def doBlock( self, entity, fileset ):
connection = urlopen( self.nodeURL + "&block=%s" % quote( entity ) )
aString = connection.read()
connection.close()
if aString[2:8] != "phedex":
print "PhEDExNotifier: bad string from server follows."
print "%s" % aString
phedex = eval( aString.replace( "null", "None" ), {}, {} )
blocks = phedex[ 'phedex' ][ 'block' ]
if len( blocks ) != 1:
print "PhEDExNotifier: Found %d blocks, expected 1, will only consider first block" % len( blocks)
files = blocks[0][ 'file' ]
for file in files:
lfn = file[ 'name' ]
events = self.getEvents( lfn )
(runs,lumis) = self.getRunLumi( lfn )
fileToAdd = File( lfn, file[ 'bytes'], events, runs[0], lumis[0] )
replicas = file[ 'replica' ]
if len( replicas ) > 0:
locations = []
for replica in replicas:
locations.append( replica[ 'node' ] )
fileToAdd.setLocation( locations )
fileset.addFile( fileToAdd )
def doBlock(self, entity, fileset):
connection = urlopen(self.nodeURL + "&block=%s" % quote(entity))
aString = connection.read()
connection.close()
if aString[2:8] != "phedex":
print "PhEDExNotifier: bad string from server follows."
print "%s" % aString
phedex = eval(aString.replace("null", "None"), {}, {})
blocks = phedex["phedex"]["block"]
if len(blocks) != 1:
print "PhEDExNotifier: Found %d blocks, expected 1, will only consider first block" % len(blocks)
files = blocks[0]["file"]
for file in files:
lfn = file["name"]
events = self.getEvents(lfn)
(runs, lumis) = self.getRunLumi(lfn)
fileToAdd = File(lfn, file["bytes"], events, runs[0], lumis[0])
replicas = file["replica"]
if len(replicas) > 0:
locations = []
for replica in replicas:
locations.append(replica["node"])
fileToAdd.setLocation(locations)
fileset.addFile(fileToAdd)
def execute(self, *args, **kwargs): #pylint: disable=unused-argument
# since https://github.com/dmwm/CRABServer/issues/5633 totalunits can be a float
# but that would confuse WMCore, therefore cast to int
totalevents = int(kwargs['task']['tm_totalunits'])
firstEvent = 1
lastEvent = totalevents
firstLumi = 1
lastLumi = 10
# Set a default of 100 events per lumi. This is set as a task
# property, as the splitting considers it independently of the file
# information provided by the fake dataset.
if not kwargs['task']['tm_events_per_lumi']:
kwargs['task']['tm_events_per_lumi'] = 100
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name="MCFakeFileSet")
newFile = File("MCFakeFile", size=1000, events=totalevents)
newFile.setLocation(self.getListOfSites())
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["block"] = 'MCFakeBlock'
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
singleMCFileset.addFile(newFile)
return Result(task=kwargs['task'], result=singleMCFileset)
def execute(self, *args, **kwargs):
totalevents = kwargs['task']['tm_totalunits']
firstEvent = 1
lastEvent = totalevents
firstLumi = 1
lastLumi = 10
# Set a default of 100 events per lumi. This is set as a task
# property, as the splitting considers it independently of the file
# information provided by the fake dataset.
if not kwargs['task']['tm_events_per_lumi']:
kwargs['task']['tm_events_per_lumi'] = 100
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCFakeFileSet")
newFile = File("MCFakeFile", size = 1000, events = totalevents)
sbj = SiteDBJSON({"key":self.config.TaskWorker.cmskey,
"cert":self.config.TaskWorker.cmscert})
newFile.setLocation(sbj.getAllCMSNames())
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["block"] = 'MCFakeBlock'
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
singleMCFileset.addFile(newFile)
return Result(task=kwargs['task'], result=singleMCFileset)
def execute(self, *args, **kwargs): #pylint: disable=unused-argument
# since https://github.com/dmwm/CRABServer/issues/5633 totalunits can be a float
# but that would confuse WMCore, therefore cast to int
totalevents = int(kwargs['task']['tm_totalunits'])
firstEvent = 1
lastEvent = totalevents
firstLumi = 1
lastLumi = 10
# Set a default of 100 events per lumi. This is set as a task
# property, as the splitting considers it independently of the file
# information provided by the fake dataset.
if not kwargs['task']['tm_events_per_lumi']:
kwargs['task']['tm_events_per_lumi'] = 100
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCFakeFileSet")
newFile = File("MCFakeFile", size = 1000, events = totalevents)
newFile.setLocation(self.getListOfSites())
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["block"] = 'MCFakeBlock'
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
singleMCFileset.addFile(newFile)
return Result(task=kwargs['task'], result=singleMCFileset)
def execute(self, *args, **kwargs):
totalevents = kwargs['task']['tm_totalunits']
firstEvent = 1
lastEvent = totalevents
firstLumi = 1
lastLumi = 10
# Set a default of 100 events per lumi. This is set as a task
# property, as the splitting considers it independently of the file
# information provided by the fake dataset.
if not kwargs['task']['tm_events_per_lumi']:
kwargs['task']['tm_events_per_lumi'] = 100
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCFakeFileSet")
newFile = File("MCFakeFile", size = 1000, events = totalevents)
if hasattr(self.config.Sites, 'available'):
newFile.setLocation(self.config.Sites.available)
else:
sbj = SiteDBJSON({"key":self.config.TaskWorker.cmskey,
"cert":self.config.TaskWorker.cmscert})
newFile.setLocation(sbj.getAllCMSNames())
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["block"] = 'MCFackBlock'
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
singleMCFileset.addFile(newFile)
return Result(task=kwargs['task'], result=singleMCFileset)
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name="TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size=1000, events=100)
newFile.setLocation("se01")
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name="TestFileset2")
newFile = File("/some/file/name", size=1000, events=100)
newFile.setLocation("se02")
self.singleFileFileset.addFile(newFile)
self.emptyFileFileset = Fileset(name="TestFileset3")
newFile = File("/some/file/name", size=1000, events=0)
newFile.setdefault("se03")
self.emptyFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset, workflow=testWorkflow, split_algo="EventBased", type="Processing"
)
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset, workflow=testWorkflow, split_algo="EventBased", type="Processing"
)
self.emptyFileSubscription = Subscription(
fileset=self.emptyFileFileset, workflow=testWorkflow, split_algo="EventBased", type="Processing"
)
return
def generateFakeMCFile(self,
numEvents=100,
firstEvent=1,
lastEvent=100,
firstLumi=1,
lastLumi=10,
existingSub=None):
# MC comes with only one MCFakeFile
newFile = File("MCFakeFileTest", size=1000, events=numEvents)
newFile.setLocation('se01')
if firstLumi == lastLumi:
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
else:
newFile.addRun(Run(1, *range(firstLumi, lastLumi)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
if existingSub is None:
singleMCFileset = Fileset(name="MCTestFileset")
singleMCFileset.addFile(newFile)
testWorkflow = Workflow()
existingSub = Subscription(fileset=singleMCFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Production")
else:
existingSub['fileset'].addFile(newFile)
return existingSub
def testAddFile(self):
"""
Testcase for the addFile method of the Fileset class
"""
#First test - Add file and check if its there
testfile = File('/tmp/lfntest',9999,9,9)
self.fileset.addFile(testfile)
assert(testfile in self.fileset.listNewFiles(), 'Couldn\'t add file ' +
'to fileset - fileset.addfile method not working')
#Second test - Add file that was already at Fileset.files ,
# and check if it gets updated
testFileSame = File('/tmp/lfntest',9999,9,9)
testFileSame.setLocation(set('dummyse.dummy.com'))
self.fileset.addFile(testFileSame)
assert(testFileSame in self.fileset.getFiles(),'Same file copy ' +
'failed - fileset.addFile not updating location of already ' +
'existing files' )
assert(testfile in self.fileset.getFiles(),'Same file copy ' +
'failed - fileset.addFile unable to remove previous file ' +
'from list')
#Third test - Add file that was already at Fileset.newfiles ,
#and check if it gets updated
assert(testFileSame in self.fileset.listNewFiles(),'Same file copy ' +
'failed - fileset.addFile not adding file to fileset.newFiles')
def returnDataStructsFile(self):
"""
_returnDataStructsFile_
Creates a dataStruct file out of this file
"""
parents = set()
for parent in self["parents"]:
parents.add(
WMFile(lfn=parent['lfn'],
size=parent['size'],
events=parent['events'],
checksums=parent['checksums'],
parents=parent['parents'],
merged=parent['merged']))
file = WMFile(lfn=self['lfn'],
size=self['size'],
events=self['events'],
checksums=self['checksums'],
parents=parents,
merged=self['merged'])
for run in self['runs']:
file.addRun(run)
for location in self['locations']:
file.setLocation(pnn=location)
return file
def testAddFile(self):
"""
_testAddFile_
Testcase for the addFile method of the Fileset class
"""
# First test - Add file and check if it's there
testfile = File('/tmp/lfntest', 9999, 9, 9)
self.fileset.addFile(testfile)
self.assertTrue(
testfile in self.fileset.listNewFiles(),
"Couldn't add file to fileset - fileset.addfile method not working"
)
# Second test - Add file that was already at Fileset.files, and check if it gets updated
testFileSame = File('/tmp/lfntest', 9999, 9, 9)
testFileSame.setLocation(set('dummyse.dummy.com'))
self.fileset.addFile(testFileSame)
self.assertTrue(
testFileSame in self.fileset.getFiles(),
'Same file copy ailed - fileset.addFile not updating location of already existing files'
)
self.assertTrue(
testfile in self.fileset.getFiles(),
'Same file copy failed - fileset.addFile unable to remove previous file from list'
)
# Third test - Add file that was already at Fileset.newfiles, and check if it gets updated
self.assertTrue(
testFileSame in self.fileset.listNewFiles(),
'Same file copy failed - fileset.addFile not adding file to fileset.newFiles'
)
def __call__(self, fileset):
"""
return a randomly sized list of files (DataStructs.File) at locations
files will always be new
"""
num_files = random.randint(0 , self.max)
for f in self.makelist(fileset):
list = []
for i in range(0, num_files):
# Decide where the file is
locs = []
for i in range(0, len(self.locations)):
if random.randint(0 , 1):
locs.append(self.locations[i])
lfn='/store/data/fake-feeder-files/notreal/%s.root' % uuid(i)
size=2000 + ((i-5) * 50)
events=1000 + ((i-3) * 150)
run = random.randint(0 , int(3.14159265 * i * self.max))
lumi = random.randint(0 ,10)
file = File(lfn, size, events, run, lumi)
file.setLocation(locs)
f.addFile(file)
return fileset
def jobConfig(self, wf, task, jobid, lfn):
"""
Create a fake job dict to upload to the ACDC server
"""
testFile = File(lfn=lfn, size=1024, events=1024)
testFile.setLocation(["T2_CH_CERN", "T2_CH_CERN_HLT"])
testFile.addRun(Run(jobid, 1, 2)) # run = jobid
testJob = self.getMinimalJob(wf, task)
testJob.addFile(testFile)
return testJob
def jobConfig(self, wf, task, jobid, lfn):
"""
Create a fake job dict to upload to the ACDC server
"""
testFile = File(lfn=lfn, size=1024, events=1024)
testFile.setLocation(["T2_CH_CERN", "T2_CH_CERN_HLT"])
testFile.addRun(Run(jobid, 1, 2)) # run = jobid
testJob = self.getMinimalJob(wf, task)
testJob.addFile(testFile)
return testJob
def execute(self, *args, **kwargs):
self.logger.info(
"Data discovery and splitting for %s using user-provided files" %
kwargs['task']['tm_taskname'])
userfiles = kwargs['task']['tm_arguments'].get('userfiles')
splitting = kwargs['task']['tm_split_algo']
total_units = kwargs['task']['tm_totalunits']
if not userfiles or splitting != 'FileBased':
if not userfiles:
msg = "No files specified to process for task %s." % kwargs[
'task']['tm_taskname']
if splitting != 'FileBased':
msg = "Data.splitting must be set to 'FileBased' when using a custom set of files."
self.logger.error("Setting %s as failed: %s" %
(kwargs['task']['tm_taskname'], msg))
configreq = {
'workflow': kwargs['task']['tm_taskname'],
'status': "FAILED",
'subresource': 'failure',
'failure': b64encode(msg)
}
self.server.post(self.resturi, data=urllib.urlencode(configreq))
raise StopHandler(msg)
if hasattr(self.config.Sites, 'available'):
locations = self.config.Sites.available
else:
sbj = SiteDBJSON({
"key": self.config.TaskWorker.cmskey,
"cert": self.config.TaskWorker.cmscert
})
locations = sbj.getAllCMSNames()
userFileset = Fileset(name=kwargs['task']['tm_taskname'])
self.logger.info("There are %d files specified by the user." %
len(userfiles))
if total_units > 0:
self.logger.info("Will run over the first %d files." % total_units)
file_counter = 0
for userfile, idx in zip(userfiles, range(len(userfiles))):
newFile = File(userfile, size=1000, events=1)
newFile.setLocation(locations)
newFile.addRun(Run(1, idx))
newFile["block"] = 'UserFilesFakeBlock'
newFile["first_event"] = 1
newFile["last_event"] = 2
userFileset.addFile(newFile)
file_counter += 1
if total_units > 0 and file_counter >= total_units:
break
return Result(task=kwargs['task'], result=userFileset)
def getOutputFile(self, fileName, outputModule, step):
"""
_getOutputFile_
Takes a fileRef object and returns a DataStructs/File object as output
"""
outputMod = self.getOutputModule(step=step, outputModule=outputModule)
if not outputMod:
return None
fileRef = getattr(outputMod.files, fileName, None)
newFile = File(locations=set())
# Locations
newFile.setLocation(getattr(fileRef, "location", None))
# Runs
runList = fileRef.runs.listSections_()
for run in runList:
lumis = getattr(fileRef.runs, run)
if isinstance(lumis, dict):
newRun = Run(int(run), *lumis.items())
else:
newRun = Run(int(run), *lumis)
newFile.addRun(newRun)
newFile["lfn"] = getattr(fileRef, "lfn", None)
newFile["pfn"] = getattr(fileRef, "pfn", None)
newFile["events"] = int(getattr(fileRef, "events", 0))
newFile["size"] = int(getattr(fileRef, "size", 0))
newFile["branches"] = getattr(fileRef, "branches", [])
newFile["input"] = getattr(fileRef, "input", [])
newFile["inputpfns"] = getattr(fileRef, "inputpfns", [])
newFile["branch_hash"] = getattr(fileRef, "branch_hash", None)
newFile["catalog"] = getattr(fileRef, "catalog", "")
newFile["guid"] = getattr(fileRef, "guid", "")
newFile["module_label"] = getattr(fileRef, "module_label", "")
newFile["checksums"] = getattr(fileRef, "checksums", {})
newFile["merged"] = bool(getattr(fileRef, "merged", False))
newFile["dataset"] = getattr(fileRef, "dataset", {})
newFile["acquisitionEra"] = getattr(fileRef, 'acquisitionEra', None)
newFile["processingVer"] = getattr(fileRef, 'processingVer', None)
newFile["validStatus"] = getattr(fileRef, 'validStatus', None)
newFile["globalTag"] = getattr(fileRef, 'globalTag', None)
newFile["prep_id"] = getattr(fileRef, 'prep_id', None)
newFile['configURL'] = getattr(fileRef, 'configURL', None)
newFile['inputPath'] = getattr(fileRef, 'inputPath', None)
newFile["outputModule"] = outputModule
newFile["fileRef"] = fileRef
return newFile
def getOutputFile(self, fileName, outputModule, step):
"""
_getOutputFile_
Takes a fileRef object and returns a DataStructs/File object as output
"""
outputMod = self.getOutputModule(step=step, outputModule=outputModule)
if not outputMod:
return None
fileRef = getattr(outputMod.files, fileName, None)
newFile = File(locations=set())
# Locations
newFile.setLocation(getattr(fileRef, "location", None))
# Runs
runList = fileRef.runs.listSections_()
for run in runList:
lumis = getattr(fileRef.runs, run)
if isinstance(lumis, dict):
newRun = Run(int(run), *listitems(lumis))
else:
newRun = Run(int(run), *lumis)
newFile.addRun(newRun)
newFile["lfn"] = getattr(fileRef, "lfn", None)
newFile["pfn"] = getattr(fileRef, "pfn", None)
newFile["events"] = int(getattr(fileRef, "events", 0))
newFile["size"] = int(getattr(fileRef, "size", 0))
newFile["branches"] = getattr(fileRef, "branches", [])
newFile["input"] = getattr(fileRef, "input", [])
newFile["inputpfns"] = getattr(fileRef, "inputpfns", [])
newFile["branch_hash"] = getattr(fileRef, "branch_hash", None)
newFile["catalog"] = getattr(fileRef, "catalog", "")
newFile["guid"] = getattr(fileRef, "guid", "")
newFile["module_label"] = getattr(fileRef, "module_label", "")
newFile["checksums"] = getattr(fileRef, "checksums", {})
newFile["merged"] = bool(getattr(fileRef, "merged", False))
newFile["dataset"] = getattr(fileRef, "dataset", {})
newFile["acquisitionEra"] = getattr(fileRef, 'acquisitionEra', None)
newFile["processingVer"] = getattr(fileRef, 'processingVer', None)
newFile["validStatus"] = getattr(fileRef, 'validStatus', None)
newFile["globalTag"] = getattr(fileRef, 'globalTag', None)
newFile["prep_id"] = getattr(fileRef, 'prep_id', None)
newFile['configURL'] = getattr(fileRef, 'configURL', None)
newFile['inputPath'] = getattr(fileRef, 'inputPath', None)
newFile["outputModule"] = outputModule
newFile["fileRef"] = fileRef
return newFile
def createResubmitSpec(self, serverUrl, couchDB):
"""
_createResubmitSpec_
Create a bogus resubmit workload.
"""
self.site = "cmssrm.fnal.gov"
workload = WMWorkloadHelper(WMWorkload("TestWorkload"))
reco = workload.newTask("reco")
workload.setOwnerDetails(name = "evansde77", group = "DMWM")
# first task uses the input dataset
reco.addInputDataset(primary = "PRIMARY", processed = "processed-v1", tier = "TIER1")
reco.data.input.splitting.algorithm = "File"
reco.setTaskType("Processing")
cmsRunReco = reco.makeStep("cmsRun1")
cmsRunReco.setStepType("CMSSW")
reco.applyTemplates()
cmsRunRecoHelper = cmsRunReco.getTypeHelper()
cmsRunRecoHelper.addOutputModule("outputRECO",
primaryDataset = "PRIMARY",
processedDataset = "processed-v2",
dataTier = "TIER2",
lfnBase = "/store/dunkindonuts",
mergedLFNBase = "/store/kfc")
dcs = DataCollectionService(url = serverUrl, database = couchDB)
def getJob(workload):
job = Job()
job["task"] = workload.getTask("reco").getPathName()
job["workflow"] = workload.name()
job["location"] = self.site
job["owner"] = "evansde77"
job["group"] = "DMWM"
return job
testFileA = WMFile(lfn = makeUUID(), size = 1024, events = 1024)
testFileA.setLocation([self.site])
testFileA.addRun(Run(1, 1, 2))
testFileB = WMFile(lfn = makeUUID(), size = 1024, events = 1024)
testFileB.setLocation([self.site])
testFileB.addRun(Run(1, 3, 4))
testJobA = getJob(workload)
testJobA.addFile(testFileA)
testJobA.addFile(testFileB)
dcs.failedJobs([testJobA])
topLevelTask = workload.getTopLevelTask()[0]
workload.truncate("Resubmit_TestWorkload", topLevelTask.getPathName(),
serverUrl, couchDB)
return workload
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name="TestFileset1")
for i in range(10):
newFile = File(makeUUID(),
size=1000,
events=100,
locations=set(["somese.cern.ch"]))
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name="TestFileset2")
newFile = File("/some/file/name",
size=1000,
events=100,
locations=set(["somese.cern.ch"]))
self.singleFileFileset.addFile(newFile)
self.multipleSiteFileset = Fileset(name="TestFileset3")
for i in range(5):
newFile = File(makeUUID(),
size=1000,
events=100,
locations=set(["somese.cern.ch"]))
newFile.setLocation("somese.cern.ch")
self.multipleSiteFileset.addFile(newFile)
for i in range(5):
newFile = File(makeUUID(), size=1000, events=100)
newFile.setLocation(["somese.cern.ch", "otherse.cern.ch"])
self.multipleSiteFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset,
workflow=testWorkflow,
split_algo="SizeBased",
type="Processing")
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset,
workflow=testWorkflow,
split_algo="SizeBased",
type="Processing")
self.multipleSiteSubscription = Subscription(
fileset=self.multipleSiteFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Processing")
return
def createFile(self, lfn, events, run, lumis, location):
"""
_createFile_
Create a file for testing
"""
newFile = File(lfn=lfn, size=1000, events=events)
lumiList = []
for lumi in range(lumis):
lumiList.append((run * lumis) + lumi)
newFile.addRun(Run(run, *lumiList))
newFile.setLocation(location)
return newFile
def createFile(lfn, events, run, lumis, location):
"""
_createFile_
Create a file for testing
"""
newFile = File(lfn=lfn, size=1000, events=events)
lumiList = []
for lumi in range(lumis):
lumiList.append((run * lumis) + lumi)
newFile.addRun(Run(run, *lumiList))
newFile.setLocation(location)
return newFile
def generateFakeMCFile(self, numEvents=100, firstEvent=1, lastEvent=100, firstLumi=1, lastLumi=10):
# MC comes with only one MCFakeFile
singleMCFileset = Fileset(name="MCTestFileset")
newFile = File("MCFakeFileTest", size=1000, events=numEvents)
newFile.setLocation("se01")
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
testWorkflow = Workflow()
singleMCFileset.addFile(newFile)
singleMCFileSubscription = Subscription(
fileset=singleMCFileset, workflow=testWorkflow, split_algo="EventBased", type="Production"
)
return singleMCFileSubscription
def generateFakeMCFile(self, numEvents = 100, firstEvent = 1,
lastEvent = 100, firstLumi = 1, lastLumi = 10):
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCTestFileset")
newFile = File("MCFakeFileTest", size = 1000, events = numEvents)
newFile.setLocation('se01')
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
testWorkflow = Workflow()
singleMCFileset.addFile(newFile)
singleMCFileSubscription = Subscription(fileset = singleMCFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Production")
return singleMCFileSubscription
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name="TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size=1000, events=100)
newFile.setLocation('se01')
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name="TestFileset2")
newFile = File("/some/file/name", size=1000, events=100)
newFile.setLocation('se02')
self.singleFileFileset.addFile(newFile)
self.emptyFileFileset = Fileset(name="TestFileset3")
newFile = File("/some/file/name", size=1000, events=0)
newFile.setdefault('se03')
self.emptyFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Processing")
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Processing")
self.emptyFileSubscription = Subscription(
fileset=self.emptyFileFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Processing")
self.performanceParams = {
'timePerEvent': None,
'memoryRequirement': 2300,
'sizePerEvent': 400
}
return
def testDataStructsFile(self):
"""
_testDataStructsFile_
Tests our ability to create a WMBS file from a DataStructs File and vice versa
"""
myThread = threading.currentThread()
testLFN = "lfn1"
testSize = 1024
testEvents = 100
testCksum = {"cksum": '1'}
testParents = set(["lfn2"])
testRun = Run(1, *[45])
testSE = "se1.cern.ch"
parentFile = File(lfn="lfn2")
parentFile.create()
testFile = File()
inputFile = WMFile(lfn=testLFN,
size=testSize,
events=testEvents,
checksums=testCksum,
parents=testParents)
inputFile.addRun(testRun)
inputFile.setLocation(se=testSE)
testFile.loadFromDataStructsFile(file=inputFile)
testFile.create()
testFile.save()
loadFile = File(lfn="lfn1")
loadFile.loadData(parentage=1)
self.assertEqual(loadFile['size'], testSize)
self.assertEqual(loadFile['events'], testEvents)
self.assertEqual(loadFile['checksums'], testCksum)
self.assertEqual(loadFile['locations'], set([testSE]))
#self.assertEqual(loadFile['parents'].pop()['lfn'], 'lfn2')
wmFile = loadFile.returnDataStructsFile()
self.assertEqual(wmFile == inputFile, True)
return
def execute(self, *args, **kwargs):
self.logger.info("Data discovery and splitting for %s using user-provided files" % kwargs['task']['tm_taskname'])
if 'tm_user_files' in kwargs['task'] and kwargs['task']['tm_user_files']:
userfiles = kwargs['task']['tm_user_files']
else: ## For backward compatibility only.
userfiles = kwargs['task']['tm_arguments'].get('userfiles')
splitting = kwargs['task']['tm_split_algo']
total_units = kwargs['task']['tm_totalunits']
if not userfiles or splitting != 'FileBased':
if not userfiles:
msg = "No files specified to process for task %s." % kwargs['task']['tm_taskname']
if splitting != 'FileBased':
msg = "Data.splitting must be set to 'FileBased' when using a custom set of files."
self.logger.error("Setting %s as failed: %s" % (kwargs['task']['tm_taskname'], msg))
configreq = {'workflow': kwargs['task']['tm_taskname'],
'status': "FAILED",
'subresource': 'failure',
'failure': b64encode(msg)}
self.server.post(self.resturi, data = urllib.urlencode(configreq))
raise StopHandler(msg)
if hasattr(self.config.Sites, 'available'):
locations = self.config.Sites.available
else:
sbj = SiteDBJSON({"key":self.config.TaskWorker.cmskey,
"cert":self.config.TaskWorker.cmscert})
locations = sbj.getAllCMSNames()
userFileset = Fileset(name = kwargs['task']['tm_taskname'])
self.logger.info("There are %d files specified by the user." % len(userfiles))
if total_units > 0:
self.logger.info("Will run over the first %d files." % total_units)
file_counter = 0
for userfile, idx in zip(userfiles, range(len(userfiles))):
newFile = File(userfile, size = 1000, events = 1)
newFile.setLocation(locations)
newFile.addRun(Run(1, idx))
newFile["block"] = 'UserFilesFakeBlock'
newFile["first_event"] = 1
newFile["last_event"] = 2
userFileset.addFile(newFile)
file_counter += 1
if total_units > 0 and file_counter >= total_units:
break
return Result(task = kwargs['task'], result = userFileset)
def execute(self, *args, **kwargs):
self.logger.info(
"Data discovery and splitting for %s using user-provided files" %
kwargs['task']['tm_taskname'])
userfiles = kwargs['task']['tm_user_files']
splitting = kwargs['task']['tm_split_algo']
total_units = kwargs['task']['tm_totalunits']
if not userfiles or splitting != 'FileBased':
if not userfiles:
msg = "No files specified to process for task %s." % kwargs[
'task']['tm_taskname']
if splitting != 'FileBased':
msg = "Data.splitting must be set to 'FileBased' when using a custom set of files."
raise TaskWorkerException(msg)
if hasattr(self.config.Sites, 'available'):
locations = self.config.Sites.available
else:
with self.config.TaskWorker.envForCMSWEB:
configDict = {
"cacheduration": 1,
"pycurl": True
} # cache duration is in hours
resourceCatalog = CRIC(logger=self.logger,
configDict=configDict)
locations = resourceCatalog.getAllPSNs()
userFileset = Fileset(name=kwargs['task']['tm_taskname'])
self.logger.info("There are %d files specified by the user." %
len(userfiles))
if total_units > 0:
self.logger.info("Will run over the first %d files." % total_units)
file_counter = 0
for userfile, idx in zip(userfiles, range(len(userfiles))):
newFile = File(userfile, size=1000, events=1)
newFile.setLocation(locations)
newFile.addRun(Run(1, idx))
newFile["block"] = 'UserFilesFakeBlock'
newFile["first_event"] = 1
newFile["last_event"] = 2
userFileset.addFile(newFile)
file_counter += 1
if total_units > 0 and file_counter >= total_units:
break
return Result(task=kwargs['task'], result=userFileset)
def getFileset(self):
"""
Get a fileset based on the task
"""
fileset = Fileset(name='Merge%s' % (type))
for i in range(0, random.randint(15, 25)):
# Use the testDir to generate a random lfn
inpFile = File(lfn="%s/%s.root" % (self.testDir, makeUUID()),
size=random.randint(200000, 1000000),
events=random.randint(1000, 2000))
inpFile.setLocation('Megiddo')
fileset.addFile(inpFile)
return fileset
def getFileset(self):
"""
Get a fileset based on the task
"""
fileset = Fileset(name='Merge%s' % (type))
for i in range(0, random.randint(15, 25)):
# Use the testDir to generate a random lfn
inpFile = File(lfn="%s/%s.root" % (self.testDir, makeUUID()),
size=random.randint(200000, 1000000),
events=random.randint(1000, 2000))
inpFile.setLocation('Megiddo')
fileset.addFile(inpFile)
return fileset
def testDataStructsFile(self):
"""
_testDataStructsFile_
Tests our ability to create a WMBS file from a DataStructs File and vice versa
"""
myThread = threading.currentThread()
testLFN = "lfn1"
testSize = 1024
testEvents = 100
testCksum = {"cksum": '1'}
testParents = set(["lfn2"])
testRun = Run( 1, *[45])
testSE = "se1.cern.ch"
parentFile = File(lfn= "lfn2")
parentFile.create()
testFile = File()
inputFile = WMFile(lfn = testLFN, size = testSize, events = testEvents, checksums = testCksum, parents = testParents)
inputFile.addRun(testRun)
inputFile.setLocation(se = testSE)
testFile.loadFromDataStructsFile(file = inputFile)
testFile.create()
testFile.save()
loadFile = File(lfn = "lfn1")
loadFile.loadData(parentage = 1)
self.assertEqual(loadFile['size'], testSize)
self.assertEqual(loadFile['events'], testEvents)
self.assertEqual(loadFile['checksums'], testCksum)
self.assertEqual(loadFile['locations'], set([testSE]))
#self.assertEqual(loadFile['parents'].pop()['lfn'], 'lfn2')
wmFile = loadFile.returnDataStructsFile()
self.assertEqual(wmFile == inputFile, True)
return
def addOutputFilesToReport(self, report):
"""
_addOutputFilesToReport_
Add output files to every output module in the step. Scale the size
and number of events in the output files appropriately.
"""
(outputSize, outputEvents) = self.determineOutputSize()
if not os.path.exists('ReportEmuTestFile.txt'):
f = open('ReportEmuTestFile.txt', 'w')
f.write('A Shubbery')
f.close()
for outputModuleName in self.step.listOutputModules():
outputModuleSection = self.step.getOutputModule(outputModuleName)
outputModuleSection.fixedLFN = False
outputModuleSection.disableGUID = False
outputLFN = "%s/%s.root" % (outputModuleSection.lfnBase,
str(makeUUID()))
outputFile = File(lfn=outputLFN,
size=outputSize,
events=outputEvents,
merged=False)
outputFile.setLocation(self.job["location"])
outputFile['pfn'] = "ReportEmuTestFile.txt"
outputFile['guid'] = "ThisIsGUID"
outputFile["checksums"] = {"adler32": "1234", "cksum": "5678"}
outputFile["dataset"] = {
"primaryDataset": outputModuleSection.primaryDataset,
"processedDataset": outputModuleSection.processedDataset,
"dataTier": outputModuleSection.dataTier,
"applicationName": "cmsRun",
"applicationVersion": self.step.getCMSSWVersion()
}
outputFile["module_label"] = outputModuleName
outputFileSection = report.addOutputFile(outputModuleName,
outputFile)
for inputFile in self.job["input_files"]:
Report.addRunInfoToFile(outputFileSection, inputFile["runs"])
return
def execute(self, *args, **kwargs):
totalevents = kwargs['task']['tm_totalunits']
firstEvent = 1
lastEvent = totalevents
firstLumi = 1
lastLumi = 10
#MC comes with only one MCFakeFile
singleMCFileset = Fileset(name = "MCFakeFileSet")
newFile = File("MCFakeFile", size = 1000, events = totalevents)
newFile.setLocation(self.config.Sites.available)
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
newFile["block"] = 'MCFackBlock'
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
singleMCFileset.addFile(newFile)
return Result(task=kwargs['task'], result=singleMCFileset)
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name = "TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.setLocation('se01')
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name = "TestFileset2")
newFile = File("/some/file/name", size = 1000, events = 100)
newFile.setLocation('se02')
self.singleFileFileset.addFile(newFile)
self.emptyFileFileset = Fileset(name = "TestFileset3")
newFile = File("/some/file/name", size = 1000, events = 0)
newFile.setdefault('se03')
self.emptyFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(fileset = self.multipleFileFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.singleFileSubscription = Subscription(fileset = self.singleFileFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.emptyFileSubscription = Subscription(fileset = self.emptyFileFileset,
workflow = testWorkflow,
split_algo = "EventBased",
type = "Processing")
self.performanceParams = {'timePerEvent' : None,
'memoryRequirement' : 2300,
'sizePerEvent' : 400}
return
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name = "TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.setLocation('blenheim')
newFile.setLocation('malpaquet')
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name = "TestFileset2")
newFile = File("/some/file/name", size = 1000, events = 100)
newFile.setLocation('blenheim')
self.singleFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(fileset = self.multipleFileFileset,
workflow = testWorkflow,
split_algo = "FileBased",
type = "Processing")
self.singleFileSubscription = Subscription(fileset = self.singleFileFileset,
workflow = testWorkflow,
split_algo = "FileBased",
type = "Processing")
#self.multipleFileSubscription.create()
#self.singleFileSubscription.create()
return
def testG_LumiMask(self):
"""
_testG_LumiMask_
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("somese.cern.ch")
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("somese.cern.ch")
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("somese.cern.ch")
testFileset = Fileset(name='Fileset')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiBased",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
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")
self.assertEqual(jobs[0]['mask'].getRunAndLumis(), {1: [[10, 14]], 2: [[20, 21]], 4: [[40, 40]]})
self.assertEqual(jobs[1]['mask'].getRunAndLumis(), {4: [[41, 41]]})
def testG_LumiMask(self):
"""
_testG_LumiMask_
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("somese.cern.ch")
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("somese.cern.ch")
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("somese.cern.ch")
testFileset = Fileset(name = 'Fileset')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testSubscription = Subscription(fileset = testFileset,
workflow = self.testWorkflow,
split_algo = "EventAwareLumiBased",
type = "Processing")
jobFactory = splitter(package = "WMCore.DataStructs",
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")
self.assertEqual(jobs[0]['mask'].getRunAndLumis(), {1: [[10, 14]], 2: [[20, 21]], 4: [[40, 40]]})
self.assertEqual(jobs[1]['mask'].getRunAndLumis(), {4: [[41, 41]]})
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name="TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size=1000, events=100)
newFile.setLocation('blenheim')
newFile.setLocation('malpaquet')
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name="TestFileset2")
newFile = File("/some/file/name", size=1000, events=100)
newFile.setLocation('blenheim')
self.singleFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset,
workflow=testWorkflow,
split_algo="FileBased",
type="Processing")
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset,
workflow=testWorkflow,
split_algo="FileBased",
type="Processing")
#self.multipleFileSubscription.create()
#self.singleFileSubscription.create()
return
def addOutputFilesToReport(self, report):
"""
_addOutputFilesToReport_
Add output files to every output module in the step. Scale the size
and number of events in the output files appropriately.
"""
(outputSize, outputEvents) = self.determineOutputSize()
if not os.path.exists('ReportEmuTestFile.txt'):
f = open('ReportEmuTestFile.txt', 'w')
f.write('A Shubbery')
f.close()
for outputModuleName in self.step.listOutputModules():
outputModuleSection = self.step.getOutputModule(outputModuleName)
outputModuleSection.fixedLFN = False
outputModuleSection.disableGUID = False
outputLFN = "%s/%s.root" % (outputModuleSection.lfnBase,
str(makeUUID()))
outputFile = File(lfn = outputLFN, size = outputSize, events = outputEvents,
merged = False)
outputFile.setLocation(self.job["location"])
outputFile['pfn'] = "ReportEmuTestFile.txt"
outputFile['guid'] = "ThisIsGUID"
outputFile["checksums"] = {"adler32": "1234", "cksum": "5678"}
outputFile["dataset"] = {"primaryDataset": outputModuleSection.primaryDataset,
"processedDataset": outputModuleSection.processedDataset,
"dataTier": outputModuleSection.dataTier,
"applicationName": "cmsRun",
"applicationVersion": self.step.getCMSSWVersion()}
outputFile["module_label"] = outputModuleName
outputFileSection = report.addOutputFile(outputModuleName, outputFile)
for inputFile in self.job["input_files"]:
Report.addRunInfoToFile(outputFileSection, inputFile["runs"])
return
def execute(self, *args, **kwargs):
self.logger.info("Data discovery and splitting for %s using user-provided files" % kwargs['task']['tm_taskname'])
userfiles = kwargs['task']['tm_user_files']
splitting = kwargs['task']['tm_split_algo']
total_units = kwargs['task']['tm_totalunits']
if not userfiles or splitting != 'FileBased':
if not userfiles:
msg = "No files specified to process for task %s." % kwargs['task']['tm_taskname']
if splitting != 'FileBased':
msg = "Data.splitting must be set to 'FileBased' when using a custom set of files."
raise TaskWorkerException(msg)
if hasattr(self.config.Sites, 'available'):
locations = self.config.Sites.available
else:
with self.config.TaskWorker.envForCMSWEB :
configDict = {"cacheduration": 1, "pycurl": True} # cache duration is in hours
resourceCatalog = CRIC(logger=self.logger, configDict=configDict)
locations = resourceCatalog.getAllPSNs()
userFileset = Fileset(name = kwargs['task']['tm_taskname'])
self.logger.info("There are %d files specified by the user." % len(userfiles))
if total_units > 0:
self.logger.info("Will run over the first %d files." % total_units)
file_counter = 0
for userfile, idx in zip(userfiles, range(len(userfiles))):
newFile = File(userfile, size = 1000, events = 1)
newFile.setLocation(locations)
newFile.addRun(Run(1, idx))
newFile["block"] = 'UserFilesFakeBlock'
newFile["first_event"] = 1
newFile["last_event"] = 2
userFileset.addFile(newFile)
file_counter += 1
if total_units > 0 and file_counter >= total_units:
break
return Result(task = kwargs['task'], result = userFileset)
def returnDataStructsFile(self):
"""
_returnDataStructsFile_
Creates a dataStruct file out of this file
"""
parents = set()
for parent in self["parents"]:
parents.add(WMFile(lfn = parent['lfn'], size = parent['size'],
events = parent['events'], checksums = parent['checksums'],
parents = parent['parents'], merged = parent['merged']))
file = WMFile(lfn = self['lfn'], size = self['size'],
events = self['events'], checksums = self['checksums'],
parents = parents, merged = self['merged'])
for run in self['runs']:
file.addRun(run)
for location in self['locations']:
file.setLocation(se = location)
return file
def createSubscription(self, nFiles, lumisPerFile, twoSites = False):
"""
_createSubscription_
Create a subscription for testing
"""
baseName = makeUUID()
testFileset = Fileset(name = baseName)
for i in range(nFiles):
newFile = File(lfn = '%s_%i' % (baseName, i), size = 1000,
events = 100)
lumis = []
for lumi in range(lumisPerFile):
lumis.append((i * 100) + lumi)
newFile.addRun(Run(i, *lumis))
newFile.setLocation('blenheim')
testFileset.addFile(newFile)
if twoSites:
for i in range(nFiles):
newFile = File(lfn = '%s_%i_2' % (baseName, i), size = 1000,
events = 100)
lumis = []
for lumi in range(lumisPerFile):
lumis.append(5 + 10 * (i * 100) + lumi) #lumis should be different
newFile.addRun(Run(i, *lumis))
newFile.setLocation('malpaquet')
testFileset.addFile(newFile)
testSubscription = Subscription(fileset = testFileset,
workflow = self.testWorkflow,
split_algo = "LumiBased",
type = "Processing")
return testSubscription
def createSubscription(self, nFiles, lumisPerFile, twoSites=False):
"""
_createSubscription_
Create a subscription for testing
"""
baseName = makeUUID()
testFileset = Fileset(name=baseName)
for i in range(nFiles):
newFile = File(lfn='%s_%i' % (baseName, i), size=1000, events=100)
lumis = []
for lumi in range(lumisPerFile):
lumis.append((i * 100) + lumi)
newFile.addRun(Run(i, *lumis))
newFile.setLocation('blenheim')
testFileset.addFile(newFile)
if twoSites:
for i in range(nFiles):
newFile = File(lfn='%s_%i_2' % (baseName, i),
size=1000,
events=100)
lumis = []
for lumi in range(lumisPerFile):
lumis.append(5 + 10 * (i * 100) +
lumi) #lumis should be different
newFile.addRun(Run(i, *lumis))
newFile.setLocation('malpaquet')
testFileset.addFile(newFile)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="LumiBased",
type="Processing")
return testSubscription
def generateFakeMCFile(self, numEvents=100, firstEvent=1, lastEvent=100,
firstLumi=1, lastLumi=10, existingSub=None):
# MC comes with only one MCFakeFile
newFile = File("MCFakeFileTest", size=1000, events=numEvents)
newFile.setLocation('se01')
if firstLumi == lastLumi:
newFile.addRun(Run(1, *range(firstLumi, lastLumi + 1)))
else:
newFile.addRun(Run(1, *range(firstLumi, lastLumi)))
newFile["first_event"] = firstEvent
newFile["last_event"] = lastEvent
if existingSub is None:
singleMCFileset = Fileset(name="MCTestFileset")
singleMCFileset.addFile(newFile)
testWorkflow = Workflow()
existingSub = Subscription(fileset=singleMCFileset,
workflow=testWorkflow,
split_algo="EventBased",
type="Production")
else:
existingSub['fileset'].addFile(newFile)
return existingSub
def __call__(self, fileset):
"""
return a randomly sized list of files (DataStructs.File) at locations
files will always be new
"""
num_files = random.randint(0, self.max)
for f in self.makelist(fileset):
list = []
for i in range(0, num_files):
# Decide where the file is
locs = []
for i in range(0, len(self.locations)):
if random.randint(0, 1):
locs.append(self.locations[i])
lfn = '/store/data/fake-feeder-files/notreal/%s.root' % uuid(i)
size = 2000 + ((i - 5) * 50)
events = 1000 + ((i - 3) * 150)
run = random.randint(0, int(3.14159265 * i * self.max))
lumi = random.randint(0, 10)
file = File(lfn, size, events, run, lumi)
file.setLocation(locs)
f.addFile(file)
return fileset
def testSetLocation(self):
"""
Test the `setLocation` method functionality
"""
testFile = File(lfn="test_file")
self.assertItemsEqual(testFile['locations'], {})
testFile.setLocation(None)
self.assertItemsEqual(testFile['locations'], {})
testFile.setLocation("")
self.assertItemsEqual(testFile['locations'], {})
testFile.setLocation([])
self.assertItemsEqual(testFile['locations'], {})
testFile.setLocation("valid_PNN")
self.assertItemsEqual(testFile['locations'], {"valid_PNN"})
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name="TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size=1000, events=100)
newFile.setLocation('blenheim')
newFile.setLocation('malpaquet')
lumis = []
for lumi in range(20):
lumis.append((i * 100) + lumi)
newFile.addRun(Run(i, *lumis))
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name="TestFileset2")
newFile = File("/some/file/name", size=1000, events=100)
newFile.setLocation('blenheim')
lumis = list(range(50, 60)) + list(range(70, 80))
newFile.addRun(Run(13, *lumis))
self.singleFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(
fileset=self.multipleFileFileset,
workflow=testWorkflow,
split_algo="FileBased",
type="Processing")
self.singleFileSubscription = Subscription(
fileset=self.singleFileFileset,
workflow=testWorkflow,
split_algo="FileBased",
type="Processing")
#self.multipleFileSubscription.create()
#self.singleFileSubscription.create()
self.performanceParams = {
'timePerEvent': 12,
'memoryRequirement': 2300,
'sizePerEvent': 400
}
return
def setUp(self):
"""
_setUp_
Create two subscriptions: One that contains a single file and one that
contains multiple files.
"""
self.multipleFileFileset = Fileset(name = "TestFileset1")
for i in range(10):
newFile = File(makeUUID(), size = 1000, events = 100)
newFile.setLocation('blenheim')
newFile.setLocation('malpaquet')
lumis = []
for lumi in range(20):
lumis.append((i * 100) + lumi)
newFile.addRun(Run(i, *lumis))
self.multipleFileFileset.addFile(newFile)
self.singleFileFileset = Fileset(name = "TestFileset2")
newFile = File("/some/file/name", size = 1000, events = 100)
newFile.setLocation('blenheim')
lumis = range(50,60) + range(70,80)
newFile.addRun(Run(13, *lumis))
self.singleFileFileset.addFile(newFile)
testWorkflow = Workflow()
self.multipleFileSubscription = Subscription(fileset = self.multipleFileFileset,
workflow = testWorkflow,
split_algo = "FileBased",
type = "Processing")
self.singleFileSubscription = Subscription(fileset = self.singleFileFileset,
workflow = testWorkflow,
split_algo = "FileBased",
type = "Processing")
#self.multipleFileSubscription.create()
#self.singleFileSubscription.create()
self.performanceParams = {'timePerEvent' : 12,
'memoryRequirement' : 2300,
'sizePerEvent' : 400}
return
def testD_NoFileSplitNoHardLimit(self):
"""
_testD_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.DataStructs",
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)
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")
#Create 7 files, each one with different lumi/event distributions
testFileset = Fileset(name = "FilesetA")
testFileA = self.createFile("/this/is/file1", 250, 0, 5, "blenheim")
testFileB = self.createFile("/this/is/file2", 600, 1, 1, "blenheim")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2, "blenheim")
testFileD = self.createFile("/this/is/file4", 100, 3, 1, "blenheim")
testFileE = self.createFile("/this/is/file5", 30, 4, 1, "blenheim")
testFileF = self.createFile("/this/is/file6", 10, 5, 1, "blenheim")
testFileG = self.createFile("/this/is/file7", 151, 6, 3, "blenheim")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testSubscription = Subscription(fileset = testFileset,
workflow = self.testWorkflow,
split_algo = "EventAwareLumiBased",
type = "Processing")
jobFactory = splitter(package = "WMCore.DataStructs",
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)
self.assertEqual(len(jobGroups), 1, "There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 8, "Eight jobs must be in the jobgroup")
self.assertEqual(jobs[0]["mask"].getRunAndLumis(), {0L : [[0L, 2L]]}, "Wrong mask for the first job")
self.assertEqual(jobs[1]["mask"].getRunAndLumis(), {0L : [[3L, 4L]]}, "Wrong mask for the second job")
self.assertEqual(jobs[2]["mask"].getRunAndLumis(), {1L : [[1L, 1L]]}, "Wrong mask for the third job")
self.assertEqual(jobs[3]["mask"].getRunAndLumis(), {2L : [[4L, 4L]]}, "Wrong mask for the fourth job")
self.assertEqual(jobs[4]["mask"].getRunAndLumis(), {2L : [[5L, 5L]]}, "Wrong mask for the fifth job")
self.assertEqual(jobs[5]["mask"].getRunAndLumis(),
{3L : [[3L, 3L]], 4L : [[4L, 4L]], 5L : [[5L, 5L]]}, "Wrong mask for the sixth job")
self.assertEqual(jobs[6]["mask"].getRunAndLumis(), {6L : [[18L, 19L]]}, "Wrong mask for the seventh job")
self.assertEqual(jobs[7]["mask"].getRunAndLumis(), {6L : [[20L, 20L]]}, "Wrong mask for the seventh job")
#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/file1", 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("malpaquet")
fileB = self.createFile('/this/is/file2', 200, 3, 5, "malpaquet")
testFileset = Fileset(name = 'FilesetB')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testSubscription = Subscription(fileset = testFileset,
workflow = self.testWorkflow,
split_algo = "EventAwareLumiBased",
type = "Processing")
jobFactory = splitter(package = "WMCore.DataStructs",
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)
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")
def testChunking(self):
"""
_testChunking_
Insert a workload and files that have several distinct sets of
locations. Verify that the chunks are created correctly and that they
only groups files that have the same set of locations. Also verify that
the chunks are pulled out of ACDC correctly.
"""
dcs = DataCollectionService(url=self.testInit.couchUrl,
database="wmcore-acdc-datacollectionsvc")
testFileA = File(lfn=makeUUID(), size=1024, events=1024)
testFileA.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileA.addRun(Run(1, 1, 2))
testFileB = File(lfn=makeUUID(), size=1024, events=1024)
testFileB.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileB.addRun(Run(1, 3, 4))
testFileC = File(lfn=makeUUID(), size=1024, events=1024)
testFileC.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileC.addRun(Run(1, 5, 6))
testJobA = self.getMinimalJob()
testJobA.addFile(testFileA)
testJobA.addFile(testFileB)
testJobA.addFile(testFileC)
testFileD = File(lfn=makeUUID(), size=1024, events=1024)
testFileD.setLocation(["cmssrm.fnal.gov"])
testFileD.addRun(Run(2, 1, 2))
testFileE = File(lfn=makeUUID(), size=1024, events=1024)
testFileE.setLocation(["cmssrm.fnal.gov"])
testFileE.addRun(Run(2, 3, 4))
testJobB = self.getMinimalJob()
testJobB.addFile(testFileD)
testJobB.addFile(testFileE)
testFileF = File(lfn=makeUUID(),
size=1024,
events=1024,
parents={"/some/parent/F"})
testFileF.setLocation(
["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"])
testFileF.addRun(Run(3, 1, 2))
testFileG = File(lfn=makeUUID(),
size=1024,
events=1024,
parents={"/some/parent/G"})
testFileG.setLocation(
["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"])
testFileG.addRun(Run(3, 3, 4))
testFileH = File(lfn=makeUUID(),
size=1024,
events=1024,
parents={"/some/parent/H"})
testFileH.setLocation(
["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"])
testFileH.addRun(Run(3, 5, 6))
testJobC = self.getMinimalJob()
testJobC.addFile(testFileF)
testJobC.addFile(testFileG)
testJobC.addFile(testFileH)
testFileI = File(lfn=makeUUID(), size=1024, events=1024, merged=True)
testFileI.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileI.addRun(Run(4, 1, 2))
testFileJ = File(lfn=makeUUID(), size=1024, events=1024, merged=True)
testFileJ.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileJ.addRun(Run(4, 3, 4))
testFileK = File(lfn=makeUUID(), size=1024, events=1024, merged=True)
testFileK.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileK.addRun(Run(4, 5, 6))
testJobD = self.getMinimalJob()
testJobD.addFile(testFileI)
testJobD.addFile(testFileJ)
testJobD.addFile(testFileK)
dcs.failedJobs([testJobA, testJobB, testJobC, testJobD])
chunks = dcs.chunkFileset("ACDCTest", "/ACDCTest/reco", chunkSize=5)
self.assertEqual(
len(chunks), 4,
"Error: There should be four chunks: %s" % len(chunks))
goldenMetaData = {
1: {
"lumis": 2,
"locations": ["castor.cern.ch", "cmssrm.fnal.gov"],
"events": 1024
},
2: {
"lumis": 4,
"locations": ["cmssrm.fnal.gov"],
"events": 2048
},
3: {
"lumis": 6,
"locations":
["castor.cern.ch", "cmssrm.fnal.gov", "srm.ral.uk"],
"events": 3072
},
5: {
"lumis": 10,
"locations": ["castor.cern.ch", "cmssrm.fnal.gov"],
"events": 5120
}
}
testFiles = [
testFileA, testFileB, testFileC, testFileI, testFileJ, testFileK
]
lastFile = testFileA
for testFile in testFiles:
if lastFile["lfn"] < testFile["lfn"]:
lastFile = testFile
testFiles.remove(lastFile)
goldenFiles = {
1: [lastFile],
2: [testFileD, testFileE],
3: [testFileF, testFileG, testFileH],
5: testFiles
}
for chunk in chunks:
chunkMetaData = dcs.getChunkInfo("ACDCTest", "/ACDCTest/reco",
chunk["offset"], chunk["files"])
self.assertEqual(chunkMetaData["files"], chunk["files"])
self.assertEqual(chunkMetaData["lumis"], chunk["lumis"])
self.assertEqual(chunkMetaData["events"], chunk["events"])
self.assertEqual(chunkMetaData["locations"], chunk["locations"])
self.assertTrue(chunk["files"] in goldenMetaData.keys(),
"Error: Extra chunk found.")
self.assertEqual(chunk["lumis"],
goldenMetaData[chunk["files"]]["lumis"],
"Error: Lumis in chunk is wrong.")
self.assertEqual(chunk["locations"],
goldenMetaData[chunk["files"]]["locations"],
"Error: Locations in chunk is wrong.")
self.assertEqual(chunk["events"],
goldenMetaData[chunk["files"]]["events"],
"Error: Events in chunk is wrong.")
del goldenMetaData[chunk["files"]]
chunkFiles = dcs.getChunkFiles("ACDCTest", "/ACDCTest/reco",
chunk["offset"], chunk["files"])
self.assertTrue(chunk["files"] in goldenFiles.keys(),
"Error: Extra chunk found.")
goldenChunkFiles = goldenFiles[chunk["files"]]
self.assertEqual(len(chunkFiles), len(goldenChunkFiles))
for chunkFile in chunkFiles:
foundFile = None
for goldenChunkFile in goldenChunkFiles:
if chunkFile["lfn"] == goldenChunkFile["lfn"]:
foundFile = goldenChunkFile
break
self.assertIsNotNone(
foundFile, "Error: Missing chunk file: %s, %s" %
(chunkFiles, goldenChunkFiles))
self.assertEqual(set(foundFile["parents"]),
chunkFile["parents"],
"Error: File parents should match.")
self.assertEqual(foundFile["merged"], chunkFile["merged"],
"Error: File merged status should match.")
self.assertEqual(foundFile["locations"],
chunkFile["locations"],
"Error: File locations should match.")
self.assertEqual(foundFile["events"], chunkFile["events"])
self.assertEqual(foundFile["size"], chunkFile["size"])
self.assertEqual(len(foundFile["runs"]),
len(chunkFile["runs"]),
"Error: Wrong number of runs.")
for run in foundFile["runs"]:
runMatch = False
for chunkRun in chunkFile["runs"]:
if chunkRun.run == run.run and chunkRun.lumis == run.lumis:
runMatch = True
break
self.assertTrue(runMatch,
"Error: Run information is wrong.")
del goldenFiles[chunk["files"]]
singleChunk = dcs.singleChunkFileset("ACDCTest", "/ACDCTest/reco")
self.assertEqual(
singleChunk, {
"offset": 0,
"files": 11,
"events": 11264,
"lumis": 22,
"locations":
{"castor.cern.ch", "cmssrm.fnal.gov", "srm.ral.uk"}
}, "Error: Single chunk metadata is wrong")
return
def testD_NoFileSplitNoHardLimit(self):
"""
_testD_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.DataStructs",
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)
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")
#Create 7 files, each one with different lumi/event distributions
testFileset = Fileset(name="FilesetA")
testFileA = self.createFile("/this/is/file1", 250, 0, 5, "blenheim")
testFileB = self.createFile("/this/is/file2", 600, 1, 1, "blenheim")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2, "blenheim")
testFileD = self.createFile("/this/is/file4", 100, 3, 1, "blenheim")
testFileE = self.createFile("/this/is/file5", 30, 4, 1, "blenheim")
testFileF = self.createFile("/this/is/file6", 10, 5, 1, "blenheim")
testFileG = self.createFile("/this/is/file7", 151, 6, 3, "blenheim")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiBased",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
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)
self.assertEqual(len(jobGroups), 1,
"There should be only one job group")
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 8, "Eight jobs must be in the jobgroup")
self.assertEqual(jobs[0]["mask"].getRunAndLumis(), {0L: [[0L, 2L]]},
"Wrong mask for the first job")
self.assertEqual(jobs[1]["mask"].getRunAndLumis(), {0L: [[3L, 4L]]},
"Wrong mask for the second job")
self.assertEqual(jobs[2]["mask"].getRunAndLumis(), {1L: [[1L, 1L]]},
"Wrong mask for the third job")
self.assertEqual(jobs[3]["mask"].getRunAndLumis(), {2L: [[4L, 4L]]},
"Wrong mask for the fourth job")
self.assertEqual(jobs[4]["mask"].getRunAndLumis(), {2L: [[5L, 5L]]},
"Wrong mask for the fifth job")
self.assertEqual(jobs[5]["mask"].getRunAndLumis(), {
3L: [[3L, 3L]],
4L: [[4L, 4L]],
5L: [[5L, 5L]]
}, "Wrong mask for the sixth job")
self.assertEqual(jobs[6]["mask"].getRunAndLumis(), {6L: [[18L, 19L]]},
"Wrong mask for the seventh job")
self.assertEqual(jobs[7]["mask"].getRunAndLumis(), {6L: [[20L, 20L]]},
"Wrong mask for the seventh job")
#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/file1", 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("malpaquet")
fileB = self.createFile('/this/is/file2', 200, 3, 5, "malpaquet")
testFileset = Fileset(name='FilesetB')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiBased",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
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)
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")
def testChunking(self):
"""
_testChunking_
Insert a workload and files that have several distinct sets of
locations. Verify that the chunks are created correctly and that they
only groups files that have the same set of locations. Also verify that
the chunks are pulled out of ACDC correctly.
"""
dcs = DataCollectionService(url = self.testInit.couchUrl,
database = "wmcore-acdc-datacollectionsvc")
def getJob():
job = Job()
job["task"] = "/ACDCTest/reco"
job["workflow"] = "ACDCTest"
job["location"] = "cmssrm.fnal.gov"
job["owner"] = "cmsdataops"
job["group"] = "cmsdataops"
return job
testFileA = File(lfn = makeUUID(), size = 1024, events = 1024)
testFileA.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileA.addRun(Run(1, 1, 2))
testFileB = File(lfn = makeUUID(), size = 1024, events = 1024)
testFileB.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileB.addRun(Run(1, 3, 4))
testFileC = File(lfn = makeUUID(), size = 1024, events = 1024)
testFileC.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileC.addRun(Run(1, 5, 6))
testJobA = getJob()
testJobA.addFile(testFileA)
testJobA.addFile(testFileB)
testJobA.addFile(testFileC)
testFileD = File(lfn = makeUUID(), size = 1024, events = 1024)
testFileD.setLocation(["cmssrm.fnal.gov"])
testFileD.addRun(Run(2, 1, 2))
testFileE = File(lfn = makeUUID(), size = 1024, events = 1024)
testFileE.setLocation(["cmssrm.fnal.gov"])
testFileE.addRun(Run(2, 3, 4))
testJobB = getJob()
testJobB.addFile(testFileD)
testJobB.addFile(testFileE)
testFileF = File(lfn = makeUUID(), size = 1024, events = 1024,
parents = set(["/some/parent/F"]))
testFileF.setLocation(["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"])
testFileF.addRun(Run(3, 1, 2))
testFileG = File(lfn = makeUUID(), size = 1024, events = 1024,
parents = set(["/some/parent/G"]))
testFileG.setLocation(["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"] )
testFileG.addRun(Run(3, 3, 4))
testFileH = File(lfn = makeUUID(), size = 1024, events = 1024,
parents = set(["/some/parent/H"]))
testFileH.setLocation(["cmssrm.fnal.gov", "castor.cern.ch", "srm.ral.uk"])
testFileH.addRun(Run(3, 5, 6))
testJobC = getJob()
testJobC.addFile(testFileF)
testJobC.addFile(testFileG)
testJobC.addFile(testFileH)
testFileI = File(lfn = makeUUID(), size = 1024, events = 1024, merged = True)
testFileI.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileI.addRun(Run(4, 1, 2))
testFileJ = File(lfn = makeUUID(), size = 1024, events = 1024, merged = True)
testFileJ.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"] )
testFileJ.addRun(Run(4, 3, 4))
testFileK = File(lfn = makeUUID(), size = 1024, events = 1024, merged = True)
testFileK.setLocation(["cmssrm.fnal.gov", "castor.cern.ch"])
testFileK.addRun(Run(4, 5, 6))
testJobD = getJob()
testJobD.addFile(testFileI)
testJobD.addFile(testFileJ)
testJobD.addFile(testFileK)
dcs.failedJobs([testJobA, testJobB, testJobC, testJobD])
chunks = dcs.chunkFileset("ACDCTest", "/ACDCTest/reco",
chunkSize = 5)
self.assertEqual(len(chunks), 4, "Error: There should be four chunks: %s" % len(chunks))
goldenMetaData = {1: {"lumis": 2, "locations": ["castor.cern.ch", "cmssrm.fnal.gov"], "events": 1024},
2: {"lumis": 4, "locations": ["cmssrm.fnal.gov"], "events": 2048},
3: {"lumis": 6, "locations": ["castor.cern.ch", "cmssrm.fnal.gov", "srm.ral.uk"], "events": 3072},
5: {"lumis": 10, "locations": ["castor.cern.ch", "cmssrm.fnal.gov"], "events": 5120}}
testFiles =[testFileA, testFileB, testFileC, testFileI, testFileJ, testFileK]
lastFile = testFileA
for testFile in testFiles:
if lastFile["lfn"] < testFile["lfn"]:
lastFile = testFile
testFiles.remove(lastFile)
goldenFiles = {1: [lastFile],
2: [testFileD, testFileE],
3: [testFileF, testFileG, testFileH],
5: testFiles}
for chunk in chunks:
chunkMetaData = dcs.getChunkInfo("ACDCTest", "/ACDCTest/reco",
chunk["offset"], chunk["files"])
self.assertEqual(chunkMetaData["files"], chunk["files"],
"Error: Metadata doesn't match.")
self.assertEqual(chunkMetaData["lumis"], chunk["lumis"],
"Error: Metadata doesn't match.")
self.assertEqual(chunkMetaData["events"], chunk["events"],
"Error: Metadata doesn't match.")
self.assertEqual(chunkMetaData["locations"], chunk["locations"],
"Error: Metadata doesn't match.")
self.assertTrue(chunk["files"] in goldenMetaData.keys(),
"Error: Extra chunk found.")
self.assertEqual(chunk["lumis"], goldenMetaData[chunk["files"]]["lumis"],
"Error: Lumis in chunk is wrong.")
self.assertEqual(chunk["locations"], goldenMetaData[chunk["files"]]["locations"],
"Error: Locations in chunk is wrong.")
self.assertEqual(chunk["events"], goldenMetaData[chunk["files"]]["events"],
"Error: Events in chunk is wrong.")
del goldenMetaData[chunk["files"]]
chunkFiles = dcs.getChunkFiles("ACDCTest", "/ACDCTest/reco",
chunk["offset"], chunk["files"])
self.assertTrue(chunk["files"] in goldenFiles.keys(),
"Error: Extra chunk found.")
goldenChunkFiles = goldenFiles[chunk["files"]]
self.assertEqual(len(chunkFiles), len(goldenChunkFiles))
for chunkFile in chunkFiles:
foundFile = None
for goldenChunkFile in goldenChunkFiles:
if chunkFile["lfn"] == goldenChunkFile["lfn"]:
foundFile = goldenChunkFile
break
self.assertTrue(foundFile != None,
"Error: Missing chunk file: %s, %s" % (chunkFiles, goldenChunkFiles))
self.assertEqual(foundFile["parents"], chunkFile["parents"],
"Error: File parents should match.")
self.assertEqual(foundFile["merged"], chunkFile["merged"],
"Error: File merged status should match.")
self.assertEqual(foundFile["locations"], chunkFile["locations"],
"Error: File locations should match.")
self.assertEqual(foundFile["events"], chunkFile["events"],
"Error: File locations should match: %s" % chunk["files"])
self.assertEqual(foundFile["size"], chunkFile["size"],
"Error: File locations should match.")
self.assertEqual(len(foundFile["runs"]), len(chunkFile["runs"]),
"Error: Wrong number of runs.")
for run in foundFile["runs"]:
runMatch = False
for chunkRun in chunkFile["runs"]:
if chunkRun.run == run.run and chunkRun.lumis == run.lumis:
runMatch = True
break
self.assertTrue(runMatch, "Error: Run information is wrong.")
del goldenFiles[chunk["files"]]
singleChunk = dcs.singleChunkFileset("ACDCTest", "/ACDCTest/reco")
self.assertEqual(singleChunk, {"offset" : 0,
"files" : 11,
"events" : 11264,
"lumis" : 22,
"locations" : set(["castor.cern.ch", "cmssrm.fnal.gov", "srm.ral.uk"])},
"Error: Single chunk metadata is wrong")
return
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.DataStructs",
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)
# One job in one job group with 100 files
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, "blenheim")
testFileB = self.createFile("/this/is/file2", 600, 1, 1, "blenheim")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2, "blenheim")
testFileD = self.createFile("/this/is/file4", 100, 3, 1, "blenheim")
testFileE = self.createFile("/this/is/file5", 30, 4, 1, "blenheim")
testFileF = self.createFile("/this/is/file6", 10, 5, 1, "blenheim")
testFileG = self.createFile("/this/is/file7", 153, 6, 3, "blenheim")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
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/file1", 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("malpaquet")
fileB = self.createFile('/this/is/file2', 200, 3, 5, "malpaquet")
testFileset = Fileset(name='FilesetB')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
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 testRunWhiteList(self):
"""
_testRunWhiteList_
Test that we can use a run white list 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("somese.cern.ch")
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("somese.cern.ch")
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("somese.cern.ch")
testFileset = Fileset(name='Fileset')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
subscription=testSubscription)
# Split with no breaks
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=725,
runWhitelist=[1, 4],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 2)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 4])
# Re-split with a break on runs
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=True,
events_per_job=595,
runWhitelist=[1, 3, 4],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 4)
self.enforceLimits(jobs=jobs, runsPerJob=1)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 3, 4])
# Re-split with a break on files
jobGroups = jobFactory(halt_job_on_file_boundaries=True,
splitOnRun=False,
events_per_job=595,
runWhitelist=[1, 2, 3],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 3)
self.enforceLimits(jobs=jobs, filesPerJob=1)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 2, 3])
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.DataStructs", 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)
# One job in one job group with 100 files
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, "blenheim")
testFileB = self.createFile("/this/is/file2", 600, 1, 1, "blenheim")
testFileC = self.createFile("/this/is/file3", 1200, 2, 2, "blenheim")
testFileD = self.createFile("/this/is/file4", 100, 3, 1, "blenheim")
testFileE = self.createFile("/this/is/file5", 30, 4, 1, "blenheim")
testFileF = self.createFile("/this/is/file6", 10, 5, 1, "blenheim")
testFileG = self.createFile("/this/is/file7", 153, 6, 3, "blenheim")
testFileset.addFile(testFileA)
testFileset.addFile(testFileB)
testFileset.addFile(testFileC)
testFileset.addFile(testFileD)
testFileset.addFile(testFileE)
testFileset.addFile(testFileF)
testFileset.addFile(testFileG)
testSubscription = Subscription(fileset=testFileset, workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork", type="Processing")
jobFactory = splitter(package="WMCore.DataStructs", 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/file1", 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("malpaquet")
fileB = self.createFile('/this/is/file2', 200, 3, 5, "malpaquet")
testFileset = Fileset(name='FilesetB')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testSubscription = Subscription(fileset=testFileset, workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork", type="Processing")
jobFactory = splitter(package="WMCore.DataStructs", 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 testRunWhiteList(self):
"""
_testRunWhiteList_
Test that we can use a run white list 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("somese.cern.ch")
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("somese.cern.ch")
lumiListC = []
for lumi in range(5):
lumiListC.append(40 + lumi)
fileC.addRun(Run(4, *lumiListC))
fileC.setLocation("somese.cern.ch")
testFileset = Fileset(name='Fileset')
testFileset.addFile(fileA)
testFileset.addFile(fileB)
testFileset.addFile(fileC)
testSubscription = Subscription(fileset=testFileset,
workflow=self.testWorkflow,
split_algo="EventAwareLumiByWork",
type="Processing")
jobFactory = splitter(package="WMCore.DataStructs",
subscription=testSubscription)
# Split with no breaks
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=False,
events_per_job=725,
runWhitelist=[1, 4],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 2)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 4])
# Re-split with a break on runs
jobGroups = jobFactory(halt_job_on_file_boundaries=False,
splitOnRun=True,
events_per_job=595,
runWhitelist=[1, 3, 4],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 4)
self.enforceLimits(jobs=jobs, runsPerJob=1)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 3, 4])
# Re-split with a break on files
jobGroups = jobFactory(halt_job_on_file_boundaries=True,
splitOnRun=False,
events_per_job=595,
runWhitelist=[1, 2, 3],
performance=self.performanceParams)
self.assertEqual(len(jobGroups), 1)
jobs = jobGroups[0].jobs
self.assertEqual(len(jobs), 3)
self.enforceLimits(jobs=jobs, filesPerJob=1)
for job in jobs:
for run in job['mask'].getRunAndLumis().keys():
self.assertIn(run, [1, 2, 3])
