def __init__(self, startRun = None, purgeTime = 48,
phedexUrl = "https://cmsweb.cern.ch/phedex/datasvc/json/prod/fileReplicas",
dbsHost = "cmsweb.cern.ch/dbs_discovery/",
dbsInstance = "cms_dbs_prod_global",
dbsPort = 443):
"""
Configure the feeder
"""
# Configure data service look up components
self.dbsHelper = DbsQueryHelper(dbsHost, dbsPort, dbsInstance)
self.phedexHelper = PhEDExQueryHelper(phedexUrl)
# Runs that are being watched
self.watchedRuns = []
# The last run that was identified as new, and run purge time
self.lastRun = 0
self.purgeTime = purgeTime * 3600 # Convert hours to seconds
# Bootstrap run list
if not startRun:
# Determine the last run registered in DBS
runs = self.dbsHelper.getRuns()
runs.sort(reverse=True)
if len(runs) == 0:
msg = "Could not bootstrap RunTransferNotifier feeder"
raise RuntimeError, msg
# There are runs, ensure we ignore them in first query
self.lastRun = runs[0]
else:
# Ensure we include the startRun in first query
self.lastRun = startRun - 1
class RunTransferNotifier(FeederImpl):
"""
Run / transfer feeder implementation
"""
class WatchedRun:
"""
Records a run that is being watched, and maintains state about
transfered files for monitored datasets
"""
def __init__(self, run):
self.run = run
self.creationTime = int(time())
self.accessTime = int(time())
self.datasetCompletion = {}
def addDatasetOfInterest(self, dataset):
"""
Adds an entry to Dataset : CompleteLocation tracking
"""
if not self.datasetCompletion.has_key(dataset):
self.datasetCompletion[dataset] = set()
self.accessTime = int(time())
def addCompletedNodes(self, dataset, nodes):
"""
Records completed transfer for a dataset to a node
"""
if isinstance(node, list):
self.datasetCompletion[dataset].extend(nodes)
else:
self.datasetCompletion[dataset].append(nodes)
self.accessTime = int(time())
def getNewSites(self, dataset, sitesWithRun):
"""
Returns all sites that have not been marked as complete
"""
return sitesWithRun - self.datasetCompletion[dataset]
def __init__(
self,
startRun=None,
purgeTime=48,
phedexUrl="https://cmsweb.cern.ch/phedex/datasvc/json/prod/fileReplicas",
dbsHost="cmsweb.cern.ch/dbs_discovery/",
dbsInstance="cms_dbs_prod_global",
dbsPort=443):
"""
Configure the feeder
"""
# Configure data service look up components
self.dbsHelper = DbsQueryHelper(dbsHost, dbsPort, dbsInstance)
self.phedexHelper = PhEDExQueryHelper(phedexUrl)
# Runs that are being watched
self.watchedRuns = []
# The last run that was identified as new, and run purge time
self.lastRun = 0
self.purgeTime = purgeTime * 3600 # Convert hours to seconds
# Bootstrap run list
if not startRun:
# Determine the last run registered in DBS
runs = self.dbsHelper.getRuns()
runs.sort(reverse=True)
if len(runs) == 0:
msg = "Could not bootstrap RunTransferNotifier feeder"
raise RuntimeError, msg
# There are runs, ensure we ignore them in first query
self.lastRun = runs[0]
else:
# Ensure we include the startRun in first query
self.lastRun = startRun - 1
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 getNewRuns(self):
"""
Queries DBS to determine what new runs are present, and adds a watcher
"""
runs = self.dbsHelper.getRuns(self.lastRun)
runs.sort()
for run in runs:
watchedRuns.append(WatchedRun(run))
self.lastRun = run
def purgeWatchedRuns(self):
"""
Purges watched runs that were last accessed longer than purgeTime ago
"""
validRuns = []
for run in self.watchedRuns:
if int(time) - run.accessTime < self.purgeTime:
validRuns.append(run)
self.watchedRuns = validRuns
class RunTransferNotifier(FeederImpl):
"""
Run / transfer feeder implementation
"""
class WatchedRun:
"""
Records a run that is being watched, and maintains state about
transfered files for monitored datasets
"""
def __init__(self, run):
self.run = run
self.creationTime = int(time())
self.accessTime = int(time())
self.datasetCompletion = {}
def addDatasetOfInterest(self, dataset):
"""
Adds an entry to Dataset : CompleteLocation tracking
"""
if dataset not in self.datasetCompletion:
self.datasetCompletion[dataset] = set()
self.accessTime = int(time())
def addCompletedNodes(self, dataset, nodes):
"""
Records completed transfer for a dataset to a node
"""
if isinstance(node, list):
self.datasetCompletion[dataset].extend(nodes)
else:
self.datasetCompletion[dataset].append(nodes)
self.accessTime = int(time())
def getNewSites(self, dataset, sitesWithRun):
"""
Returns all sites that have not been marked as complete
"""
return sitesWithRun - self.datasetCompletion[dataset]
def __init__(self, startRun = None, purgeTime = 48,
phedexUrl = "https://cmsweb.cern.ch/phedex/datasvc/json/prod/fileReplicas",
dbsHost = "cmsweb.cern.ch/dbs_discovery/",
dbsInstance = "cms_dbs_prod_global",
dbsPort = 443):
"""
Configure the feeder
"""
# Configure data service look up components
self.dbsHelper = DbsQueryHelper(dbsHost, dbsPort, dbsInstance)
self.phedexHelper = PhEDExQueryHelper(phedexUrl)
# Runs that are being watched
self.watchedRuns = []
# The last run that was identified as new, and run purge time
self.lastRun = 0
self.purgeTime = purgeTime * 3600 # Convert hours to seconds
# Bootstrap run list
if not startRun:
# Determine the last run registered in DBS
runs = self.dbsHelper.getRuns()
runs.sort(reverse=True)
if len(runs) == 0:
msg = "Could not bootstrap RunTransferNotifier feeder"
raise RuntimeError, msg
# There are runs, ensure we ignore them in first query
self.lastRun = runs[0]
else:
# Ensure we include the startRun in first query
self.lastRun = startRun - 1
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 getNewRuns(self):
"""
Queries DBS to determine what new runs are present, and adds a watcher
"""
runs = self.dbsHelper.getRuns(self.lastRun)
runs.sort()
for run in runs:
watchedRuns.append(WatchedRun(run))
self.lastRun = run
def purgeWatchedRuns(self):
"""
Purges watched runs that were last accessed longer than purgeTime ago
"""
validRuns = []
for run in self.watchedRuns:
if int(time) - run.accessTime < self.purgeTime:
validRuns.append(run)
self.watchedRuns = validRuns
