def testParallelProcessing(self):
"""
_testParallelProcessing_
Verify that merging works correctly when multiple processing
subscriptions are run over the same input files. The merging algorithm
should ignore processing jobs that feed into different merge
subscriptions.
"""
locationAction = self.daoFactory(classname="Locations.New")
locationAction.execute(siteName="T2_CH_CERN", pnn="T2_CH_CERN")
locationAction.execute(siteName="T1_US_FNAL", pnn="T2_CH_CERN")
mergeFilesetA = Fileset(name="mergeFilesetA")
mergeFilesetB = Fileset(name="mergeFilesetB")
mergeFilesetA.create()
mergeFilesetB.create()
mergeMergedFilesetA = Fileset(name="mergeMergedFilesetA")
mergeMergedFilesetB = Fileset(name="mergeMergedFilesetB")
mergeMergedFilesetA.create()
mergeMergedFilesetB.create()
mergeWorkflow = Workflow(name="mergeWorkflow",
spec="bogus",
owner="Steve",
task="Test")
mergeWorkflow.create()
mergeSubscriptionA = Subscription(fileset=mergeFilesetA,
workflow=mergeWorkflow,
split_algo="WMBSMergeBySize")
mergeSubscriptionB = Subscription(fileset=mergeFilesetB,
workflow=mergeWorkflow,
split_algo="WMBSMergeBySize")
mergeSubscriptionA.create()
mergeSubscriptionB.create()
inputFileset = Fileset(name="inputFileset")
inputFileset.create()
inputFileA = File(lfn="inputLFNA")
inputFileB = File(lfn="inputLFNB")
inputFileA.create()
inputFileB.create()
procWorkflowA = Workflow(name="procWorkflowA",
spec="bunk2",
owner="Steve",
task="Test")
procWorkflowA.create()
procWorkflowA.addOutput("output", mergeFilesetA, mergeMergedFilesetA)
procWorkflowB = Workflow(name="procWorkflowB",
spec="bunk3",
owner="Steve",
task="Test2")
procWorkflowB.create()
procWorkflowB.addOutput("output", mergeFilesetB, mergeMergedFilesetB)
procSubscriptionA = Subscription(fileset=inputFileset,
workflow=procWorkflowA,
split_algo="EventBased")
procSubscriptionA.create()
procSubscriptionB = Subscription(fileset=inputFileset,
workflow=procWorkflowB,
split_algo="EventBased")
procSubscriptionB.create()
jobGroupA = JobGroup(subscription=procSubscriptionA)
jobGroupA.create()
jobGroupB = JobGroup(subscription=procSubscriptionB)
jobGroupB.create()
changeStateDAO = self.daoFactory(classname="Jobs.ChangeState")
testJobA = Job()
testJobA.addFile(inputFileA)
testJobA.create(jobGroupA)
testJobA["state"] = "cleanout"
testJobA["oldstate"] = "new"
testJobA["couch_record"] = "somejive"
testJobA["retry_count"] = 0
testJobA["outcome"] = "success"
testJobA.save()
testJobB = Job()
testJobB.addFile(inputFileB)
testJobB.create(jobGroupA)
testJobB["state"] = "cleanout"
testJobB["oldstate"] = "new"
testJobB["couch_record"] = "somejive"
testJobB["retry_count"] = 0
testJobB["outcome"] = "success"
testJobB.save()
testJobC = Job()
testJobC.addFile(inputFileA)
testJobC.create(jobGroupB)
testJobC["state"] = "cleanout"
testJobC["oldstate"] = "new"
testJobC["couch_record"] = "somejive"
testJobC["retry_count"] = 0
testJobC["outcome"] = "success"
testJobC.save()
testJobD = Job()
testJobD.addFile(inputFileA)
testJobD.create(jobGroupB)
testJobD["state"] = "cleanout"
testJobD["oldstate"] = "new"
testJobD["couch_record"] = "somejive"
testJobD["retry_count"] = 0
testJobD["outcome"] = "failure"
testJobD.save()
testJobE = Job()
testJobE.addFile(inputFileB)
testJobE.create(jobGroupB)
testJobE["state"] = "cleanout"
testJobE["oldstate"] = "new"
testJobE["couch_record"] = "somejive"
testJobE["retry_count"] = 0
testJobE["outcome"] = "success"
testJobE.save()
testJobF = Job()
testJobF.addFile(inputFileB)
testJobF.create(jobGroupB)
testJobF["state"] = "cleanout"
testJobF["oldstate"] = "new"
testJobF["couch_record"] = "somejive"
testJobF["retry_count"] = 0
testJobF["outcome"] = "failure"
testJobF.save()
changeStateDAO.execute(
[testJobA, testJobB, testJobC, testJobD, testJobE, testJobF])
fileA = File(lfn="fileA",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileA.addRun(Run(1, *[45]))
fileA.create()
fileA.addParent(inputFileA["lfn"])
fileB = File(lfn="fileB",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileB.addRun(Run(1, *[45]))
fileB.create()
fileB.addParent(inputFileB["lfn"])
jobGroupA.output.addFile(fileA)
jobGroupA.output.addFile(fileB)
jobGroupA.output.commit()
mergeFilesetA.addFile(fileA)
mergeFilesetA.addFile(fileB)
mergeFilesetA.commit()
fileC = File(lfn="fileC",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileC.addRun(Run(1, *[45]))
fileC.create()
fileC.addParent(inputFileA["lfn"])
fileD = File(lfn="fileD",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileD.addRun(Run(1, *[45]))
fileD.create()
fileD.addParent(inputFileB["lfn"])
jobGroupB.output.addFile(fileC)
jobGroupB.output.addFile(fileD)
mergeFilesetB.addFile(fileC)
mergeFilesetB.addFile(fileD)
mergeFilesetB.commit()
splitter = SplitterFactory()
jobFactory = splitter(package="WMCore.WMBS",
subscription=mergeSubscriptionB)
result = jobFactory(min_merge_size=1,
max_merge_size=20000,
max_merge_events=7169)
assert len(result) == 0, \
"Error: No merge jobs should have been created."
fileE = File(lfn="fileE",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileE.addRun(Run(1, *[45]))
fileE.create()
fileE.addParent(inputFileA["lfn"])
fileF = File(lfn="fileF",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileF.addRun(Run(1, *[45]))
fileF.create()
fileF.addParent(inputFileB["lfn"])
jobGroupB.output.addFile(fileE)
jobGroupB.output.addFile(fileF)
mergeFilesetB.addFile(fileE)
mergeFilesetB.addFile(fileF)
mergeFilesetB.commit()
testJobD["outcome"] = "success"
testJobD.save()
testJobF["outcome"] = "success"
testJobF.save()
changeStateDAO.execute([testJobD, testJobF])
result = jobFactory(min_merge_size=1,
max_merge_size=20000,
max_merge_events=7169)
assert len(result) == 1, \
"Error: One merge job should have been created: %s" % len(result)
return
def testGetOutputMapDAO(self):
"""
_testGetOutputMapDAO_
Verify the proper behavior of the GetOutputMapDAO for a variety of
different processing chains.
"""
recoOutputFileset = Fileset(name="RECO")
recoOutputFileset.create()
mergedRecoOutputFileset = Fileset(name="MergedRECO")
mergedRecoOutputFileset.create()
alcaOutputFileset = Fileset(name="ALCA")
alcaOutputFileset.create()
mergedAlcaOutputFileset = Fileset(name="MergedALCA")
mergedAlcaOutputFileset.create()
dqmOutputFileset = Fileset(name="DQM")
dqmOutputFileset.create()
mergedDqmOutputFileset = Fileset(name="MergedDQM")
mergedDqmOutputFileset.create()
cleanupFileset = Fileset(name="Cleanup")
cleanupFileset.create()
testWorkflow = Workflow(spec="wf001.xml",
owner="Steve",
name="TestWF",
task="None")
testWorkflow.create()
testWorkflow.addOutput("output", recoOutputFileset,
mergedRecoOutputFileset)
testWorkflow.addOutput("ALCARECOStreamCombined", alcaOutputFileset,
mergedAlcaOutputFileset)
testWorkflow.addOutput("DQM", dqmOutputFileset, mergedDqmOutputFileset)
testWorkflow.addOutput("output", cleanupFileset)
testWorkflow.addOutput("ALCARECOStreamCombined", cleanupFileset)
testWorkflow.addOutput("DQM", cleanupFileset)
testRecoMergeWorkflow = Workflow(spec="wf002.xml",
owner="Steve",
name="TestRecoMergeWF",
task="None")
testRecoMergeWorkflow.create()
testRecoMergeWorkflow.addOutput("anything", mergedRecoOutputFileset,
mergedRecoOutputFileset)
testRecoProcWorkflow = Workflow(spec="wf004.xml",
owner="Steve",
name="TestRecoProcWF",
task="None")
testRecoProcWorkflow.create()
testAlcaChildWorkflow = Workflow(spec="wf003.xml",
owner="Steve",
name="TestAlcaChildWF",
task="None")
testAlcaChildWorkflow.create()
inputFile = File(lfn="/path/to/some/lfn",
size=600000,
events=60000,
locations="cmssrm.fnal.gov")
inputFile.create()
testFileset = Fileset(name="TestFileset")
testFileset.create()
testFileset.addFile(inputFile)
testFileset.commit()
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Processing")
testMergeRecoSubscription = Subscription(
fileset=recoOutputFileset,
workflow=testRecoMergeWorkflow,
split_algo="WMBSMergeBySize",
type="Merge")
testProcRecoSubscription = Subscription(fileset=recoOutputFileset,
workflow=testRecoProcWorkflow,
split_algo="FileBased",
type="Processing")
testChildAlcaSubscription = Subscription(
fileset=alcaOutputFileset,
workflow=testAlcaChildWorkflow,
split_algo="FileBased",
type="Processing")
testSubscription.create()
testMergeRecoSubscription.create()
testProcRecoSubscription.create()
testChildAlcaSubscription.create()
testSubscription.acquireFiles()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testJob = Job(name="SplitJobA", files=[inputFile])
testJob.create(group=testJobGroup)
testJob["state"] = "complete"
testJob.save()
outputMapAction = self.daoFactory(classname="Jobs.GetOutputMap")
outputMap = outputMapAction.execute(jobID=testJob["id"])
assert len(outputMap.keys()) == 3, \
"Error: Wrong number of outputs for primary workflow."
goldenMap = {
"output": (recoOutputFileset.id, mergedRecoOutputFileset.id),
"ALCARECOStreamCombined":
(alcaOutputFileset.id, mergedAlcaOutputFileset.id),
"DQM": (dqmOutputFileset.id, mergedDqmOutputFileset.id)
}
for outputID in outputMap.keys():
for outputFilesets in outputMap[outputID]:
if outputFilesets["merged_output_fileset"] == None:
self.assertEqual(outputFilesets["output_fileset"],
cleanupFileset.id,
"Error: Cleanup fileset is wrong.")
continue
self.assertTrue(outputID in goldenMap.keys(),
"Error: Output identifier is missing.")
self.assertEqual(outputFilesets["output_fileset"],
goldenMap[outputID][0],
"Error: Output fileset is wrong.")
self.assertEqual(outputFilesets["merged_output_fileset"],
goldenMap[outputID][1],
"Error: Merged output fileset is wrong.")
del goldenMap[outputID]
self.assertEqual(len(goldenMap.keys()), 0,
"Error: Missing output maps.")
return
def setUp(self):
"""
_setUp_
"""
self.testInit = TestInit(__file__)
self.testInit.setLogging()
self.testInit.setDatabaseConnection()
self.testInit.setSchema(customModules = ["T0.WMBS"])
self.splitterFactory = SplitterFactory(package = "T0.JobSplitting")
myThread = threading.currentThread()
daoFactory = DAOFactory(package = "T0.WMBS",
logger = logging,
dbinterface = myThread.dbi)
wmbsDaoFactory = DAOFactory(package = "WMCore.WMBS",
logger = logging,
dbinterface = myThread.dbi)
myThread.dbi.processData("""INSERT INTO wmbs_location
(id, site_name, state)
VALUES (1, 'SomeSite', 1)
""", transaction = False)
myThread.dbi.processData("""INSERT INTO wmbs_location_pnn
(location, pnn)
VALUES (1, 'SomePNN')
""", transaction = False)
insertRunDAO = daoFactory(classname = "RunConfig.InsertRun")
insertRunDAO.execute(binds = { 'RUN' : 1,
'HLTKEY' : "someHLTKey" },
transaction = False)
insertLumiDAO = daoFactory(classname = "RunConfig.InsertLumiSection")
insertLumiDAO.execute(binds = { 'RUN' : 1,
'LUMI' : 1 },
transaction = False)
insertStreamDAO = daoFactory(classname = "RunConfig.InsertStream")
insertStreamDAO.execute(binds = { 'STREAM' : "Express" },
transaction = False)
insertStreamFilesetDAO = daoFactory(classname = "RunConfig.InsertStreamFileset")
insertStreamFilesetDAO.execute(1, "Express", "TestFileset1")
insertStreamerDAO = daoFactory(classname = "RunConfig.InsertStreamer")
insertStreamerDAO.execute(binds = { 'RUN' : 1,
'LUMI' : 1,
'STREAM' : "Express",
'TIME' : int(time.time()),
'LFN' : "/streamer",
'FILESIZE' : 0,
'EVENTS' : 0 },
transaction = False)
insertPromptCalibrationDAO = daoFactory(classname = "RunConfig.InsertPromptCalibration")
insertPromptCalibrationDAO.execute( { 'RUN' : 1,
'STREAM' : "Express" },
transaction = False)
self.fileset1 = Fileset(name = "TestFileset1")
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 = "Condition",
type = "Condition")
self.subscription1.create()
# set parentage chain and sqlite fileset
alcaRecoFile = File("/alcareco", size = 0, events = 0)
alcaRecoFile.addRun(Run(1, *[1]))
alcaRecoFile.setLocation("SomePNN", immediateSave = False)
alcaRecoFile.create()
alcaPromptFile = File("/alcaprompt", size = 0, events = 0)
alcaPromptFile.addRun(Run(1, *[1]))
alcaPromptFile.setLocation("SomePNN", immediateSave = False)
alcaPromptFile.create()
sqliteFile = File("/sqlite", size = 0, events = 0)
sqliteFile.create()
self.fileset1.addFile(sqliteFile)
self.fileset1.commit()
results = myThread.dbi.processData("""SELECT lfn FROM wmbs_file_details
""",
transaction = False)[0].fetchall()
setParentageDAO = wmbsDaoFactory(classname = "Files.SetParentage")
setParentageDAO.execute(binds = [ { 'parent' : "/streamer",
'child' : "/alcareco" },
{ 'parent' : "/alcareco",
'child' : "/alcaprompt" },
{ 'parent' : "/alcaprompt",
'child' : "/sqlite" } ],
transaction = False)
# default split parameters
self.splitArgs = {}
self.splitArgs['runNumber'] = 1
self.splitArgs['streamName'] = "Express"
return
def test_AutoIncrementCheck(self):
"""
_AutoIncrementCheck_
Test and see whether we can find and set the auto_increment values
"""
myThread = threading.currentThread()
if not myThread.dialect.lower() == 'mysql':
return
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()
incrementDAO = self.daoFactory(classname="Jobs.AutoIncrementCheck")
incrementDAO.execute()
testJob = Job()
testJob.create(group=testJobGroup)
self.assertEqual(testJob.exists(), 1)
incrementDAO.execute()
testJob = Job()
testJob.create(group=testJobGroup)
self.assertEqual(testJob.exists(), 2)
incrementDAO.execute(input=10)
testJob = Job()
testJob.create(group=testJobGroup)
self.assertEqual(testJob.exists(), 11)
incrementDAO.execute(input=5)
testJob = Job()
testJob.create(group=testJobGroup)
self.assertEqual(testJob.exists(), 12)
return
def testFailJobInput(self):
"""
_testFailJobInput_
Test the Jobs.FailInput DAO and verify that it doesn't affect other
jobs/subscriptions that run over the same files.
"""
testWorkflow = Workflow(spec="spec.xml",
owner="Steve",
name="wf001",
task="Test")
bogusWorkflow = Workflow(spec="spec1.xml",
owner="Steve",
name="wf002",
task="Test")
testWorkflow.create()
bogusWorkflow.create()
testFileset = Fileset(name="TestFileset")
bogusFileset = Fileset(name="BogusFileset")
testFileset.create()
bogusFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow)
bogusSubscription = Subscription(fileset=bogusFileset,
workflow=bogusWorkflow)
testSubscription.create()
bogusSubscription.create()
testFileA = File(lfn=makeUUID(), locations="setest.site.ch")
testFileB = File(lfn=makeUUID(), locations="setest.site.ch")
testFileC = File(lfn=makeUUID(), locations="setest.site.ch")
testFileA.create()
testFileB.create()
testFileC.create()
testFileset.addFile([testFileA, testFileB, testFileC])
bogusFileset.addFile([testFileA, testFileB, testFileC])
testFileset.commit()
bogusFileset.commit()
testSubscription.completeFiles([testFileA, testFileB, testFileC])
bogusSubscription.acquireFiles([testFileA, testFileB, testFileC])
testJobGroup = JobGroup(subscription=testSubscription)
bogusJobGroup = JobGroup(subscription=bogusSubscription)
testJobGroup.create()
bogusJobGroup.create()
testJobA = Job(name="TestJobA",
files=[testFileA, testFileB, testFileC])
testJobB = Job(name="TestJobB",
files=[testFileA, testFileB, testFileC])
bogusJob = Job(name="BogusJob",
files=[testFileA, testFileB, testFileC])
testJobA.create(group=testJobGroup)
testJobB.create(group=testJobGroup)
bogusJob.create(group=bogusJobGroup)
testJobA.failInputFiles()
testJobB.failInputFiles()
self.assertEqual(len(testSubscription.filesOfStatus("Available")), 0)
self.assertEqual(len(testSubscription.filesOfStatus("Acquired")), 0)
self.assertEqual(len(testSubscription.filesOfStatus("Failed")), 3)
self.assertEqual(len(testSubscription.filesOfStatus("Completed")), 0)
changeStateAction = self.daoFactory(classname="Jobs.ChangeState")
testJobB["state"] = "cleanout"
changeStateAction.execute([testJobB])
# Try again
testJobA.failInputFiles()
# Should now be failed
self.assertEqual(len(testSubscription.filesOfStatus("Available")), 0)
self.assertEqual(len(testSubscription.filesOfStatus("Acquired")), 0)
self.assertEqual(len(testSubscription.filesOfStatus("Failed")), 3)
self.assertEqual(len(testSubscription.filesOfStatus("Completed")), 0)
# bogus should be unchanged
self.assertEqual(len(bogusSubscription.filesOfStatus("Available")), 0)
self.assertEqual(len(bogusSubscription.filesOfStatus("Acquired")), 3)
self.assertEqual(len(bogusSubscription.filesOfStatus("Failed")), 0)
self.assertEqual(len(bogusSubscription.filesOfStatus("Completed")), 0)
return
def algorithm(self, *args, **kwargs):
"""
_algorithm_
An event base splitting algorithm. All available files are split into a
set number of events per job.
"""
eventsPerJob = int(kwargs.get("events_per_job", 100))
eventsPerLumi = int(kwargs.get("events_per_lumi", eventsPerJob))
getParents = kwargs.get("include_parents", False)
lheInput = kwargs.get("lheInputFiles", False)
collectionName = kwargs.get('collectionName', None)
timePerEvent, sizePerEvent, memoryRequirement = \
self.getPerformanceParameters(kwargs.get('performance', {}))
acdcFileList = []
# If we have runLumi info, we need to load it from couch
if collectionName:
try:
from WMCore.ACDC.DataCollectionService import DataCollectionService
couchURL = kwargs.get('couchURL')
couchDB = kwargs.get('couchDB')
filesetName = kwargs.get('filesetName')
collectionName = kwargs.get('collectionName')
owner = kwargs.get('owner')
group = kwargs.get('group')
logging.info('Creating jobs for ACDC fileset %s' % filesetName)
dcs = DataCollectionService(couchURL, couchDB)
acdcFileList = dcs.getProductionACDCInfo(
collectionName, filesetName, owner, group)
except Exception as ex:
msg = "Exception while trying to load goodRunList\n"
msg += "Refusing to create any jobs.\n"
msg += str(ex)
msg += str(traceback.format_exc())
logging.error(msg)
return
totalJobs = 0
locationDict = self.sortByLocation()
for location in locationDict:
self.newGroup()
fileList = locationDict[location]
getRunLumiInformation = False
for f in fileList:
if f['lfn'].startswith("MCFakeFile"):
#We have one MCFakeFile, then it needs run information
getRunLumiInformation = True
break
if getRunLumiInformation:
if self.package == 'WMCore.WMBS':
loadRunLumi = self.daoFactory(
classname="Files.GetBulkRunLumi")
fileLumis = loadRunLumi.execute(files=fileList)
for f in fileList:
lumiDict = fileLumis.get(f['id'], {})
for run in lumiDict.keys():
f.addRun(run=Run(run, *lumiDict[run]))
for f in fileList:
currentEvent = f['first_event']
eventsInFile = f['events']
runs = list(f['runs'])
#We got the runs, clean the file.
f['runs'] = set()
if getParents:
parentLFNs = self.findParent(lfn=f['lfn'])
for lfn in parentLFNs:
parent = File(lfn=lfn)
f['parents'].add(parent)
if acdcFileList:
if f['lfn'] in [x['lfn'] for x in acdcFileList]:
totalJobs = self.createACDCJobs(
f, acdcFileList, timePerEvent, sizePerEvent,
memoryRequirement, lheInput, eventsPerJob,
eventsPerLumi, totalJobs)
continue
elif not f['lfn'].startswith("MCFakeFile"):
# Very very uncommon, but it has real input dataset
if eventsInFile >= eventsPerJob:
while currentEvent < eventsInFile:
self.newJob(name=self.getJobName(length=totalJobs))
self.currentJob.addFile(f)
if eventsPerJob + currentEvent < eventsInFile:
jobTime = eventsPerJob * timePerEvent
diskRequired = eventsPerJob * sizePerEvent
self.currentJob["mask"].setMaxAndSkipEvents(
eventsPerJob, currentEvent)
else:
jobTime = (eventsInFile -
currentEvent) * timePerEvent
diskRequired = (eventsInFile -
currentEvent) * sizePerEvent
self.currentJob["mask"].setMaxAndSkipEvents(
None, currentEvent)
self.currentJob.addResourceEstimates(
jobTime=jobTime,
memory=memoryRequirement,
disk=diskRequired)
logging.debug(
"Job created for real input with %s" %
self.currentJob)
currentEvent += eventsPerJob
totalJobs += 1
else:
self.newJob(name=self.getJobName(length=totalJobs))
self.currentJob.addFile(f)
jobTime = eventsInFile * timePerEvent
diskRequired = eventsInFile * sizePerEvent
self.currentJob.addResourceEstimates(
jobTime=jobTime,
memory=memoryRequirement,
disk=diskRequired)
logging.debug(
"Last job created for real input with %s" %
self.currentJob)
totalJobs += 1
else:
#This assumes there's only one run which is the case for MC
lumis = runs[0].lumis
(firstLumi, lastLumi) = (min(lumis), max(lumis))
currentLumi = firstLumi
totalEvents = 0
if eventsInFile >= eventsPerJob:
while totalEvents < eventsInFile:
self.newJob(name=self.getJobName(length=totalJobs))
self.currentJob.addFile(f)
self.currentJob.addBaggageParameter(
"lheInputFiles", lheInput)
lumisPerJob = int(
ceil(float(eventsPerJob) / eventsPerLumi))
#Limit the number of events to a unsigned 32bit int
eventsRemaining = eventsInFile - totalEvents
if (currentEvent + eventsPerJob -
1) > (2**32 - 1) and (currentEvent +
eventsRemaining -
1) > (2**32 - 1):
currentEvent = 1
if eventsRemaining > eventsPerJob:
self.currentJob["mask"].setMaxAndSkipEvents(
eventsPerJob, currentEvent)
self.currentJob["mask"].setMaxAndSkipLumis(
lumisPerJob, currentLumi)
jobTime = eventsPerJob * timePerEvent
diskRequired = eventsPerJob * sizePerEvent
else:
jobTime = eventsRemaining * timePerEvent
diskRequired = eventsRemaining * sizePerEvent
lumisPerJob = int(
ceil(
float(eventsRemaining) /
eventsPerLumi))
self.currentJob["mask"].setMaxAndSkipEvents(
eventsRemaining, currentEvent)
self.currentJob["mask"].setMaxAndSkipLumis(
lumisPerJob, currentLumi)
currentLumi += lumisPerJob
currentEvent += eventsPerJob
totalEvents += eventsPerJob
totalJobs += 1
self.currentJob.addResourceEstimates(
jobTime=jobTime,
memory=memoryRequirement,
disk=diskRequired)
else:
self.newJob(name=self.getJobName(length=totalJobs))
self.currentJob.addFile(f)
#For MC we use firstEvent instead of skipEvents so set it to 1
#We must check for events going over 2**32 - 1 here too
if (eventsInFile + currentEvent - 1) > (2**32 - 1):
currentEvent = 1
self.currentJob["mask"].setMaxAndSkipEvents(
eventsInFile, currentEvent)
self.currentJob["mask"].setMaxAndSkipLumis(
lastLumi - currentLumi + 1, currentLumi)
jobTime = eventsInFile * timePerEvent
diskRequired = eventsInFile * sizePerEvent
self.currentJob.addResourceEstimates(
jobTime=jobTime,
memory=memoryRequirement,
disk=diskRequired)
totalJobs += 1
def stuffWMBS(self, workflowURL, name):
"""
_stuffWMBS_
Insert some dummy jobs, jobgroups, filesets, files and subscriptions
into WMBS to test job creation. Three completed job groups each
containing several files are injected. Another incomplete job group is
also injected. Also files are added to the "Mergeable" subscription as
well as to the output fileset for their jobgroups.
"""
locationAction = self.daoFactory(classname = "Locations.New")
locationAction.execute(siteName = "s1", seName = "somese.cern.ch")
changeStateDAO = self.daoFactory(classname = "Jobs.ChangeState")
mergeFileset = Fileset(name = "mergeFileset")
mergeFileset.create()
bogusFileset = Fileset(name = "bogusFileset")
bogusFileset.create()
mergeWorkflow = Workflow(spec = workflowURL, owner = "mnorman",
name = name, task="/TestWorkload/ReReco")
mergeWorkflow.create()
mergeSubscription = Subscription(fileset = mergeFileset,
workflow = mergeWorkflow,
split_algo = "ParentlessMergeBySize")
mergeSubscription.create()
bogusSubscription = Subscription(fileset = bogusFileset,
workflow = mergeWorkflow,
split_algo = "ParentlessMergeBySize")
file1 = File(lfn = "file1", size = 1024, events = 1024, first_event = 0,
locations = set(["somese.cern.ch"]))
file1.addRun(Run(1, *[45]))
file1.create()
file2 = File(lfn = "file2", size = 1024, events = 1024,
first_event = 1024, locations = set(["somese.cern.ch"]))
file2.addRun(Run(1, *[45]))
file2.create()
file3 = File(lfn = "file3", size = 1024, events = 1024,
first_event = 2048, locations = set(["somese.cern.ch"]))
file3.addRun(Run(1, *[45]))
file3.create()
file4 = File(lfn = "file4", size = 1024, events = 1024,
first_event = 3072, locations = set(["somese.cern.ch"]))
file4.addRun(Run(1, *[45]))
file4.create()
fileA = File(lfn = "fileA", size = 1024, events = 1024,
first_event = 0, locations = set(["somese.cern.ch"]))
fileA.addRun(Run(1, *[46]))
fileA.create()
fileB = File(lfn = "fileB", size = 1024, events = 1024,
first_event = 1024, locations = set(["somese.cern.ch"]))
fileB.addRun(Run(1, *[46]))
fileB.create()
fileC = File(lfn = "fileC", size = 1024, events = 1024,
first_event = 2048, locations = set(["somese.cern.ch"]))
fileC.addRun(Run(1, *[46]))
fileC.create()
fileI = File(lfn = "fileI", size = 1024, events = 1024,
first_event = 0, locations = set(["somese.cern.ch"]))
fileI.addRun(Run(2, *[46]))
fileI.create()
fileII = File(lfn = "fileII", size = 1024, events = 1024,
first_event = 1024, locations = set(["somese.cern.ch"]))
fileII.addRun(Run(2, *[46]))
fileII.create()
fileIII = File(lfn = "fileIII", size = 1024, events = 102400,
first_event = 2048, locations = set(["somese.cern.ch"]))
fileIII.addRun(Run(2, *[46]))
fileIII.create()
fileIV = File(lfn = "fileIV", size = 102400, events = 1024,
first_event = 3072, locations = set(["somese.cern.ch"]))
fileIV.addRun(Run(2, *[46]))
fileIV.create()
for file in [file1, file2, file3, file4, fileA, fileB, fileC, fileI,
fileII, fileIII, fileIV]:
mergeFileset.addFile(file)
bogusFileset.addFile(file)
mergeFileset.commit()
bogusFileset.commit()
return
def __call__(self, filesets):
"""
The algorithm itself
"""
# Update run list
self.getNewRuns()
# Do per fileset work, abandon fileset processing on exception
for fileset in filesets:
ds = fileset.name
try:
# Do per run work
watchCompleteFiles = []
for watch in self.watchedRuns:
# Ensure watcher has dataset listed
watch.addDatasetOfInterest(ds)
# Query DBS to find all blocks for this run / dataset
(files, blocks, fileInfoMap) = \
self.dbsHelper.getFileInfo(watch.run, ds)
# Now determine all required parent blocks
parentBlocks = set()
if fileset.requireParents:
parentDs = self.dbsHelper.getParentDataset(ds)
parentBlocks = self.dbsHelper.getBlockInfo(
watch.run, parentDs)
# Final list of all required blocks
allBlocks = blocks[:]
allBlocks.update(parentBlocks)
# Find all sites where all blocks are complete
sites = self.phedexHelper.getCompleteSites(blocks)
# Get sites with newly completed transfers
newSites = watch.getNewSites(ds, sites)
if len(newSites) > 0:
# Add the files for these blocks to the fileset
for file in fileInfoMap:
fi = fileInfoMap[file]
# First add parent file
if fileset.requireParents:
parentFile = File(lfn=fi["file.parent"])
parentFile.save()
parentFile.setLocation(newSites)
# Add actual file
fileToAdd = File(lfn=file,
size=fi["file.size"],
events=fi["file.events"],
run=watch.run,
umi=fi["file.lumi"])
if not fileToAdd.exists(
) and fileset.requireParents:
fileToAdd.addParent(fi["file.parent"])
# Add new locations but don't persist immediately
fileToAdd.setLocations(newSites,
immediateSave=False)
# Add the file to the new file list
fileset.addFile(fileToAdd)
# Add the site info to the watcher list
watchCompleteFiles.append([watch, ds, newSites])
# Commit the fileset
fileset.commit()
# Add the watched runs
for a in watchCompleteFiles:
a[0].addCompletedNodes(a[1], a[2])
except:
# Reset the watch list so we re-evaluate next call
watchCompleteFiles = []
# Purge old runs
self.purgeWatchedRuns()
def verifyFileMetaData(self, jobID, fwkJobReportFiles):
"""
_verifyFileMetaData_
Verify that all the files that were output by a job made it into WMBS
correctly. Compare the contents of WMBS to the files in the frameworks
job report.
Note that fwkJobReportFiles is a list of DataStructs File objects.
"""
testJob = Job(id=jobID)
testJob.loadData()
inputLFNs = []
for inputFile in testJob["input_files"]:
inputLFNs.append(inputFile["lfn"])
for fwkJobReportFile in fwkJobReportFiles:
outputFile = File(lfn=fwkJobReportFile["lfn"])
outputFile.loadData(parentage=1)
assert outputFile["events"] == int(fwkJobReportFile["events"]), \
"Error: Output file has wrong events: %s, %s" % \
(outputFile["events"], fwkJobReportFile["events"])
assert outputFile["size"] == int(fwkJobReportFile["size"]), \
"Error: Output file has wrong size: %s, %s" % \
(outputFile["size"], fwkJobReportFile["size"])
for ckType in fwkJobReportFile["checksums"].keys():
assert ckType in outputFile["checksums"].keys(), \
"Error: Output file is missing checksums: %s" % ckType
assert outputFile["checksums"][ckType] == fwkJobReportFile["checksums"][ckType], \
"Error: Checksums don't match."
assert len(fwkJobReportFile["checksums"].keys()) == \
len(outputFile["checksums"].keys()), \
"Error: Wrong number of checksums."
jobType = self.getJobTypeAction.execute(jobID=jobID)
if jobType == "Merge":
assert str(outputFile["merged"]) == "True", \
"Error: Merge jobs should output merged files."
else:
assert outputFile["merged"] == fwkJobReportFile["merged"], \
"Error: Output file merged output is wrong: %s, %s" % \
(outputFile["merged"], fwkJobReportFile["merged"])
assert len(outputFile["locations"]) == 1, \
"Error: outputfile should have one location: %s" % outputFile["locations"]
assert list(outputFile["locations"])[0] == list(fwkJobReportFile["locations"])[0], \
"Error: wrong location for file."
assert len(outputFile["parents"]) == len(inputLFNs), \
"Error: Output file has wrong number of parents."
for outputParent in outputFile["parents"]:
assert outputParent["lfn"] in inputLFNs, \
"Error: Unknown parent file: %s" % outputParent["lfn"]
fwjrRuns = {}
for run in fwkJobReportFile["runs"]:
fwjrRuns[run.run] = run.lumis
for run in outputFile["runs"]:
assert run.run in fwjrRuns, \
"Error: Extra run in output: %s" % run.run
for lumi in run:
assert lumi in fwjrRuns[run.run], \
"Error: Extra lumi: %s" % lumi
fwjrRuns[run.run].remove(lumi)
if len(fwjrRuns[run.run]) == 0:
del fwjrRuns[run.run]
assert len(fwjrRuns.keys()) == 0, \
"Error: Missing runs, lumis: %s" % fwjrRuns
testJobGroup = JobGroup(id=testJob["jobgroup"])
testJobGroup.loadData()
jobGroupFileset = testJobGroup.output
jobGroupFileset.loadData()
assert outputFile["id"] in jobGroupFileset.getFiles(type = "id"), \
"Error: output file not in jobgroup fileset."
if testJob["mask"]["FirstEvent"] == None:
assert outputFile["first_event"] == 0, \
"Error: first event not set correctly: 0, %s" % \
outputFile["first_event"]
else:
assert testJob["mask"]["FirstEvent"] == outputFile["first_event"], \
"Error: last event not set correctly: %s, %s" % \
(testJob["mask"]["FirstEvent"], outputFile["first_event"])
return
def testLumiMask(self):
"""
_testLumiMask_
Test that we can use a lumi-mask to filter good runs/lumis.
"""
splitter = SplitterFactory()
# Create 3 files with 100 events per lumi:
# - file1 with 1 run of 8 lumis
# - file2 with 2 runs of 2 lumis each
# - file3 with 1 run of 5 lumis
fileA = File(lfn="/this/is/file1", size=1000, events=800)
fileB = File(lfn="/this/is/file2", size=1000, events=400)
fileC = File(lfn="/this/is/file3", size=1000, events=500)
lumiListA = []
for lumi in range(8):
lumiListA.append(10 + lumi)
fileA.addRun(Run(1, *lumiListA))
fileA.setLocation("T1_US_FNAL_Disk")
lumiListB1 = []
lumiListB2 = []
for lumi in range(2):
lumiListB1.append(20 + lumi)
lumiListB2.append(30 + lumi)
fileB.addRun(Run(2, *lumiListB1))
fileB.addRun(Run(3, *lumiListB2))
fileB.setLocation("T1_US_FNAL_Disk")
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("T1_US_FNAL_Disk")
testFileset = Fileset(name='Fileset')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# Use a lumi-mask = {1: [[10,14]], 2: [[20,21]], 4: [[40,41]]}
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=850,
runs=['1', '2', '4'],
lumis=['10,14', '20,21', '40,41'],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1,
"There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 2, "Two jobs must be in the jobgroup")
processedLumis = LumiList()
for job in jobs:
processedLumis += LumiList(
compactList=job['mask'].getRunAndLumis())
correctLumis = LumiList(compactList={
1: [[10, 14]],
2: [[20, 21]],
4: [[40, 41]]
})
self.assertEqual(processedLumis.getCMSSWString(),
correctLumis.getCMSSWString())
def testLumiCorrections(self):
"""
_testLumiCorrections_
Test the splitting algorithm can handle lumis which cross multiple files.
No need for applyLumiCorrection=True
"""
splitter = SplitterFactory()
testSubscription = self.createSubscription(nFiles=2,
lumisPerFile=2,
twoSites=False,
nEventsPerFile=150)
files = testSubscription.getFileset().getFiles()
self.assertEqual(len(files), 2)
# Two files with 2 lumis each: file0 has run0 and lumis 0,1 - file1 has run1 and lumis 2,3 - each 150 events
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
jobGroups = jobFactory(events_per_job=50,
halt_job_on_file_boundaries=False,
splitOnRun=False,
performance=self.performanceParams)
# The splitting algorithm will assume 75 events per lumi so we will have one job per lumi
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 4)
# Recreate the same subscription as before, but with duplicated lumis
testFileset = Fileset(name='FilesetA')
fileA = File(lfn="/this/is/file1", size=1000, events=150)
lumiListA = [0, 1, 42]
fileA.addRun(Run(0, *lumiListA))
fileA.setLocation("T1_US_FNAL_Disk")
testFileset.addFile(fileA)
fileB = File(lfn="/this/is/file2", size=1000, events=150)
lumiListB = [2, 3, 42]
fileB.addRun(Run(0, *lumiListB))
fileB.setLocation("T1_US_FNAL_Disk")
testFileset.addFile(fileB)
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
jobGroups = jobFactory(events_per_job=50,
halt_job_on_file_boundaries=True,
performance=self.performanceParams)
# Now we will have: file0: Run0 and lumis [0, 1, 42] file1: Run0 and lumis [2, 3, 42]
# With 50 events per lumi, one job per lumi, one will have two files on lumi 42
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 5)
n1files = 0
n2files = 0
lumi1files = []
lumi2files = []
for job in jobs:
runLumis = job['mask'].getRunAndLumis()
lumis = runLumis[0]
self.assertEqual(len(runLumis), 1)
self.assertEqual(len(lumis), 1)
self.assertEqual(lumis[0][0],
lumis[0][1]) # Make sure only one lumi per job
if len(job['input_files']) == 1:
n1files += 1
lumi1files.append(lumis[0][0])
elif len(job['input_files']) == 2:
n2files += 1
lumi2files.append(lumis[0][0])
else:
self.fail("At least one job has nFiles =! 1, 2")
self.assertEqual(n1files, 4)
self.assertEqual(n2files, 1)
self.assertItemsEqual(lumi1files, [0, 1, 2, 3])
self.assertItemsEqual(lumi2files, [42])
def testNoFileSplitNoHardLimit(self):
"""
_testNoFileSplitNoHardLimit_
In this case we don't split on file boundaries, check different combination of files
make sure we make the most of the splitting, e.g. include many zero event files in
a single job.
"""
splitter = SplitterFactory()
# Create 100 files with 7 lumi per file and 0 events per lumi on average.
testSubscription = self.createSubscription(nFiles=100,
lumisPerFile=7,
twoSites=False,
nEventsPerFile=0)
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# First test, the optimal settings are 360 events per job. As we have files with 0 events per lumi, this will
# configure the splitting to a single job containing all files
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=360,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 0)
# since there are not enough events to be splitted, let's close the fileset and get it going
fileset = testSubscription.getFileset()
fileset.markOpen(False)
# One job in one job group with 100 files
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=360,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 1)
self.assertEqual(len(jobs[0]['input_files']), 100)
# Create 7 files, each one with different lumi/event distributions
testFileset = Fileset(name="FilesetA")
testFileA = self.createFile("/this/is/file1", 250, 0, 5,
"T1_US_FNAL_Disk")
testFileB = self.createFile("/this/is/file2", 600, 1, 1,
"T1_US_FNAL_Disk")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2,
"T1_US_FNAL_Disk")
testFileD = self.createFile("/this/is/file4", 100, 3, 1,
"T1_US_FNAL_Disk")
testFileE = self.createFile("/this/is/file5", 30, 4, 1,
"T1_US_FNAL_Disk")
testFileF = self.createFile("/this/is/file6", 10, 5, 1,
"T1_US_FNAL_Disk")
testFileG = self.createFile("/this/is/file7", 153, 6, 3,
"T1_US_FNAL_Disk")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# Split the work targeting 150 events per job
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=150,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 7)
# Test interactions of this algorithm with splitOnRun = True
# Make 2 files, one with 3 runs and a second one with the last run of the first
fileA = File(lfn="/this/is/file10", size=1000, events=2400)
lumiListA = []
lumiListB = []
lumiListC = []
for lumi in range(8):
lumiListA.append(1 + lumi)
lumiListB.append(1 + lumi)
lumiListC.append(1 + lumi)
fileA.addRun(Run(1, *lumiListA))
fileA.addRun(Run(2, *lumiListA))
fileA.addRun(Run(3, *lumiListA))
fileA.setLocation("T2_CH_CERN")
fileB = self.createFile('/this/is/file11', 200, 3, 5, "T2_CH_CERN")
testFileset = Fileset(name='FilesetB')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# The settings for this splitting are 700 events per job
jobGroups = jobFactory(splitOnRun=True,
halt_job_on_file_boundaries=False,
events_per_job=700,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 7)
# Make sure each job has one run
for job in jobs:
self.assertEqual(len(job['mask'].getRunAndLumis()), 1)
def _createThisSubscription(self, initialCounter=1):
"""
Private function to create a fileset and subscription with
different fileset and file names
:param initialCounter: just a simple integer to be appended to files
:return: an splitter instance (jobFactory)
"""
splitter = SplitterFactory()
# Create 3 files with 100 events per lumi:
# - file1 with 1 run of 8 lumis
# - file2 with 2 runs of 2 lumis each
# - file3 with 1 run of 5 lumis
testFileset = Fileset(name='Fileset%s' % initialCounter)
fileA = File(lfn="/this/is/file%s" % initialCounter,
size=1000,
events=800)
lumiListA = []
for lumi in range(8):
lumiListA.append(10 + lumi)
fileA.addRun(Run(1, *lumiListA))
fileA.setLocation("T1_US_FNAL_Disk")
initialCounter = int(initialCounter) + 1
fileB = File(lfn="/this/is/file%s" % initialCounter,
size=1000,
events=400)
lumiListB1 = []
lumiListB2 = []
for lumi in range(2):
lumiListB1.append(20 + lumi)
lumiListB2.append(30 + lumi)
fileB.addRun(Run(2, *lumiListB1))
fileB.addRun(Run(3, *lumiListB2))
fileB.setLocation("T1_US_FNAL_Disk")
initialCounter = int(initialCounter) + 1
fileC = File(lfn="/this/is/file%s" % initialCounter,
size=1000,
events=500)
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("T1_US_FNAL_Disk")
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
return jobFactory
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 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 test10(self):
"""
_test10_
Test merging of multiple lumis with holes in the lumi sequence
Hole is due to no streamer files for the lumi
Multi lumi input
"""
mySplitArgs = self.splitArgs.copy()
for lumi in [1, 2, 5]:
for i in range(2):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.addRun(Run(1, *[lumi]))
newFile.setLocation("SomeSE", immediateSave = False)
newFile.create()
self.fileset2.addFile(newFile)
self.fileset2.commit()
jobFactory = self.splitterFactory(package = "WMCore.WMBS",
subscription = self.subscription2)
mySplitArgs['maxInputEvents'] = 500
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), 0,
"ERROR: JobFactory should have returned no JobGroup")
self.insertClosedLumiDAO.execute(binds = { 'RUN' : 1,
'LUMI' : 4,
'STREAM' : "A",
'FILECOUNT' : 1,
'INSERT_TIME' : self.currentTime,
'CLOSE_TIME' : self.currentTime },
transaction = False)
self.feedStreamersDAO.execute(transaction = False)
self.fileset1.loadData()
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
for fileid in self.fileset1.getFiles(type = 'id'):
self.acquireFilesDAO.execute(self.subscription1['id'], fileid,
transaction = False)
jobGroups = jobFactory(**mySplitArgs)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
for fileid in self.fileset1.getFiles(type = 'id'):
self.completeFilesDAO.execute(self.subscription1['id'], fileid,
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")
job = jobGroups[0].jobs[0]
self.assertEqual(len(job.getFiles()), 4,
"ERROR: Job does not process 4 files")
return
def testReportHandling(self):
"""
_testReportHandling_
Verify that we're able to parse a CMSSW report, convert it to a Report()
style report, pickle it and then have the accountant process it.
"""
self.procPath = os.path.join(
WMCore.WMBase.getTestBase(),
"WMCore_t/FwkJobReport_t/CMSSWProcessingReport.xml")
myReport = Report("cmsRun1")
myReport.parse(self.procPath)
# Fake some metadata that should be added by the stageout scripts.
for fileRef in myReport.getAllFileRefsFromStep("cmsRun1"):
fileRef.size = 1024
fileRef.location = "cmssrm.fnal.gov"
fwjrPath = os.path.join(self.tempDir, "ProcReport.pkl")
cmsRunStep = myReport.retrieveStep("cmsRun1")
cmsRunStep.status = 0
myReport.setTaskName('/TestWF/None')
myReport.persist(fwjrPath)
self.setFWJRAction.execute(jobID=self.testJob["id"], fwjrPath=fwjrPath)
pFile = DBSBufferFile(lfn="/path/to/some/lfn",
size=600000,
events=60000)
pFile.setAlgorithm(appName="cmsRun",
appVer="UNKNOWN",
appFam="RECO",
psetHash="GIBBERISH",
configContent="MOREGIBBERISH")
pFile.setDatasetPath("/bogus/dataset/path")
#pFile.addRun(Run(1, *[45]))
pFile.create()
config = self.createConfig(workerThreads=1)
accountant = JobAccountantPoller(config)
accountant.setup()
accountant.algorithm()
self.verifyJobSuccess(self.testJob["id"])
self.verifyFileMetaData(self.testJob["id"],
myReport.getAllFilesFromStep("cmsRun1"))
inputFile = File(
lfn=
"/store/backfill/2/unmerged/WMAgentCommissioining10/MinimumBias/RECO/rereco_GR09_R_34X_V5_All_v1/0000/outputRECORECO.root"
)
inputFile.load()
self.testMergeJob = Job(name="testMergeJob", files=[inputFile])
self.testMergeJob.create(group=self.mergeJobGroup)
self.testMergeJob["state"] = "complete"
self.stateChangeAction.execute(jobs=[self.testMergeJob])
self.mergePath = os.path.join(
WMCore.WMBase.getTestBase(),
"WMCore_t/FwkJobReport_t/CMSSWMergeReport.xml")
myReport = Report("mergeReco")
myReport.parse(self.mergePath)
# Fake some metadata that should be added by the stageout scripts.
for fileRef in myReport.getAllFileRefsFromStep("mergeReco"):
fileRef.size = 1024
fileRef.location = "cmssrm.fnal.gov"
fileRef.dataset = {
"applicationName": "cmsRun",
"applicationVersion": "CMSSW_3_4_2_patch1",
"primaryDataset": "MinimumBias",
"processedDataset": "Rereco-v1",
"dataTier": "RECO"
}
fwjrPath = os.path.join(self.tempDir, "MergeReport.pkl")
myReport.setTaskName('/MergeWF/None')
cmsRunStep = myReport.retrieveStep("mergeReco")
cmsRunStep.status = 0
myReport.persist(fwjrPath)
self.setFWJRAction.execute(jobID=self.testMergeJob["id"],
fwjrPath=fwjrPath)
accountant.algorithm()
self.verifyJobSuccess(self.testMergeJob["id"])
self.verifyFileMetaData(self.testMergeJob["id"],
myReport.getAllFilesFromStep("mergeReco"))
return
def test00(self):
"""
_test00_
Test that the job name prefix feature works
Test event threshold (single job creation)
Test that only closed lumis are used
Test check on express release
"""
insertClosedLumiBinds = []
for lumi in [1]:
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' : "Express",
'FILECOUNT' : filecount,
'INSERT_TIME' : self.currentTime,
'CLOSE_TIME' : 0 } )
self.fileset1.commit()
jobFactory = self.splitterFactory(package = "WMCore.WMBS",
subscription = self.subscription1)
jobGroups = jobFactory(maxInputEvents = 200)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
self.insertClosedLumiDAO.execute(binds = insertClosedLumiBinds,
transaction = False)
jobGroups = jobFactory(maxInputEvents = 200)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
self.finalCloseLumis()
jobGroups = jobFactory(maxInputEvents = 200)
self.assertEqual(len(jobGroups), 0,
"ERROR: JobFactory should have returned no JobGroup")
self.releaseExpressDAO.execute(binds = { 'RUN' : 1 }, transaction = False)
jobGroups = jobFactory(maxInputEvents = 200)
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("Express-"),
"ERROR: Job has wrong name")
self.assertEqual(self.getNumActiveSplitLumis(), 0,
"ERROR: Split lumis were created")
return
def setUp(self):
"""
_setUp_
Setup the database and WMBS for the test.
"""
self.testInit = TestInit(__file__)
self.testInit.setLogging()
self.testInit.setDatabaseConnection()
self.testInit.setSchema(
customModules=["WMComponent.DBS3Buffer", "WMCore.WMBS"],
useDefault=False)
myThread = threading.currentThread()
self.daofactory = DAOFactory(package="WMCore.WMBS",
logger=myThread.logger,
dbinterface=myThread.dbi)
self.dbsfactory = DAOFactory(package="WMComponent.DBS3Buffer",
logger=myThread.logger,
dbinterface=myThread.dbi)
locationAction = self.daofactory(classname="Locations.New")
locationAction.execute(siteName="site1", seName="cmssrm.fnal.gov")
inputFile = File(lfn="/path/to/some/lfn",
size=10,
events=10,
locations="cmssrm.fnal.gov")
inputFile.create()
inputFileset = Fileset(name="InputFileset")
inputFileset.create()
inputFileset.addFile(inputFile)
inputFileset.commit()
unmergedFileset = Fileset(name="UnmergedFileset")
unmergedFileset.create()
mergedFileset = Fileset(name="MergedFileset")
mergedFileset.create()
procWorkflow = Workflow(spec="wf001.xml",
owner="Steve",
name="TestWF",
task="/TestWF/None")
procWorkflow.create()
procWorkflow.addOutput("outputRECORECO", unmergedFileset)
mergeWorkflow = Workflow(spec="wf002.xml",
owner="Steve",
name="MergeWF",
task="/MergeWF/None")
mergeWorkflow.create()
mergeWorkflow.addOutput("Merged", mergedFileset)
insertWorkflow = self.dbsfactory(classname="InsertWorkflow")
insertWorkflow.execute("TestWF", "/TestWF/None", 0, 0, 0, 0)
insertWorkflow.execute("MergeWF", "/MergeWF/None", 0, 0, 0, 0)
self.procSubscription = Subscription(fileset=inputFileset,
workflow=procWorkflow,
split_algo="FileBased",
type="Processing")
self.procSubscription.create()
self.procSubscription.acquireFiles()
self.mergeSubscription = Subscription(fileset=unmergedFileset,
workflow=mergeWorkflow,
split_algo="WMBSMergeBySize",
type="Merge")
self.mergeSubscription.create()
self.procJobGroup = JobGroup(subscription=self.procSubscription)
self.procJobGroup.create()
self.mergeJobGroup = JobGroup(subscription=self.mergeSubscription)
self.mergeJobGroup.create()
self.testJob = Job(name="testJob", files=[inputFile])
self.testJob.create(group=self.procJobGroup)
self.testJob["state"] = "complete"
myThread = threading.currentThread()
self.daofactory = DAOFactory(package="WMCore.WMBS",
logger=myThread.logger,
dbinterface=myThread.dbi)
self.stateChangeAction = self.daofactory(classname="Jobs.ChangeState")
self.setFWJRAction = self.daofactory(classname="Jobs.SetFWJRPath")
self.getJobTypeAction = self.daofactory(classname="Jobs.GetType")
locationAction = self.daofactory(classname="Locations.New")
locationAction.execute(siteName="cmssrm.fnal.gov")
self.stateChangeAction.execute(jobs=[self.testJob])
self.tempDir = tempfile.mkdtemp()
return
class WMBSHelper(WMConnectionBase):
"""
_WMBSHelper_
Interface between the WorkQueue and WMBS.
"""
def __init__(self,
wmSpec,
taskName,
blockName=None,
mask=None,
cachepath='.'):
"""
_init_
Initialize DAOs and other things needed.
"""
self.block = blockName
self.mask = mask
self.wmSpec = wmSpec
self.topLevelTask = wmSpec.getTask(taskName)
self.cachepath = cachepath
self.isDBS = True
self.topLevelFileset = None
self.topLevelSubscription = None
self.mergeOutputMapping = {}
# Initiate the pieces you need to run your own DAOs
WMConnectionBase.__init__(self, "WMCore.WMBS")
myThread = threading.currentThread()
self.dbsDaoFactory = DAOFactory(package="WMComponent.DBS3Buffer",
logger=myThread.logger,
dbinterface=myThread.dbi)
# DAOs from WMBS for file commit
self.setParentage = self.daofactory(classname="Files.SetParentage")
self.setFileRunLumi = self.daofactory(classname="Files.AddRunLumi")
self.setFileLocation = self.daofactory(
classname="Files.SetLocationForWorkQueue")
self.setFileAddChecksum = self.daofactory(
classname="Files.AddChecksumByLFN")
self.addFileAction = self.daofactory(classname="Files.Add")
self.addToFileset = self.daofactory(classname="Files.AddDupsToFileset")
self.getLocations = self.daofactory(classname="Locations.ListSites")
self.getLocationInfo = self.daofactory(
classname="Locations.GetSiteInfo")
# DAOs from DBSBuffer for file commit
self.dbsCreateFiles = self.dbsDaoFactory(
classname="DBSBufferFiles.Add")
self.dbsSetLocation = self.dbsDaoFactory(
classname="DBSBufferFiles.SetLocationByLFN")
self.dbsInsertLocation = self.dbsDaoFactory(
classname="DBSBufferFiles.AddLocation")
self.dbsSetChecksum = self.dbsDaoFactory(
classname="DBSBufferFiles.AddChecksumByLFN")
self.dbsInsertWorkflow = self.dbsDaoFactory(classname="InsertWorkflow")
# Added for file creation bookkeeping
self.dbsFilesToCreate = []
self.addedLocations = []
self.wmbsFilesToCreate = []
self.insertedBogusDataset = -1
return
def createSandbox(self):
"""Create the runtime sandbox"""
sandboxCreator = SandboxCreator()
sandboxCreator.makeSandbox(self.cachepath, self.wmSpec)
def createTopLevelFileset(self, topLevelFilesetName=None):
"""
_createTopLevelFileset_
Create the top level fileset for the workflow. If the name of the top
level fileset is not given create one.
"""
if topLevelFilesetName == None:
filesetName = (
"%s-%s" %
(self.wmSpec.name(), self.wmSpec.getTopLevelTask()[0].name()))
if self.block:
filesetName += "-%s" % self.block
if self.mask:
from hashlib import md5
mask_string = ",".join(
["%s=%s" % (x, self.mask[x]) for x in sorted(self.mask)])
filesetName += "-%s" % md5(mask_string).hexdigest()
else:
filesetName = topLevelFilesetName
self.topLevelFileset = Fileset(filesetName)
self.topLevelFileset.create()
return
def outputFilesetName(self, task, outputModuleName):
"""
_outputFilesetName_
Generate an output fileset name for the given task and output module.
"""
if task.taskType() == "Merge":
outputFilesetName = "%s/merged-%s" % (task.getPathName(),
outputModuleName)
else:
outputFilesetName = "%s/unmerged-%s" % (task.getPathName(),
outputModuleName)
return outputFilesetName
def createSubscription(self, task, fileset, alternativeFilesetClose=False):
"""
_createSubscription_
Create subscriptions in WMBS for all the tasks in the spec. This
includes filesets, workflows and the output map for each task.
"""
# create runtime sandbox for workflow
self.createSandbox()
#FIXME: Let workflow put in values if spec is missing them
workflow = Workflow(
spec=self.wmSpec.specUrl(),
owner=self.wmSpec.getOwner()["name"],
dn=self.wmSpec.getOwner().get("dn", "unknown"),
group=self.wmSpec.getOwner().get("group", "unknown"),
owner_vogroup=self.wmSpec.getOwner().get("vogroup", "DEFAULT"),
owner_vorole=self.wmSpec.getOwner().get("vorole", "DEFAULT"),
name=self.wmSpec.name(),
task=task.getPathName(),
wfType=self.wmSpec.getDashboardActivity(),
alternativeFilesetClose=alternativeFilesetClose)
workflow.create()
subscription = Subscription(fileset=fileset,
workflow=workflow,
split_algo=task.jobSplittingAlgorithm(),
type=task.getPrimarySubType())
if subscription.exists():
subscription.load()
msg = "Subscription %s already exists for %s (you may ignore file insertion messages below, existing files wont be duplicated)"
self.logger.info(msg % (subscription['id'], task.getPathName()))
else:
subscription.create()
for site in task.siteWhitelist():
subscription.addWhiteBlackList([{
"site_name": site,
"valid": True
}])
for site in task.siteBlacklist():
subscription.addWhiteBlackList([{
"site_name": site,
"valid": False
}])
if self.topLevelSubscription == None:
self.topLevelSubscription = subscription
logging.info("Top level subscription created: %s" %
subscription["id"])
else:
logging.info("Child subscription created: %s" % subscription["id"])
outputModules = task.getOutputModulesForTask()
for outputModule in outputModules:
for outputModuleName in outputModule.listSections_():
outputFileset = Fileset(
self.outputFilesetName(task, outputModuleName))
outputFileset.create()
outputFileset.markOpen(True)
mergedOutputFileset = None
for childTask in task.childTaskIterator():
if childTask.data.input.outputModule == outputModuleName:
if childTask.taskType() == "Merge":
mergedOutputFileset = Fileset(
self.outputFilesetName(childTask, "Merged"))
mergedOutputFileset.create()
mergedOutputFileset.markOpen(True)
primaryDataset = getattr(
getattr(outputModule, outputModuleName),
"primaryDataset", None)
if primaryDataset != None:
self.mergeOutputMapping[
mergedOutputFileset.id] = primaryDataset
self.createSubscription(childTask, outputFileset,
alternativeFilesetClose)
if mergedOutputFileset == None:
workflow.addOutput(outputModuleName, outputFileset,
outputFileset)
else:
workflow.addOutput(outputModuleName, outputFileset,
mergedOutputFileset)
return self.topLevelSubscription
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:
raise RuntimeError, 'Invalid value "%s" for %s' % (
self.mask.get(key), key)
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]['se_name'])
except StandardError, ex:
self.logger.error(
'Error getting storage element for "%s": %s' %
(site, str(ex)))
if not locations:
raise RuntimeError, "No locations to inject Monte Carlo work to, unable to proceed"
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 = 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 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 __call__(self, filesetToProcess):
"""
The algorithm itself
"""
# Get configuration
initObj = WMInit()
initObj.setLogging()
initObj.setDatabaseConnection(os.getenv("DATABASE"), \
os.getenv('DIALECT'), os.getenv("DBSOCK"))
myThread = threading.currentThread()
daofactory = DAOFactory(package = "WMCore.WMBS" , \
logger = myThread.logger, \
dbinterface = myThread.dbi)
lastFileset = daofactory(classname="Fileset.ListFilesetByTask")
lastWorkflow = daofactory(classname="Workflow.LoadFromTask")
subsRun = daofactory(\
classname = "Subscriptions.LoadFromFilesetWorkflow")
successJob = daofactory(classname="Subscriptions.SucceededJobs")
allJob = daofactory(classname="Subscriptions.Jobs")
fileInFileset = daofactory(classname="Files.InFileset")
# Get the start Run if asked
startRun = (filesetToProcess.name).split(":")[3]
logging.debug("the T0Feeder is processing %s" % \
filesetToProcess.name)
logging.debug("the fileset name %s" % \
(filesetToProcess.name).split(":")[0])
fileType = (filesetToProcess.name).split(":")[2]
crabTask = filesetToProcess.name.split(":")[0]
LASTIME = filesetToProcess.lastUpdate
tries = 1
while True:
try:
myRequester = JSONRequests(url="vocms52.cern.ch:8889")
requestResult = myRequester.get("/tier0/runs")
except:
logging.debug("T0Reader call error...")
if tries == self.maxRetries:
return
else:
tries += 1
continue
logging.debug("T0ASTRunChain feeder queries done ...")
now = time.time()
break
for listRun in requestResult[0]:
if startRun != 'None' and int(listRun['run']) >= int(startRun):
if listRun['status'] =='CloseOutExport' or listRun\
['status']=='Complete' or listRun['status']=='CloseOutT1Skimming':
crabWorkflow = lastWorkflow.execute(task=crabTask)
crabFileset = lastFileset.execute\
(task=crabTask)
crabrunFileset = Fileset(\
name = crabFileset[0]["name"].split(':')[0].split\
('-Run')[0]+ '-Run' + str(listRun['run']) + ":" + \
":".join(crabFileset[0]['name'].split(':')[1:]) )
if crabrunFileset.exists() > 0:
crabrunFileset.load()
currSubs = subsRun.execute\
(crabrunFileset.id, crabWorkflow[0]['id'])
if currSubs:
listsuccessJob = successJob.execute(\
subscription=currSubs['id'])
listallJob = allJob.execute(\
subscription=currSubs['id'])
if len(listsuccessJob) == len(listallJob):
for currid in listsuccessJob:
currjob = Job(id=currid)
currjob.load()
logging.debug("Reading FJR %s" %
currjob['fwjr_path'])
jobReport = readJobReport(
currjob['fwjr_path'])
if len(jobReport) > 0:
if jobReport[0].files:
for newFile in jobReport[0].files:
logging.debug(\
"Output path %s" %newFile['LFN'])
newFileToAdd = File(\
lfn=newFile['LFN'], locations ='caf.cern.ch')
LOCK.acquire()
if newFileToAdd.exists\
() == False :
newFileToAdd.create()
else:
newFileToAdd.loadData()
LOCK.release()
listFile = \
fileInFileset.execute(filesetToProcess.id)
if {'fileid': \
newFileToAdd['id']} not in listFile:
filesetToProcess.addFile(\
newFileToAdd)
filesetToProcess\
.setLastUpdate(now)
filesetToProcess.commit()
logging.debug(\
"new file created/loaded and added by T0ASTRunChain...")
elif jobReport[0].analysisFiles:
for newFile in jobReport\
[0].analysisFiles:
logging.debug(\
"Ouput path %s " %newFile['LFN'])
newFileToAdd = File(\
lfn=newFile['LFN'], locations ='caf.cern.ch')
LOCK.acquire()
if newFileToAdd.exists\
() == False :
newFileToAdd.create()
else:
newFileToAdd.loadData()
LOCK.release()
listFile = \
fileInFileset.execute(filesetToProcess.id)
if {'fileid': newFileToAdd\
['id']} not in listFile:
logging.debug\
("%s loaded and added by T0ASTRunChain" %newFile['LFN'])
filesetToProcess.addFile\
(newFileToAdd)
filesetToProcess.\
setLastUpdate(now)
filesetToProcess.commit()
logging.debug(\
"new file created/loaded added by T0ASTRunChain...")
else:
break #Missed fjr - Try next time
# Commit the fileset
logging.debug("Test purge in T0ASTRunChain ...")
filesetToProcess.load()
LASTIME = filesetToProcess.lastUpdate
# For re-opned fileset or empty, try until the purge time
if (int(now) / 3600 - LASTIME / 3600) > self.reopenTime:
filesetToProcess.setLastUpdate(time.time())
filesetToProcess.commit()
if (int(now) / 3600 - LASTIME / 3600) > self.purgeTime:
filesetToProcess.markOpen(False)
logging.debug("Purge Done...")
def testMask(self):
"""
_testMask_
Test the new mask setup
"""
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['mask'].addRunAndLumis(run=100, lumis=[101, 102])
testJob['mask'].addRunAndLumis(run=200, lumis=[201, 202])
testJob.create(group=testJobGroup)
loadJob = Job(id=testJob.exists())
loadJob.loadData()
runs = loadJob['mask'].getRunAndLumis()
self.assertEqual(len(runs), 2)
self.assertEqual(runs[100], [[101, 102]])
self.assertEqual(runs[200], [[201, 202]])
bigRun = Run(100, *[101, 102, 103, 104])
badRun = Run(300, *[1001, 1002])
result = loadJob['mask'].filterRunLumisByMask([bigRun, badRun])
self.assertEqual(len(result), 1)
alteredRun = result.pop()
self.assertEqual(alteredRun.run, 100)
self.assertEqual(alteredRun.lumis, [101, 102])
run0 = Run(300, *[1001, 1002])
run1 = Run(300, *[1001, 1002])
loadJob['mask'].filterRunLumisByMask([run0, run1])
return
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.testInit = TestInit(__file__)
self.testInit.setLogging()
self.testInit.setDatabaseConnection()
self.testInit.setSchema(customModules=["WMCore.WMBS"],
useDefault=False)
myThread = threading.currentThread()
daofactory = DAOFactory(package="WMCore.WMBS",
logger=myThread.logger,
dbinterface=myThread.dbi)
locationAction = daofactory(classname="Locations.New")
locationAction.execute(siteName="site1", seName="somese.cern.ch")
self.multipleFileFileset = Fileset(name="TestFileset1")
self.multipleFileFileset.create()
for i in range(10):
newFile = File(makeUUID(),
size=1000,
events=100,
locations="somese.cern.ch")
newFile.addRun(Run(i, *[45 + i]))
newFile.create()
self.multipleFileFileset.addFile(newFile)
self.multipleFileFileset.commit()
self.singleFileFileset = Fileset(name="TestFileset2")
self.singleFileFileset.create()
newFile = File("/some/file/name",
size=1000,
events=100,
locations="somese.cern.ch")
newFile.addRun(Run(1, *[45]))
newFile.create()
self.singleFileFileset.addFile(newFile)
self.singleFileFileset.commit()
self.multipleFileRunset = Fileset(name="TestFileset3")
self.multipleFileRunset.create()
for i in range(10):
newFile = File(makeUUID(),
size=1000,
events=100,
locations="somese.cern.ch")
newFile.addRun(Run(i / 3, *[45]))
newFile.create()
self.multipleFileRunset.addFile(newFile)
self.multipleFileRunset.commit()
self.singleRunFileset = Fileset(name="TestFileset4")
self.singleRunFileset.create()
for i in range(10):
newFile = File(makeUUID(),
size=1000,
events=100,
locations="somese.cern.ch")
newFile.addRun(Run(1, *[45]))
newFile.create()
self.singleRunFileset.addFile(newFile)
self.singleRunFileset.commit()
self.singleRunMultipleLumi = Fileset(name="TestFileset5")
self.singleRunMultipleLumi.create()
for i in range(10):
newFile = File(makeUUID(),
size=1000,
events=100,
locations="somese.cern.ch")
newFile.addRun(Run(1, *[45 + i]))
newFile.create()
self.singleRunMultipleLumi.addFile(newFile)
self.singleRunMultipleLumi.commit()
testWorkflow = Workflow(spec="spec.xml",
owner="mnorman",
name="wf001",
task="Test")
testWorkflow.create()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset,
workflow=testWorkflow,
split_algo="RunBased",
type="Processing")
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset,
workflow=testWorkflow,
split_algo="RunBased",
type="Processing")
self.multipleRunSubscription = Subscription(
fileset=self.multipleFileRunset,
workflow=testWorkflow,
split_algo="RunBased",
type="Processing")
self.singleRunSubscription = Subscription(
fileset=self.singleRunFileset,
workflow=testWorkflow,
split_algo="RunBased",
type="Processing")
self.singleRunMultipleLumiSubscription = Subscription(
fileset=self.singleRunMultipleLumi,
workflow=testWorkflow,
split_algo="RunBased",
type="Processing")
self.multipleFileSubscription.create()
self.singleFileSubscription.create()
self.multipleRunSubscription.create()
self.singleRunSubscription.create()
self.singleRunMultipleLumiSubscription.create()
return
def testGetOutputParentLFNs(self):
"""
_testGetOutputParentLFNs_
Verify that the getOutputDBSParentLFNs() method returns the correct
parent LFNs.
"""
testWorkflow = Workflow(spec="spec.xml",
owner="Simon",
name="wf001",
task="Test")
testWorkflow.create()
testWMBSFileset = Fileset(name="TestFileset")
testWMBSFileset.create()
testSubscription = Subscription(fileset=testWMBSFileset,
workflow=testWorkflow)
testSubscription.create()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testFileA = File(lfn="/this/is/a/lfnA",
size=1024,
events=10,
merged=True)
testFileB = File(lfn="/this/is/a/lfnB",
size=1024,
events=10,
merged=True)
testFileC = File(lfn="/this/is/a/lfnC",
size=1024,
events=10,
merged=False)
testFileD = File(lfn="/this/is/a/lfnD",
size=1024,
events=10,
merged=False)
testFileE = File(lfn="/this/is/a/lfnE",
size=1024,
events=10,
merged=True)
testFileF = File(lfn="/this/is/a/lfnF",
size=1024,
events=10,
merged=True)
testFileA.create()
testFileB.create()
testFileC.create()
testFileD.create()
testFileE.create()
testFileF.create()
testFileE.addChild(testFileC["lfn"])
testFileF.addChild(testFileD["lfn"])
testJobA = Job(name="TestJob", files=[testFileA, testFileB])
testJobA["couch_record"] = "somecouchrecord"
testJobA["location"] = "test.site.ch"
testJobA.create(group=testJobGroup)
testJobA.associateFiles()
testJobB = Job(name="TestJobB", files=[testFileC, testFileD])
testJobB["couch_record"] = "somecouchrecord"
testJobB["location"] = "test.site.ch"
testJobB.create(group=testJobGroup)
testJobB.associateFiles()
goldenLFNs = ["/this/is/a/lfnA", "/this/is/a/lfnB"]
parentLFNs = testJobA.getOutputDBSParentLFNs()
for parentLFN in parentLFNs:
assert parentLFN in goldenLFNs, \
"ERROR: Unknown lfn: %s" % parentLFN
goldenLFNs.remove(parentLFN)
assert len(goldenLFNs) == 0, \
"ERROR: LFNs are missing: %s" % goldenLFNs
goldenLFNs = ["/this/is/a/lfnE", "/this/is/a/lfnF"]
parentLFNs = testJobB.getOutputDBSParentLFNs()
for parentLFN in parentLFNs:
assert parentLFN in goldenLFNs, \
"ERROR: Unknown lfn: %s" % parentLFN
goldenLFNs.remove(parentLFN)
assert len(goldenLFNs) == 0, \
"ERROR: LFNs are missing..."
return
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.testInit = TestInit(__file__)
self.testInit.setLogging()
self.testInit.setDatabaseConnection()
self.testInit.setSchema(customModules = ["WMCore.WMBS"],
useDefault = False)
myThread = threading.currentThread()
daofactory = DAOFactory(package = "WMCore.WMBS",
logger = myThread.logger,
dbinterface = myThread.dbi)
locationAction = daofactory(classname = "Locations.New")
locationAction.execute(siteName = 's1', seName = "somese.cern.ch")
locationAction.execute(siteName = 's2', seName = "otherse.cern.ch")
self.multipleFileFileset = Fileset(name = "TestFileset1")
self.multipleFileFileset.create()
parentFile = File('/parent/lfn/', size = 1000, events = 100,
locations = set(["somese.cern.ch"]))
parentFile.create()
for i in range(10):
newFile = File(makeUUID(), size = 1000, events = 100,
locations = set(["somese.cern.ch"]))
newFile.create()
newFile.addParent(lfn = parentFile['lfn'])
self.multipleFileFileset.addFile(newFile)
self.multipleFileFileset.commit()
self.singleFileFileset = Fileset(name = "TestFileset2")
self.singleFileFileset.create()
newFile = File("/some/file/name", size = 1000, events = 100,
locations = set(["somese.cern.ch"]))
newFile.create()
self.singleFileFileset.addFile(newFile)
self.singleFileFileset.commit()
self.multipleSiteFileset = Fileset(name = "TestFileset3")
self.multipleSiteFileset.create()
for i in range(5):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.setLocation("somese.cern.ch")
newFile.create()
self.multipleSiteFileset.addFile(newFile)
for i in range(5):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.setLocation(["somese.cern.ch","otherse.cern.ch"])
newFile.create()
self.multipleSiteFileset.addFile(newFile)
self.multipleSiteFileset.commit()
testWorkflow = Workflow(spec = "spec.xml", owner = "Steve",
name = "wf001", task="Test")
testWorkflow.create()
self.multipleFileSubscription = Subscription(fileset = self.multipleFileFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.multipleFileSubscription.create()
self.singleFileSubscription = Subscription(fileset = self.singleFileFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.singleFileSubscription.create()
self.multipleSiteSubscription = Subscription(fileset = self.multipleSiteFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.multipleSiteSubscription.create()
return
def testCompleteJobInput(self):
"""
_testCompleteJobInput_
Verify the correct output of the CompleteInput DAO. This should mark
the input for a job as complete once all the jobs that run over a
particular file have complete successfully.
"""
testWorkflow = Workflow(spec="spec.xml",
owner="Steve",
name="wf001",
task="Test")
bogusWorkflow = Workflow(spec="spec1.xml",
owner="Steve",
name="wf002",
task="Test")
testWorkflow.create()
bogusWorkflow.create()
testFileset = Fileset(name="TestFileset")
bogusFileset = Fileset(name="BogusFileset")
testFileset.create()
bogusFileset.create()
testSubscription = Subscription(fileset=testFileset,
workflow=testWorkflow)
bogusSubscription = Subscription(fileset=bogusFileset,
workflow=bogusWorkflow)
testSubscription.create()
bogusSubscription.create()
testFileA = File(lfn=makeUUID(), locations="setest.site.ch")
testFileB = File(lfn=makeUUID(), locations="setest.site.ch")
testFileA.create()
testFileB.create()
testFileset.addFile([testFileA, testFileB])
bogusFileset.addFile([testFileA, testFileB])
testFileset.commit()
bogusFileset.commit()
testSubscription.acquireFiles([testFileA, testFileB])
bogusSubscription.acquireFiles([testFileA, testFileB])
testJobGroup = JobGroup(subscription=testSubscription)
bogusJobGroup = JobGroup(subscription=bogusSubscription)
testJobGroup.create()
bogusJobGroup.create()
testJobA = Job(name="TestJobA", files=[testFileA])
testJobB = Job(name="TestJobB", files=[testFileA])
testJobC = Job(name="TestJobC", files=[testFileB])
bogusJob = Job(name="BogusJob", files=[testFileA, testFileB])
testJobA.create(group=testJobGroup)
testJobB.create(group=testJobGroup)
testJobC.create(group=testJobGroup)
bogusJob.create(group=bogusJobGroup)
testJobA["outcome"] = "success"
testJobB["outcome"] = "failure"
testJobC["outcome"] = "success"
testJobA.save()
testJobB.save()
testJobC.save()
testJobA.completeInputFiles()
compFiles = len(testSubscription.filesOfStatus("Completed"))
assert compFiles == 0, \
"Error: test sub has wrong number of complete files: %s" % compFiles
testJobB["outcome"] = "success"
testJobB.save()
testJobB.completeInputFiles()
availFiles = len(testSubscription.filesOfStatus("Available"))
assert availFiles == 0, \
"Error: test sub has wrong number of available files: %s" % availFiles
acqFiles = len(testSubscription.filesOfStatus("Acquired"))
assert acqFiles == 1, \
"Error: test sub has wrong number of acquired files: %s" % acqFiles
compFiles = len(testSubscription.filesOfStatus("Completed"))
assert compFiles == 1, \
"Error: test sub has wrong number of complete files: %s" % compFiles
failFiles = len(testSubscription.filesOfStatus("Failed"))
assert failFiles == 0, \
"Error: test sub has wrong number of failed files: %s" % failFiles
availFiles = len(bogusSubscription.filesOfStatus("Available"))
assert availFiles == 0, \
"Error: test sub has wrong number of available files: %s" % availFiles
acqFiles = len(bogusSubscription.filesOfStatus("Acquired"))
assert acqFiles == 2, \
"Error: test sub has wrong number of acquired files: %s" % acqFiles
compFiles = len(bogusSubscription.filesOfStatus("Completed"))
assert compFiles == 0, \
"Error: test sub has wrong number of complete files: %s" % compFiles
failFiles = len(bogusSubscription.filesOfStatus("Failed"))
assert failFiles == 0, \
"Error: test sub has wrong number of failed files: %s" % failFiles
return
def testB_NoFileSplitNoHardLimit(self):
"""
_testB_NoFileSplitNoHardLimit_
In this case we don't split on file boundaries, check different combination of files
make sure we make the most of the splitting, e.g. include many zero event files in
a single job.
"""
splitter = SplitterFactory()
# Create 100 files with 7 lumi per file and 0 events per lumi on average.
testSubscription = self.createSubscription(nFiles=100, lumisPerFile=7, twoSites=False,
nEventsPerFile=0)
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# First test, the optimal settings are 360 events per job
# As we have files with 0 events per lumi, this will configure the splitting to
# a single job containing all files
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=360,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 0, "There aren't enough events, so it should have 0 job groups")
# we close this fileset to get it moving
fileset = testSubscription.getFileset()
fileset.markOpen(False)
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=360,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1, "There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 1, "There should be 1 job")
self.assertEqual(len(jobs[0]['input_files']), 100, "All 100 files must be in the job")
self.assertEqual(jobs[0]['estimatedMemoryUsage'], 2300)
self.assertEqual(jobs[0]['estimatedDiskUsage'], 0)
self.assertEqual(jobs[0]['estimatedJobTime'], 0)
# Create 7 files, each one with different lumi/event distributions
testFileset = Fileset(name="FilesetA")
testFileset.create()
testFileA = self.createFile("/this/is/file1", 250, 0, 5, "T2_CH_CERN")
testFileB = self.createFile("/this/is/file2", 600, 1, 1, "T2_CH_CERN")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2, "T2_CH_CERN")
testFileD = self.createFile("/this/is/file4", 100, 3, 1, "T2_CH_CERN")
testFileE = self.createFile("/this/is/file5", 30, 4, 1, "T2_CH_CERN")
testFileF = self.createFile("/this/is/file6", 10, 5, 1, "T2_CH_CERN")
testFileG = self.createFile("/this/is/file7", 151, 6, 3, "T2_CH_CERN")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testFileset.commit()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiBased",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# Optimal settings are: jobs with 150 events per job
# This means, the first file must be splitted in 3 lumis per job which would leave room
# for another lumi in the second job, but the second file has a lumi too big for that
# The 3rd job only contains the second file, the fourth and fifth job split the third file
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=150,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1, "There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 7, "7 jobs must be in the jobgroup")
self.assertEqual(jobs[0]["mask"].getRunAndLumis(), {0: [[0, 2]]}, "Wrong mask for the first job")
self.assertEqual(jobs[0]["estimatedJobTime"], 150 * 12)
self.assertEqual(jobs[0]["estimatedDiskUsage"], 150 * 400)
self.assertEqual(jobs[1]["mask"].getRunAndLumis(), {0: [[3, 4]]}, "Wrong mask for the second job")
self.assertEqual(jobs[1]["estimatedJobTime"], 100 * 12)
self.assertEqual(jobs[1]["estimatedDiskUsage"], 100 * 400)
self.assertEqual(jobs[2]["mask"].getRunAndLumis(), {1: [[1, 1]]}, "Wrong mask for the third job")
self.assertEqual(jobs[2]["estimatedJobTime"], 600 * 12)
self.assertEqual(jobs[2]["estimatedDiskUsage"], 600 * 400)
self.assertEqual(jobs[3]["mask"].getRunAndLumis(), {2: [[4, 4]]}, "Wrong mask for the fourth job")
self.assertEqual(jobs[3]["estimatedJobTime"], 600 * 12)
self.assertEqual(jobs[3]["estimatedDiskUsage"], 600 * 400)
self.assertEqual(jobs[4]["mask"].getRunAndLumis(), {2: [[5, 5]]}, "Wrong mask for the fifth job")
self.assertEqual(jobs[4]["estimatedJobTime"], 600 * 12)
self.assertEqual(jobs[4]["estimatedDiskUsage"], 600 * 400)
self.assertEqual(jobs[5]["mask"].getRunAndLumis(),
{3: [[3, 3]], 4: [[4, 4]], 5: [[5, 5]]}, "Wrong mask for the sixth job")
self.assertEqual(jobs[5]["estimatedJobTime"], 140 * 12)
self.assertEqual(jobs[5]["estimatedDiskUsage"], 140 * 400)
self.assertEqual(jobs[6]["mask"].getRunAndLumis(), {6: [[18, 20]]}, "Wrong mask for the seventh job")
self.assertEqual(jobs[6]["estimatedJobTime"], 150 * 12)
self.assertEqual(jobs[6]["estimatedDiskUsage"], 150 * 400)
for job in jobs:
self.assertEqual(job["estimatedMemoryUsage"], 2300)
# Test interactions of this algorithm with splitOnRun = True
# Make 2 files, one with 3 runs and a second one with the last run of the first
fileA = File(lfn="/this/is/file1a", size=1000,
events=2400)
lumiListA = []
lumiListB = []
lumiListC = []
for lumi in range(8):
lumiListA.append(1 + lumi)
lumiListB.append(1 + lumi)
lumiListC.append(1 + lumi)
fileA.addRun(Run(1, *lumiListA))
fileA.addRun(Run(2, *lumiListA))
fileA.addRun(Run(3, *lumiListA))
fileA.setLocation("T1_US_FNAL_Disk")
fileB = self.createFile('/this/is/file2a', 200, 3, 5, "T1_US_FNAL_Disk")
testFileset = Fileset(name='FilesetB')
testFileset.create()
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.commit()
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiBased",
type="Processing")
testSubscription.create()
jobFactory = splitter(package="WMCore.WMBS",
subscription=testSubscription)
# The settings for this splitting are 700 events per job
jobGroups = jobFactory(splitOnRun=True,
halt_job_on_file_boundaries=False,
events_per_job=700,
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1, "There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 6, "Six jobs must be in the jobgroup")
self.assertEqual(jobs[0]["estimatedJobTime"], 700 * 12)
self.assertEqual(jobs[0]["estimatedDiskUsage"], 700 * 400)
self.assertEqual(jobs[1]["estimatedJobTime"], 100 * 12)
self.assertEqual(jobs[1]["estimatedDiskUsage"], 100 * 400)
self.assertEqual(jobs[2]["estimatedJobTime"], 700 * 12)
self.assertEqual(jobs[2]["estimatedDiskUsage"], 700 * 400)
self.assertEqual(jobs[3]["estimatedJobTime"], 100 * 12)
self.assertEqual(jobs[3]["estimatedDiskUsage"], 100 * 400)
self.assertEqual(jobs[4]["estimatedJobTime"], 700 * 12)
self.assertEqual(jobs[4]["estimatedDiskUsage"], 700 * 400)
self.assertEqual(jobs[5]["estimatedJobTime"], 300 * 12)
self.assertEqual(jobs[5]["estimatedDiskUsage"], 300 * 400)
def createTestJobGroup(self,
nJobs=10,
retry_count=1,
workloadPath='test',
fwjrPath=None,
workloadName=makeUUID()):
"""
Creates a group of several jobs
"""
myThread = threading.currentThread()
myThread.transaction.begin()
testWorkflow = Workflow(spec=workloadPath,
owner="cmsdataops",
group="cmsdataops",
name=workloadName,
task="/TestWorkload/ReReco")
testWorkflow.create()
testWMBSFileset = Fileset(name="TestFileset")
testWMBSFileset.create()
testSubscription = Subscription(fileset=testWMBSFileset,
workflow=testWorkflow)
testSubscription.create()
testJobGroup = JobGroup(subscription=testSubscription)
testJobGroup.create()
testFile0 = File(lfn="/this/is/a/parent", size=1024, events=10)
testFile0.addRun(Run(10, *[12312]))
testFile0.setLocation('malpaquet')
testFileA = File(lfn="/this/is/a/lfnA",
size=1024,
events=10,
first_event=88)
testFileA.addRun(Run(10, *[12312, 12313]))
testFileA.setLocation('malpaquet')
testFileB = File(lfn="/this/is/a/lfnB",
size=1024,
events=10,
first_event=88)
testFileB.addRun(Run(10, *[12314, 12315, 12316]))
testFileB.setLocation('malpaquet')
testFile0.create()
testFileA.create()
testFileB.create()
testFileA.addParent(lfn="/this/is/a/parent")
testFileB.addParent(lfn="/this/is/a/parent")
for i in range(0, nJobs):
testJob = Job(name=makeUUID())
testJob['retry_count'] = retry_count
testJob['retry_max'] = 10
testJob['mask'].addRunAndLumis(run=10, lumis=[12312])
testJob['mask'].addRunAndLumis(run=10, lumis=[12314, 12316])
testJob['mask']['FirstEvent'] = 100
testJob['cache_dir'] = os.path.join(self.testDir, testJob['name'])
testJob['fwjr_path'] = fwjrPath
os.mkdir(testJob['cache_dir'])
testJobGroup.add(testJob)
testJob.create(group=testJobGroup)
testJob.addFile(testFileA)
testJob.addFile(testFileB)
testJob.save()
testJobGroup.commit()
testSubscription.acquireFiles(files=[testFileA, testFileB])
testSubscription.save()
myThread.transaction.commit()
return testJobGroup
def setupForKillTest(self, baAPI=None):
"""
_setupForKillTest_
Inject a workflow into WMBS that has a processing task, a merge task and
a cleanup task. Inject files into the various tasks at various
processing states (acquired, complete, available...). Also create jobs
for each subscription in various states.
"""
myThread = threading.currentThread()
daoFactory = DAOFactory(package="WMCore.WMBS",
logger=myThread.logger,
dbinterface=myThread.dbi)
dummyLocationAction = daoFactory(classname="Locations.New")
changeStateAction = daoFactory(classname="Jobs.ChangeState")
resourceControl = ResourceControl()
resourceControl.insertSite(siteName='site1',
pnn='goodse.cern.ch',
ceName='site1',
plugin="TestPlugin")
resourceControl.insertThreshold(siteName='site1', taskType='Processing', \
maxSlots=10000, pendingSlots=10000)
userDN = 'someDN'
userAction = daoFactory(classname="Users.New")
userAction.execute(dn=userDN,
group_name='DEFAULT',
role_name='DEFAULT')
inputFileset = Fileset("input")
inputFileset.create()
inputFileA = File("lfnA", locations="goodse.cern.ch")
inputFileB = File("lfnB", locations="goodse.cern.ch")
inputFileC = File("lfnC", locations="goodse.cern.ch")
inputFileA.create()
inputFileB.create()
inputFileC.create()
inputFileset.addFile(inputFileA)
inputFileset.addFile(inputFileB)
inputFileset.addFile(inputFileC)
inputFileset.commit()
unmergedOutputFileset = Fileset("unmerged")
unmergedOutputFileset.create()
unmergedFileA = File("ulfnA", locations="goodse.cern.ch")
unmergedFileB = File("ulfnB", locations="goodse.cern.ch")
unmergedFileC = File("ulfnC", locations="goodse.cern.ch")
unmergedFileA.create()
unmergedFileB.create()
unmergedFileC.create()
unmergedOutputFileset.addFile(unmergedFileA)
unmergedOutputFileset.addFile(unmergedFileB)
unmergedOutputFileset.addFile(unmergedFileC)
unmergedOutputFileset.commit()
mainProcWorkflow = Workflow(spec="spec1",
owner="Steve",
name="Main",
task="Proc")
mainProcWorkflow.create()
mainProcMergeWorkflow = Workflow(spec="spec1",
owner="Steve",
name="Main",
task="ProcMerge")
mainProcMergeWorkflow.create()
mainCleanupWorkflow = Workflow(spec="spec1",
owner="Steve",
name="Main",
task="Cleanup")
mainCleanupWorkflow.create()
self.mainProcSub = Subscription(fileset=inputFileset,
workflow=mainProcWorkflow,
type="Processing")
self.mainProcSub.create()
self.mainProcSub.acquireFiles(inputFileA)
self.mainProcSub.completeFiles(inputFileB)
procJobGroup = JobGroup(subscription=self.mainProcSub)
procJobGroup.create()
self.procJobA = Job(name="ProcJobA")
self.procJobA["state"] = "new"
self.procJobA["location"] = "site1"
self.procJobB = Job(name="ProcJobB")
self.procJobB["state"] = "executing"
self.procJobB["location"] = "site1"
self.procJobC = Job(name="ProcJobC")
self.procJobC["state"] = "complete"
self.procJobC["location"] = "site1"
self.procJobA.create(procJobGroup)
self.procJobB.create(procJobGroup)
self.procJobC.create(procJobGroup)
self.mainMergeSub = Subscription(fileset=unmergedOutputFileset,
workflow=mainProcMergeWorkflow,
type="Merge")
self.mainMergeSub.create()
self.mainMergeSub.acquireFiles(unmergedFileA)
self.mainMergeSub.failFiles(unmergedFileB)
mergeJobGroup = JobGroup(subscription=self.mainMergeSub)
mergeJobGroup.create()
self.mergeJobA = Job(name="MergeJobA")
self.mergeJobA["state"] = "exhausted"
self.mergeJobA["location"] = "site1"
self.mergeJobB = Job(name="MergeJobB")
self.mergeJobB["state"] = "cleanout"
self.mergeJobB["location"] = "site1"
self.mergeJobC = Job(name="MergeJobC")
self.mergeJobC["state"] = "new"
self.mergeJobC["location"] = "site1"
self.mergeJobA.create(mergeJobGroup)
self.mergeJobB.create(mergeJobGroup)
self.mergeJobC.create(mergeJobGroup)
self.mainCleanupSub = Subscription(fileset=unmergedOutputFileset,
workflow=mainCleanupWorkflow,
type="Cleanup")
self.mainCleanupSub.create()
self.mainCleanupSub.acquireFiles(unmergedFileA)
self.mainCleanupSub.completeFiles(unmergedFileB)
cleanupJobGroup = JobGroup(subscription=self.mainCleanupSub)
cleanupJobGroup.create()
self.cleanupJobA = Job(name="CleanupJobA")
self.cleanupJobA["state"] = "new"
self.cleanupJobA["location"] = "site1"
self.cleanupJobB = Job(name="CleanupJobB")
self.cleanupJobB["state"] = "executing"
self.cleanupJobB["location"] = "site1"
self.cleanupJobC = Job(name="CleanupJobC")
self.cleanupJobC["state"] = "complete"
self.cleanupJobC["location"] = "site1"
self.cleanupJobA.create(cleanupJobGroup)
self.cleanupJobB.create(cleanupJobGroup)
self.cleanupJobC.create(cleanupJobGroup)
jobList = [
self.procJobA, self.procJobB, self.procJobC, self.mergeJobA,
self.mergeJobB, self.mergeJobC, self.cleanupJobA, self.cleanupJobB,
self.cleanupJobC
]
changeStateAction.execute(jobList)
if baAPI:
for job in jobList:
job['plugin'] = 'TestPlugin'
job['userdn'] = userDN
job['usergroup'] = 'DEFAULT'
job['userrole'] = 'DEFAULT'
job['custom']['location'] = 'site1'
baAPI.createNewJobs(wmbsJobs=jobList)
# We'll create an unrelated workflow to verify that it isn't affected
# by the killing code.
bogusFileset = Fileset("dontkillme")
bogusFileset.create()
bogusFileA = File("bogus/lfnA", locations="goodse.cern.ch")
bogusFileA.create()
bogusFileset.addFile(bogusFileA)
bogusFileset.commit()
bogusWorkflow = Workflow(spec="spec2",
owner="Steve",
name="Bogus",
task="Proc")
bogusWorkflow.create()
self.bogusSub = Subscription(fileset=bogusFileset,
workflow=bogusWorkflow,
type="Processing")
self.bogusSub.create()
self.bogusSub.acquireFiles(bogusFileA)
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 stuffWMBS(self, injected=True):
"""
_stuffWMBS_
Insert some dummy jobs, jobgroups, filesets, files and subscriptions
into WMBS to test job creation. Three completed job groups each
containing several files are injected. Another incomplete job group is
also injected. Also files are added to the "Mergeable" subscription as
well as to the output fileset for their jobgroups.
"""
locationAction = self.daoFactory(classname="Locations.New")
locationAction.execute(siteName="T2_CH_CERN", pnn="T2_CH_CERN")
locationAction.execute(siteName="T1_US_FNAL", pnn="T2_CH_CERN")
changeStateDAO = self.daoFactory(classname="Jobs.ChangeState")
self.mergeFileset = Fileset(name="mergeFileset")
self.mergeFileset.create()
self.bogusFileset = Fileset(name="bogusFileset")
self.bogusFileset.create()
self.mergeMergedFileset = Fileset(name="mergeMergedFileset")
self.mergeMergedFileset.create()
self.bogusMergedFileset = Fileset(name="bogusMergedFileset")
self.bogusMergedFileset.create()
mergeWorkflow = Workflow(name="mergeWorkflow",
spec="bunk2",
owner="Steve",
task="Test")
mergeWorkflow.create()
markWorkflow = self.daoFactory(
classname="Workflow.MarkInjectedWorkflows")
markWorkflow.execute(names=[mergeWorkflow.name], injected=injected)
self.mergeSubscription = Subscription(fileset=self.mergeFileset,
workflow=mergeWorkflow,
split_algo="WMBSMergeBySize")
self.mergeSubscription.create()
self.bogusSubscription = Subscription(fileset=self.bogusFileset,
workflow=mergeWorkflow,
split_algo="WMBSMergeBySize")
inputFileset = Fileset(name="inputFileset")
inputFileset.create()
inputWorkflow = Workflow(name="inputWorkflow",
spec="input",
owner="Steve",
task="Test")
inputWorkflow.create()
inputWorkflow.addOutput("output", self.mergeFileset,
self.mergeMergedFileset)
inputWorkflow.addOutput("output2", self.bogusFileset,
self.bogusMergedFileset)
bogusInputWorkflow = Workflow(name="bogusInputWorkflow",
spec="input",
owner="Steve",
task="Test")
bogusInputWorkflow.create()
inputSubscription = Subscription(fileset=inputFileset,
workflow=inputWorkflow)
inputSubscription.create()
bogusInputSubscription = Subscription(fileset=inputFileset,
workflow=bogusInputWorkflow)
bogusInputSubscription.create()
parentFile1 = File(lfn="parentFile1")
parentFile1.create()
parentFile2 = File(lfn="parentFile2")
parentFile2.create()
parentFile3 = File(lfn="parentFile3")
parentFile3.create()
parentFile4 = File(lfn="parentFile4")
parentFile4.create()
self.parentFileSite2 = File(lfn="parentFileSite2")
self.parentFileSite2.create()
jobGroup1 = JobGroup(subscription=inputSubscription)
jobGroup1.create()
jobGroup2 = JobGroup(subscription=inputSubscription)
jobGroup2.create()
jobGroup3 = JobGroup(subscription=bogusInputSubscription)
jobGroup3.create()
testJob1 = Job()
testJob1.addFile(parentFile1)
testJob1.create(jobGroup1)
testJob1["state"] = "cleanout"
testJob1["oldstate"] = "new"
testJob1["couch_record"] = "somejive"
testJob1["retry_count"] = 0
testJob1["outcome"] = "success"
testJob1.save()
changeStateDAO.execute([testJob1])
testJob1A = Job()
testJob1A.addFile(parentFile1)
testJob1A.create(jobGroup3)
testJob1A["state"] = "cleanout"
testJob1A["oldstate"] = "new"
testJob1A["couch_record"] = "somejive"
testJob1A["retry_count"] = 0
testJob1A["outcome"] = "failure"
testJob1A.save()
changeStateDAO.execute([testJob1A])
testJob2 = Job()
testJob2.addFile(parentFile2)
testJob2.create(jobGroup1)
testJob2["state"] = "cleanout"
testJob2["oldstate"] = "new"
testJob2["couch_record"] = "somejive"
testJob2["retry_count"] = 0
testJob2["outcome"] = "success"
testJob2.save()
changeStateDAO.execute([testJob2])
testJob3 = Job()
testJob3.addFile(parentFile3)
testJob3.create(jobGroup2)
testJob3["state"] = "cleanout"
testJob3["oldstate"] = "new"
testJob3["couch_record"] = "somejive"
testJob3["retry_count"] = 0
testJob3["outcome"] = "success"
testJob3.save()
changeStateDAO.execute([testJob3])
testJob4 = Job()
testJob4.addFile(parentFile4)
testJob4.create(jobGroup2)
testJob4["state"] = "cleanout"
testJob4["oldstate"] = "new"
testJob4["couch_record"] = "somejive"
testJob4["retry_count"] = 0
testJob4["outcome"] = "failure"
testJob4.save()
changeStateDAO.execute([testJob4])
# We'll simulate a failed split by event job that the merger should
# ignore.
parentFile5 = File(lfn="parentFile5")
parentFile5.create()
testJob5 = Job()
testJob5.addFile(parentFile5)
testJob5.create(jobGroup2)
testJob5["state"] = "cleanout"
testJob5["oldstate"] = "new"
testJob5["couch_record"] = "somejive"
testJob5["retry_count"] = 0
testJob5["outcome"] = "success"
testJob5.save()
changeStateDAO.execute([testJob5])
testJob6 = Job()
testJob6.addFile(parentFile5)
testJob6.create(jobGroup2)
testJob6["state"] = "cleanout"
testJob6["oldstate"] = "new"
testJob6["couch_record"] = "somejive"
testJob6["retry_count"] = 0
testJob6["outcome"] = "failure"
testJob6.save()
changeStateDAO.execute([testJob6])
testJob7 = Job()
testJob7.addFile(self.parentFileSite2)
testJob7.create(jobGroup2)
testJob7["state"] = "cleanout"
testJob7["oldstate"] = "new"
testJob7["couch_record"] = "somejive"
testJob7["retry_count"] = 0
testJob7["outcome"] = "success"
testJob7.save()
changeStateDAO.execute([testJob7])
badFile1 = File(lfn="badFile1",
size=10241024,
events=10241024,
first_event=0,
locations=set(["T2_CH_CERN"]))
badFile1.addRun(Run(1, *[45]))
badFile1.create()
badFile1.addParent(parentFile5["lfn"])
file1 = File(lfn="file1",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
file1.addRun(Run(1, *[45]))
file1.create()
file1.addParent(parentFile1["lfn"])
file2 = File(lfn="file2",
size=1024,
events=1024,
first_event=1024,
locations=set(["T2_CH_CERN"]))
file2.addRun(Run(1, *[45]))
file2.create()
file2.addParent(parentFile1["lfn"])
file3 = File(lfn="file3",
size=1024,
events=1024,
first_event=2048,
locations=set(["T2_CH_CERN"]))
file3.addRun(Run(1, *[45]))
file3.create()
file3.addParent(parentFile1["lfn"])
file4 = File(lfn="file4",
size=1024,
events=1024,
first_event=3072,
locations=set(["T2_CH_CERN"]))
file4.addRun(Run(1, *[45]))
file4.create()
file4.addParent(parentFile1["lfn"])
fileA = File(lfn="fileA",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileA.addRun(Run(1, *[46]))
fileA.create()
fileA.addParent(parentFile2["lfn"])
fileB = File(lfn="fileB",
size=1024,
events=1024,
first_event=1024,
locations=set(["T2_CH_CERN"]))
fileB.addRun(Run(1, *[46]))
fileB.create()
fileB.addParent(parentFile2["lfn"])
fileC = File(lfn="fileC",
size=1024,
events=1024,
first_event=2048,
locations=set(["T2_CH_CERN"]))
fileC.addRun(Run(1, *[46]))
fileC.create()
fileC.addParent(parentFile2["lfn"])
fileI = File(lfn="fileI",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileI.addRun(Run(2, *[46]))
fileI.create()
fileI.addParent(parentFile3["lfn"])
fileII = File(lfn="fileII",
size=1024,
events=1024,
first_event=1024,
locations=set(["T2_CH_CERN"]))
fileII.addRun(Run(2, *[46]))
fileII.create()
fileII.addParent(parentFile3["lfn"])
fileIII = File(lfn="fileIII",
size=1024,
events=1024,
first_event=2048,
locations=set(["T2_CH_CERN"]))
fileIII.addRun(Run(2, *[46]))
fileIII.create()
fileIII.addParent(parentFile3["lfn"])
fileIV = File(lfn="fileIV",
size=1024,
events=1024,
first_event=3072,
locations=set(["T2_CH_CERN"]))
fileIV.addRun(Run(2, *[46]))
fileIV.create()
fileIV.addParent(parentFile3["lfn"])
fileX = File(lfn="badFileA",
size=1024,
events=1024,
first_event=0,
locations=set(["T2_CH_CERN"]))
fileX.addRun(Run(1, *[47]))
fileX.create()
fileX.addParent(parentFile4["lfn"])
fileY = File(lfn="badFileB",
size=1024,
events=1024,
first_event=1024,
locations=set(["T2_CH_CERN"]))
fileY.addRun(Run(1, *[47]))
fileY.create()
fileY.addParent(parentFile4["lfn"])
fileZ = File(lfn="badFileC",
size=1024,
events=1024,
first_event=2048,
locations=set(["T2_CH_CERN"]))
fileZ.addRun(Run(1, *[47]))
fileZ.create()
fileZ.addParent(parentFile4["lfn"])
jobGroup1.output.addFile(file1)
jobGroup1.output.addFile(file2)
jobGroup1.output.addFile(file3)
jobGroup1.output.addFile(file4)
jobGroup1.output.addFile(fileA)
jobGroup1.output.addFile(fileB)
jobGroup1.output.addFile(fileC)
jobGroup1.output.commit()
jobGroup2.output.addFile(fileI)
jobGroup2.output.addFile(fileII)
jobGroup2.output.addFile(fileIII)
jobGroup2.output.addFile(fileIV)
jobGroup2.output.addFile(fileX)
jobGroup2.output.addFile(fileY)
jobGroup2.output.addFile(fileZ)
jobGroup2.output.addFile(badFile1)
jobGroup2.output.commit()
for file in [
file1, file2, file3, file4, fileA, fileB, fileC, fileI, fileII,
fileIII, fileIV, fileX, fileY, fileZ, badFile1
]:
self.mergeFileset.addFile(file)
self.bogusFileset.addFile(file)
self.mergeFileset.commit()
self.bogusFileset.commit()
return
