations="test.site.ch")
testFileA.create()
testFileB.create()
testFileset.addFile([testFileA, testFileB])
testFileset.commit()
testSubscription.acquireFiles([testFileA, testFileB])
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testJob = Job()
testJob.create(group=testJobGroup)
self.assertEqual(testJob.loadOutputID(), testJobGroup.output.id)
return
def filesOfStatus(self, status, limit = 0, loadChecksums = True, doingJobSplitting = False):
"""
_filesOfStatus_
Return a Set of File objects that have the given status with respect
to this subscription.
"""
existingTransaction = self.beginTransaction()
status = status.title()
files = set()
if limit > 0:
action = self.daofactory(classname = "Subscriptions.Get%sFilesByLimit" % status)
fileList = action.execute(self["id"], limit, conn = self.getDBConn(),
transaction = self.existingTransaction())
else:
action = self.daofactory(classname = "Subscriptions.Get%sFiles" % status)
fileList = action.execute(self["id"], conn = self.getDBConn(),
transaction = self.existingTransaction())
if doingJobSplitting:
fileInfoAct = self.daofactory(classname = "Files.GetForJobSplittingByID")
else:
fileInfoAct = self.daofactory(classname = "Files.GetByID")
fileInfoDict = fileInfoAct.execute(file = [x["file"] for x in fileList],
conn = self.getDBConn(),
transaction = self.existingTransaction())
#Run through all files
for f in fileList:
fl = File(id = f['file'])
if loadChecksums:
fl.loadChecksum()
fl.update(fileInfoDict[f['file']])
if 'locations' in f.keys():
fl.setLocation(f['locations'], immediateSave = False)
files.add(fl)
self.commitTransaction(existingTransaction)
return files
def test01(self):
"""
_test01_
Test size and event triggers for single lumis (they are ignored)
Test latency trigger (timed out)
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1]:
for i in range(2):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomePNN", immediateSave=False)
newFile.create()
self.fileset2.addFile(newFile)
self.fileset2.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription2)
mySplitArgs['maxInputSize'] = 1
mySplitArgs['maxInputFiles'] = 1
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
time.sleep(1)
mySplitArgs['maxLatency'] = 1
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create a single job")
return
def test00(self):
"""
_test00_
Test that the job name prefix feature works
Test latency trigger (wait and 0)
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1]:
for i in range(2):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomePNN", immediateSave=False)
newFile.create()
self.fileset2.addFile(newFile)
self.fileset2.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription2)
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
mySplitArgs['maxLatency'] = 0
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create a single job")
job = jobGroups[0].jobs[0]
self.assertTrue(job['name'].startswith("ExpressMerge-"),
"ERROR: Job has wrong name")
return
def test02(self):
"""
_test02_
Test multi lumis
"""
insertClosedLumiBinds = []
for lumi in [1,2]:
filecount = 1
for i in range(filecount):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave = False)
newFile.create()
self.fileset1.addFile(newFile)
insertClosedLumiBinds.append( { 'RUN' : 1,
'LUMI' : lumi,
'STREAM' : "Express",
'FILECOUNT' : filecount,
'INSERT_TIME' : self.currentTime,
'CLOSE_TIME' : self.currentTime } )
self.fileset1.commit()
jobFactory = self.splitterFactory(package = "WMCore.WMBS",
subscription = self.subscription1)
self.insertClosedLumiDAO.execute(binds = insertClosedLumiBinds,
transaction = False)
self.releaseExpressDAO.execute(binds = { 'RUN' : 1 }, transaction = False)
jobGroups = jobFactory(maxInputEvents = 100)
self.assertEqual(len(jobGroups[0].jobs), 2,
"ERROR: JobFactory didn't create two jobs")
self.assertEqual(self.getNumActiveSplitLumis(), 0,
"ERROR: Split lumis were created")
return
def createFile(self, lfn, events, run, lumis, location):
"""
_createFile_
Create a file for testing
"""
newFile = File(lfn = lfn, size = 1000,
events = events)
lumiList = []
for lumi in range(lumis):
lumiList.append((run * lumis) + lumi)
newFile.addRun(Run(run, *lumiList))
newFile.setLocation(location)
return newFile
def testLocationMerging(self):
"""
_testLocationMerging_
Verify that files residing on different SEs are not merged together in
the same job.
"""
self.stuffWMBS()
locationAction = self.daoFactory(classname="Locations.New")
locationAction.execute(siteName="s2", seName="somese2.cern.ch")
fileSite2 = File(lfn="fileSite2",
size=4098,
events=1024,
first_event=0,
locations=set(["somese2.cern.ch"]))
fileSite2.addRun(Run(1, *[46]))
fileSite2.create()
self.mergeFileset.addFile(fileSite2)
self.mergeFileset.commit()
splitter = SplitterFactory()
jobFactory = splitter(package="WMCore.WMBS",
subscription=self.mergeSubscription)
result = jobFactory(min_merge_size=4097,
max_merge_size=99999999,
max_merge_events=999999999,
merge_across_runs=False)
assert len(result) == 1, \
"ERROR: More than one JobGroup returned."
assert len(result[0].jobs) == 3, \
"ERROR: Three jobs should have been returned."
for job in result[0].jobs:
firstInputFile = job.getFiles()[0]
baseLocation = list(firstInputFile["locations"])[0]
for inputFile in job.getFiles():
assert len(inputFile["locations"]) == 1, \
"Error: Wrong number of locations"
assert list(inputFile["locations"])[0] == baseLocation, \
"Error: Wrong location."
return
def testLocationMerging(self):
"""
_testLocationMerging_
Verify that files residing on different SEs are not merged together in
the same job.
"""
self.stuffWMBS()
locationAction = self.daoFactory(classname="Locations.New")
locationAction.execute(siteName="T1_UK_RAL", pnn="T1_UK_RAL_Disk")
fileSite2 = File(lfn="fileRAL",
size=4098,
events=1024,
first_event=0,
locations=set(["T1_UK_RAL_Disk"]))
fileSite2.addRun(Run(1, *[46]))
fileSite2.create()
self.mergeFileset.addFile(fileSite2)
self.mergeFileset.commit()
splitter = SplitterFactory()
jobFactory = splitter(package="WMCore.WMBS",
subscription=self.mergeSubscription)
result = jobFactory(min_merge_size=4097,
max_merge_size=99999999,
max_merge_events=999999999,
merge_across_runs=False)
assert len(result) == 1, \
"ERROR: More than one JobGroup returned."
assert len(result[0].jobs) == 3, \
"ERROR: Three jobs should have been returned."
ralJobs = 0
fnalJobs = 0
for job in result[0].jobs:
if job["possiblePSN"] == set(["T1_UK_RAL"]):
ralJobs += 1
elif job["possiblePSN"] == set(["T1_US_FNAL"]):
fnalJobs += 1
self.assertEqual(ralJobs, 1)
self.assertEqual(fnalJobs, 2)
return
def injectFilesFromDBS(inputFileset, datasetPath):
"""
_injectFilesFromDBS_
"""
print "injecting files from %s into %s, please wait..." % (
datasetPath, inputFileset.name)
args = {}
args[
"url"] = "http://cmsdbsprod.cern.ch/cms_dbs_prod_global/servlet/DBSServlet"
args["version"] = "DBS_2_0_9"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFiles(path=datasetPath,
retriveList=["retrive_lumi", "retrive_run"])
dbsResults = dbsResults[0:10]
print " found %d files, inserting into wmbs..." % (len(dbsResults))
for dbsResult in dbsResults:
myFile = File(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations="cmssrm.fnal.gov",
merged=True)
myRun = Run(runNumber=dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.lumis.append(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
dbsFile = DBSBufferFile(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations="cmssrm.fnal.gov",
status="LOCAL")
dbsFile.setDatasetPath(datasetPath)
dbsFile.setAlgorithm(appName="cmsRun",
appVer="Unknown",
appFam="Unknown",
psetHash="Unknown",
configContent="Unknown")
dbsFile.create()
inputFileset.commit()
inputFileset.markOpen(False)
return
def addMCFakeFile(self):
"""Add a fake file for wmbs to run production over"""
needed = ['FirstEvent', 'FirstLumi', 'FirstRun', 'LastEvent', 'LastLumi', 'LastRun']
for key in needed:
if self.mask and self.mask.get(key) is None:
msg = 'Invalid value "%s" for %s' % (self.mask.get(key), key)
raise WorkQueueWMBSException(msg)
locations = set()
for site in self.getLocations.execute(conn=self.getDBConn(),
transaction=self.existingTransaction()):
try:
siteInfo = self.getLocationInfo.execute(site, conn=self.getDBConn(),
transaction=self.existingTransaction())
if not siteInfo:
self.logger.info('Skipping MonteCarlo injection to site "%s" as unknown to wmbs' % site)
continue
locations.add(siteInfo[0]['pnn'])
except Exception as ex:
self.logger.error('Error getting storage element for "%s": %s' % (site, str(ex)))
if not locations:
msg = 'No locations to inject Monte Carlo work to, unable to proceed'
raise WorkQueueWMBSException(msg)
mcFakeFileName = ("MCFakeFile-%s" % self.topLevelFileset.name).encode('ascii', 'ignore')
wmbsFile = File(lfn=mcFakeFileName,
first_event=self.mask['FirstEvent'],
last_event=self.mask['LastEvent'],
events=self.mask['LastEvent'] - self.mask['FirstEvent'] + 1, # inclusive range
locations=locations,
merged=False, # merged causes dbs parentage relation
)
if self.mask:
lumis = range(self.mask['FirstLumi'], self.mask['LastLumi'] + 1) # inclusive range
wmbsFile.addRun(Run(self.mask['FirstRun'], *lumis)) # assume run number static
else:
wmbsFile.addRun(Run(1, 1))
wmbsFile['inFileset'] = True # file is not a parent
logging.info("WMBS MC Fake File: %s on Location: %s", wmbsFile['lfn'], wmbsFile['newlocations'])
self.wmbsFilesToCreate.append(wmbsFile)
totalFiles = self.topLevelFileset.addFilesToWMBSInBulk(self.wmbsFilesToCreate,
self.wmSpec.name(),
isDBS=self.isDBS)
self.topLevelFileset.markOpen(False)
return totalFiles
def test07(self):
"""
_test07_
Test over merge
one small lumi, one large lumi (small below min size,
large below max size, but both together above max size)
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1, 2]:
for i in range(2):
newFile = File(makeUUID(), size = 1000 * lumi * lumi, events = 100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomePNN", immediateSave = False)
newFile.create()
self.fileset2.addFile(newFile)
self.fileset2.commit()
jobFactory = self.splitterFactory(package = "WMCore.WMBS",
subscription = self.subscription2)
mySplitArgs['minInputSize'] = 3000
mySplitArgs['maxInputSize'] = 9000
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create one job")
job = jobGroups[0].jobs[0]
self.assertEqual(len(job.getFiles()), 4,
"ERROR: Job does not process 4 files")
return
def createFileCollection(self,
name,
nSubs,
nFiles,
workflowURL='test',
site=None):
"""
_createFileCollection_
Create a collection of files for splitting into jobs
"""
myThread = threading.currentThread()
testWorkflow = Workflow(spec=workflowURL,
owner="mnorman",
name=name,
task="/TestWorkload/ReReco")
testWorkflow.create()
for sub in range(nSubs):
nameStr = '%s-%i' % (name, sub)
testFileset = Fileset(name=nameStr)
testFileset.create()
for f in range(nFiles):
# pick a random site
if not site:
tmpSite = 'se.%s' % (random.choice(self.sites))
else:
tmpSite = 'se.%s' % (site)
testFile = File(lfn="/lfn/%s/%i" % (nameStr, f),
size=1024,
events=10)
testFile.setLocation(tmpSite)
testFile.create()
testFileset.addFile(testFile)
testFileset.commit()
testFileset.markOpen(isOpen=0)
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow,
type="Processing",
split_algo="FileBased")
testSubscription.create()
return
def test03(self):
"""
_test03_
Test input size threshold on multi lumis
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1, 2]:
for i in range(2):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomePNN", immediateSave=False)
newFile.create()
self.fileset2.addFile(newFile)
self.fileset2.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription2)
mySplitArgs['maxInputSize'] = 1
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
time.sleep(1)
mySplitArgs['maxLatency'] = 1
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 2,
"ERROR: JobFactory didn't create two jobs")
return
def setUp(self):
"""
_setUp_
"""
self.testInit = TestInit(__file__)
self.testInit.setLogging()
self.testInit.setDatabaseConnection()
self.testInit.setSchema(customModules=["WMCore.WMBS"])
self.splitterFactory = SplitterFactory(package="WMCore.JobSplitting")
myThread = threading.currentThread()
self.myThread = myThread
daoFactory = DAOFactory(package="WMCore.WMBS",
logger=logging,
dbinterface=myThread.dbi)
self.WMBSFactory = daoFactory
config = self.getConfig()
self.changer = ChangeState(config)
myResourceControl = ResourceControl()
myResourceControl.insertSite("T1_US_FNAL", 10, 20, "T1_US_FNAL_Disk", "T1_US_FNAL")
myResourceControl.insertSite("T1_US_FNAL", 10, 20, "T3_US_FNALLPC", "T1_US_FNAL")
myResourceControl.insertSite("T2_CH_CERN", 10, 20, "T2_CH_CERN", "T2_CH_CERN")
self.fileset1 = Fileset(name="TestFileset1")
for fileNum in range(11):
newFile = File("/some/file/name%d" % fileNum, size=1000, events=100)
newFile.addRun(Run(1, *[1]))
newFile.setLocation('T1_US_FNAL_Disk')
self.fileset1.addFile(newFile)
self.fileset1.create()
workflow1 = Workflow(spec="spec.xml", owner="hufnagel", name="TestWorkflow1", task="Test")
workflow1.create()
self.subscription1 = Subscription(fileset=self.fileset1,
workflow=workflow1,
split_algo="Harvest",
type="Harvesting")
self.subscription1.create()
self.configFile = EmulatorSetup.setupWMAgentConfig()
return
def injectFilesFromDBS(inputFileset, datasetPath):
"""
_injectFilesFromDBS_
"""
print("injecting files from %s into %s, please wait..." %
(datasetPath, inputFileset.name))
args = {}
args["url"] = "https://cmsweb.cern.ch/dbs/prod/global/DBSReader"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFileArray(
path=datasetPath, retriveList=["retrive_lumi", "retrive_run"])
# Limiter on number of files
dbsResults = dbsResults[0:20]
print(" found %d files, inserting into wmbs..." % (len(dbsResults)))
for dbsResult in dbsResults:
myFile = File(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations="cmssrm.fnal.gov",
merged=True)
myRun = Run(runNumber=dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.appendLumi(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
dbsFile = DBSBufferFile(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations="cmssrm.fnal.gov",
status="NOTUPLOADED")
dbsFile.setDatasetPath(datasetPath)
dbsFile.setAlgorithm(appName="cmsRun",
appVer="Unknown",
appFam="Unknown",
psetHash="Unknown",
configContent="Unknown")
dbsFile.create()
inputFileset.commit()
inputFileset.markOpen(False)
return
def createFile(lfn, events, run, lumis, location, lumiMultiplier=None):
"""
_createFile_
Create a file for testing
"""
if lumiMultiplier is None:
lumiMultiplier = run
newFile = File(lfn=lfn, size=1000, events=events)
lumiList = []
for lumi in range(lumis):
lumiList.append((lumiMultiplier * lumis) + lumi)
newFile.addRun(Run(run, *lumiList))
newFile.setLocation(location)
return newFile
def fileLumiMaps(self, filesAtLocation=None, getParents=False, lumiMask=None):
"""
Args:
filesAtLocation: the list of file objects at a particular location
getParents: Attach a list of parents files to the file
lumiMask: the lumiMask for the step. Files excluded by the lumiMask are dropped
Returns:
lumisByFile: LumiList for each LFN
eventsByLumi: Estimate (for now) of how many events are in each lumi
"""
lumisByFile = {}
eventsByLumi = defaultdict(lambda: defaultdict(float))
self.filesByLumi = defaultdict(lambda: defaultdict(list))
for fileObj in filesAtLocation:
lfn = fileObj['lfn']
eventsInFile = fileObj['events']
lumisInFile = self.countLumis(fileObj['runs'])
if getParents:
parentLFNs = self.findParent(lfn=lfn)
for lfn in parentLFNs:
parent = File(lfn=lfn)
fileObj['parents'].add(parent)
runsAndLumis = {str(runLumi.run): runLumi.lumis for runLumi in fileObj['runs']}
lumiList = LumiList(runsAndLumis=runsAndLumis)
if lumiMask: # Apply the mask if there is one
lumiList &= lumiMask
if lumiList: # Skip files with no lumis of interest
lumisByFile[lfn] = lumiList
for runLumi in fileObj['runs']:
run = runLumi.run
for lumi in runLumi.lumis:
if (runLumi.run, lumi) in lumiList:
self.filesByLumi[run][lumi].append(fileObj)
eventsByLumi[run][lumi] = round(eventsInFile / lumisInFile)
return lumisByFile, eventsByLumi
def addMCFakeFile(self):
"""Add a fake file for wmbs to run production over"""
needed = ['FirstEvent', 'FirstLumi', 'FirstRun', 'LastEvent', 'LastLumi', 'LastRun']
for key in needed:
if self.mask and self.mask.get(key) is None:
msg = 'Invalid value "%s" for %s' % (self.mask.get(key), key)
raise WorkQueueWMBSException(msg)
locations = set()
siteInfo = self.getLocationInfo.execute(conn=self.getDBConn(),
transaction=self.existingTransaction())
for site in siteInfo:
if site['pnn'] in self.commonLocation:
locations.add(site['pnn'])
if not locations:
msg = 'No locations to inject Monte Carlo work to, unable to proceed'
raise WorkQueueWMBSException(msg)
mcFakeFileName = "MCFakeFile-%s" % self.topLevelFileset.name
wmbsFile = File(lfn=mcFakeFileName,
first_event=self.mask['FirstEvent'],
last_event=self.mask['LastEvent'],
events=self.mask['LastEvent'] - self.mask['FirstEvent'] + 1, # inclusive range
locations=locations,
merged=False, # merged causes dbs parentage relation
)
if self.mask:
lumis = list(range(self.mask['FirstLumi'], self.mask['LastLumi'] + 1)) # inclusive range
wmbsFile.addRun(Run(self.mask['FirstRun'], *lumis)) # assume run number static
else:
wmbsFile.addRun(Run(1, 1))
wmbsFile['inFileset'] = True # file is not a parent
logging.debug("WMBS MC Fake File: %s on Location: %s", wmbsFile['lfn'], wmbsFile['newlocations'])
self.wmbsFilesToCreate.add(wmbsFile)
totalFiles = self.topLevelFileset.addFilesToWMBSInBulk(self.wmbsFilesToCreate,
self.wmSpec.name(),
isDBS=self.isDBS)
self.topLevelFileset.markOpen(False)
return totalFiles
def generateFakeMCFile(self,
numEvents=100,
firstEvent=1,
lastEvent=100,
firstLumi=1,
lastLumi=10,
index=1,
existingSub=None):
"""
_generateFakeMCFile_
Generates a fake MC file for testing production EventBased
creation of jobs, it creates a single file subscription if no
existing subscription is provided.
"""
# MC comes with MCFakeFile(s)
newFile = File("MCFakeFile-some-hash-%s" % str(index).zfill(5),
size=1000,
events=numEvents,
locations=set(["somese.cern.ch"]))
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
newFile.create()
if existingSub is None:
singleMCFileset = Fileset(name="MCTestFileset-%i" % index)
singleMCFileset.create()
singleMCFileset.addFile(newFile)
singleMCFileset.commit()
testWorkflow = Workflow(spec="spec.xml",
owner="Steve",
name="wf001",
task="Test")
testWorkflow.create()
singleMCFileSubscription = Subscription(fileset=singleMCFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Production")
singleMCFileSubscription.create()
return singleMCFileSubscription
else:
existingSub['fileset'].addFile(newFile)
existingSub['fileset'].commit()
return existingSub
def createJobCollection(self, name, nSubs, nFiles, workflowURL='test'):
"""
_createJobCollection_
Create a collection of jobs
"""
myThread = threading.currentThread()
testWorkflow = Workflow(spec=workflowURL,
owner="mnorman",
name=name,
task="/TestWorkload/ReReco")
testWorkflow.create()
for sub in range(nSubs):
nameStr = '%s-%i' % (name, sub)
myThread.transaction.begin()
testFileset = Fileset(name=nameStr)
testFileset.create()
for f in range(nFiles):
# pick a random site
site = random.choice(self.sites)
testFile = File(lfn="/lfn/%s/%i" % (nameStr, f),
size=1024,
events=10)
testFile.setLocation(site)
testFile.create()
testFileset.addFile(testFile)
testFileset.commit()
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow,
type="Processing",
split_algo="FileBased")
testSubscription.create()
myThread.transaction.commit()
return
def injectFilesFromDBS(inputFileset, datasetPath):
"""
_injectFilesFromDBS_
"""
print "injecting files from %s into %s, please wait..." % (
datasetPath, inputFileset.name)
args = {}
args[
"url"] = "http://cmsdbsprod.cern.ch/cms_dbs_prod_global/servlet/DBSServlet"
args["version"] = "DBS_2_0_9"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFiles(
path=datasetPath,
retriveList=["retrive_block", "retrive_lumi", "retrive_run"])
# NOTE : this is to limit the number of jobs to create ... simply using first 10 files get for the needed dataset
dbsResults = dbsResults[0:2]
print " found %d files, inserting into wmbs..." % (len(dbsResults))
for dbsResult in dbsResults:
myFile = File(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations=set([
'srm.ciemat.es', 'storm-se-01.ba.infn.it',
'storage01.lcg.cscs.ch'
]))
myRun = Run(runNumber=dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.lumis.append(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
inputFileset.commit()
inputFileset.markOpen(False)
return
def testAddRunSet(self):
"""
_testAddRunSet_
Test the ability to add run and lumi information to a file.
"""
testFile = File(lfn="/this/is/a/lfn",
size=1024,
events=10,
checksums={'cksum': 1},
locations="se1.fnal.gov")
testFile.create()
runSet = set()
runSet.add(Run(1, *[45]))
runSet.add(Run(2, *[67, 68]))
testFile.addRunSet(runSet)
assert (runSet - testFile["runs"]) == set(), \
"Error: addRunSet is not updating set correctly"
return
def injectFilesFromDBS(inputFileset, datasetPath, runsWhiteList=[]):
"""
_injectFilesFromDBS_
"""
print "injecting files from %s into %s, please wait..." % (
datasetPath, inputFileset.name)
args = {}
args[
"url"] = "http://cmsdbsprod.cern.ch/cms_dbs_prod_global/servlet/DBSServlet"
args["version"] = "DBS_2_1_1"
args["mode"] = "GET"
dbsApi = DbsApi(args)
dbsResults = dbsApi.listFiles(path=datasetPath,
retriveList=["retrive_lumi", "retrive_run"])
print " found %d files, inserting into wmbs..." % (len(dbsResults))
for dbsResult in dbsResults:
if runsWhiteList and str(
dbsResult["LumiList"][0]["RunNumber"]) not in runsWhiteList:
continue
myFile = File(lfn=dbsResult["LogicalFileName"],
size=dbsResult["FileSize"],
events=dbsResult["NumberOfEvents"],
checksums={"cksum": dbsResult["Checksum"]},
locations="cmssrm.fnal.gov",
merged=True)
myRun = Run(runNumber=dbsResult["LumiList"][0]["RunNumber"])
for lumi in dbsResult["LumiList"]:
myRun.lumis.append(lumi["LumiSectionNumber"])
myFile.addRun(myRun)
myFile.create()
inputFileset.addFile(myFile)
if len(inputFileset) < 1:
raise Exception, "No files were selected!"
inputFileset.commit()
inputFileset.markOpen(False)
return
def test06(self):
"""
_test06_
Test repacking of 3 lumis
2 small lumis (single job), followed by a big one (multiple jobs)
files for lumi 1 and 2 are below multi-lumi thresholds
files for lumi 3 are above single-lumi threshold
"""
mySplitArgs = self.splitArgs.copy()
insertClosedLumiBinds = []
for lumi in [1, 2, 3]:
filecount = 2
for i in range(filecount):
if lumi == 3:
nevents = 500
else:
nevents = 100
newFile = File(makeUUID(), size=1000, events=nevents)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave=False)
newFile.create()
self.fileset1.addFile(newFile)
insertClosedLumiBinds.append({
'RUN': 1,
'LUMI': lumi,
'STREAM': "A",
'FILECOUNT': filecount,
'INSERT_TIME': self.currentTime,
'CLOSE_TIME': self.currentTime
})
self.fileset1.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription1)
self.insertClosedLumiDAO.execute(binds=insertClosedLumiBinds,
transaction=False)
mySplitArgs['maxInputEvents'] = 900
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 3,
"ERROR: JobFactory didn't create three jobs")
self.assertEqual(len(jobGroups[0].jobs[0].getFiles()), 4,
"ERROR: first job does not process 4 files")
self.assertEqual(len(jobGroups[0].jobs[1].getFiles()), 1,
"ERROR: second job does not process 1 file")
self.assertEqual(len(jobGroups[0].jobs[2].getFiles()), 1,
"ERROR: third job does not process 1 file")
return
def test05(self):
"""
_test05_
Test repacking of multiple lumis with holes in the lumi sequence
Multi lumi input
"""
mySplitArgs = self.splitArgs.copy()
insertClosedLumiBinds = []
for lumi in [1, 2, 4]:
filecount = 2
for i in range(filecount):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave=False)
newFile.create()
self.fileset1.addFile(newFile)
insertClosedLumiBinds.append({
'RUN': 1,
'LUMI': lumi,
'STREAM': "A",
'FILECOUNT': filecount,
'INSERT_TIME': self.currentTime,
'CLOSE_TIME': self.currentTime
})
self.fileset1.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription1)
self.insertClosedLumiDAO.execute(binds=insertClosedLumiBinds,
transaction=False)
mySplitArgs['maxInputFiles'] = 5
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
self.insertClosedLumiDAO.execute(binds={
'RUN': 1,
'LUMI': 3,
'STREAM': "A",
'FILECOUNT': 0,
'INSERT_TIME': self.currentTime,
'CLOSE_TIME': self.currentTime
},
transaction=False)
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create one job")
self.assertEqual(len(jobGroups[0].jobs[0].getFiles()), 4,
"ERROR: first job does not process 4 files")
return
def test03(self):
"""
_test03_
Test single lumi event threshold
Single lumi input
"""
mySplitArgs = self.splitArgs.copy()
insertClosedLumiBinds = []
for lumi in [1]:
filecount = 8
for i in range(filecount):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave=False)
newFile.create()
self.fileset1.addFile(newFile)
insertClosedLumiBinds.append({
'RUN': 1,
'LUMI': lumi,
'STREAM': "A",
'FILECOUNT': filecount,
'INSERT_TIME': self.currentTime,
'CLOSE_TIME': self.currentTime
})
self.fileset1.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription1)
self.insertClosedLumiDAO.execute(binds=insertClosedLumiBinds,
transaction=False)
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
mySplitArgs['maxInputEvents'] = 650
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 2,
"ERROR: JobFactory didn't create two jobs")
job = jobGroups[0].jobs[0]
self.assertEqual(len(job.getFiles()), 6,
"ERROR: Job does not process 6 files")
job = jobGroups[0].jobs[1]
self.assertEqual(len(job.getFiles()), 2,
"ERROR: Job does not process 2 files")
self.assertEqual(self.getNumActiveSplitLumis(), 1,
"ERROR: Split lumis were not created")
return
def test00(self):
"""
_test00_
Test that the job name prefix feature works
Test multi lumi size threshold
Multi lumi input
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1, 2, 3, 4]:
filecount = 2
for i in range(filecount):
newFile = File(makeUUID(), size=1000, events=100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave=False)
newFile.create()
self.fileset1.addFile(newFile)
self.fileset1.commit()
jobFactory = self.splitterFactory(package="WMCore.WMBS",
subscription=self.subscription1)
mySplitArgs['maxSizeMultiLumi'] = self.splitArgs['maxSizeMultiLumi']
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
mySplitArgs['maxSizeMultiLumi'] = 5000
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create a single job")
job = jobGroups[0].jobs[0]
self.assertTrue(job['name'].startswith("Repack-"),
"ERROR: Job has wrong name")
self.assertEqual(len(job.getFiles()), 4,
"ERROR: Job does not process 4 files")
self.fileset1.markOpen(False)
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 1,
"ERROR: JobFactory didn't return one JobGroup")
self.assertEqual(len(jobGroups[0].jobs), 1,
"ERROR: JobFactory didn't create a single job")
job = jobGroups[0].jobs[0]
self.assertTrue(job['name'].startswith("Repack-"),
"ERROR: Job has wrong name")
self.assertEqual(len(job.getFiles()), 4,
"ERROR: Job does not process 4 files")
self.assertEqual(self.getNumActiveSplitLumis(), 0,
"ERROR: Split lumis were created")
return
def makeNJobs(self,
name,
task,
nJobs,
jobGroup,
fileset,
sub,
site=None,
bl=[],
wl=[]):
"""
_makeNJobs_
Make and return a WMBS Job and File
This handles all those damn add-ons
"""
# Set the CacheDir
cacheDir = os.path.join(self.testDir, 'CacheDir')
for n in range(nJobs):
# First make a file
#site = self.sites[0]
testFile = File(lfn="/singleLfn/%s/%s" % (name, n),
size=1024,
events=10)
if site:
testFile.setLocation(site)
else:
for tmpSite in self.sites:
testFile.setLocation('se.%s' % (tmpSite))
testFile.create()
fileset.addFile(testFile)
fileset.commit()
index = 0
for f in fileset.files:
index += 1
testJob = Job(name='%s-%i' % (name, index))
testJob.addFile(f)
testJob["location"] = f.getLocations()[0]
testJob['custom']['location'] = f.getLocations()[0]
testJob['task'] = task.getPathName()
testJob['sandbox'] = task.data.input.sandbox
testJob['spec'] = os.path.join(self.testDir, 'basicWorkload.pcl')
testJob['mask']['FirstEvent'] = 101
testJob['owner'] = 'tapas'
testJob["siteBlacklist"] = bl
testJob["siteWhitelist"] = wl
testJob['ownerDN'] = 'tapas'
testJob['ownerRole'] = 'cmsrole'
testJob['ownerGroup'] = 'phgroup'
jobCache = os.path.join(cacheDir, 'Sub_%i' % (sub),
'Job_%i' % (index))
os.makedirs(jobCache)
testJob.create(jobGroup)
testJob['cache_dir'] = jobCache
testJob.save()
jobGroup.add(testJob)
output = open(os.path.join(jobCache, 'job.pkl'), 'w')
pickle.dump(testJob, output)
output.close()
return testJob, testFile
def createTestJobGroup(self, nJobs, subType="Processing", retryOnce=False):
"""
_createTestJobGroup_
Creates a group of several jobs
"""
testWorkflow = Workflow(spec="spec.xml",
owner="Simon",
name=makeUUID(),
task="Test")
testWorkflow.create()
testWMBSFileset = Fileset(name="TestFileset")
testWMBSFileset.create()
testSubscription = Subscription(fileset=testWMBSFileset,
workflow=testWorkflow,
type=subType)
testSubscription.create()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testFileA = File(lfn="/this/is/a/lfnA", size=1024, events=10)
testFileA.addRun(Run(10, *[12312]))
testFileA.setLocation('malpaquet')
testFileB = File(lfn="/this/is/a/lfnB", size=1024, events=10)
testFileB.addRun(Run(10, *[12312]))
testFileA.setLocation('malpaquet')
testFileA.create()
testFileB.create()
for _ in range(0, nJobs):
testJob = Job(name=makeUUID())
testJob.addFile(testFileA)
testJob.addFile(testFileB)
testJob['cache_dir'] = os.path.join(self.testDir, testJob['name'])
os.mkdir(testJob['cache_dir'])
testJobGroup.add(testJob)
testJobGroup.commit()
if retryOnce:
self.increaseRetry.execute(testJobGroup.jobs)
return testJobGroup
def algorithm(self, *args, **kwargs):
"""
_algorithm_
Split up all the available files such that each job will process a
maximum of 'files_per_job'. If the 'files_per_job' parameters is not
passed in jobs will process a maximum of 10 files.
"""
filesPerJob = int(kwargs.get("files_per_job", 10))
jobsPerGroup = int(kwargs.get("jobs_per_group", 0))
runBoundaries = kwargs.get("respect_run_boundaries", False)
getParents = kwargs.get("include_parents", False)
filesInJob = 0
listOfFiles = []
# Grab the fileset
fileset = self.subscription.getFileset()
try:
fileset.load()
except AttributeError as ae:
pass
#Get a dictionary of sites, files
locationDict = self.sortByLocation()
for location in locationDict:
#Now we have all the files in a certain location
fileList = locationDict[location]
filesInJob = 0
jobsInGroup = 0
fileCounter = 0
if len(fileList) == 0:
#No files for this location
#This isn't supposed to happen, but better safe then sorry
continue
if len(fileList) < filesPerJob and fileset.open:
continue
self.newGroup()
jobRun = None
for f in fileList:
if getParents:
parentLFNs = self.findParent(lfn = f['lfn'])
for lfn in parentLFNs:
parent = File(lfn = lfn)
f['parents'].add(parent)
fileRun = f.get('minrun', None)
if filesInJob == 0 or filesInJob == filesPerJob or (runBoundaries and fileRun != jobRun):
if jobsPerGroup:
if jobsInGroup > jobsPerGroup:
self.newGroup()
jobsInGroup = 0
filesInJob = 0
# We've finished a job. Should we create another one?
if len(fileList) - fileCounter < filesPerJob and fileset.open:
# If we don't have the full job's worth of files
# and the fileset is still open
# then we shouldn't create a new job
continue
self.newJob(name = self.getJobName())
jobsInGroup += 1
jobRun = fileRun
filesInJob += 1
self.currentJob.addFile(f)
fileCounter += 1
listOfFiles.append(f)
return
