def __init__( self, csPath = None, ftsSites = None, ftsHistoryViews = None ):
"""c'tor
:param self: self reference
:param str csPath: CS path
:param list ftsSites: list of FTSSites
:param list ftsHistoryViews: list of FTSHistoryViews
"""
# ## config path
self.csPath = csPath
# # fts sites
ftsSites = ftsSites if ftsSites else []
# # history views
ftsHistoryViews = ftsHistoryViews if ftsHistoryViews else []
# # own sub logger
self.log = gLogger.getSubLogger( "FTSStrategy", child = True )
self.log.setLevel( gConfig.getValue( self.csPath + "/LogLevel", "DEBUG" ) )
for ftsSite in ftsSites:
self.log.info( "FTSSite: %-16s FTSServer=%s" % ( ftsSite.Name, ftsSite.FTSServer ) )
# # CS options
self.log.info( "Supported strategies = %s" % ", ".join( self.supportedStrategies ) )
self.activeStrategies = gConfig.getValue( "%s/%s" % ( self.csPath, "ActiveStrategies" ), ["MinimiseTotalWait"] )
self.log.info( "ActiveStrategies = %s" % ", ".join( self.activeStrategies ) )
self.numberOfStrategies = len( self.activeStrategies )
self.log.info( "Number of active strategies = %s" % self.numberOfStrategies )
self.sigma = gConfig.getValue( "%s/%s" % ( self.csPath, "HopSigma" ), 5 )
self.log.info( "HopSigma = %s" % self.sigma )
self.schedulingType = gConfig.getValue( "%s/%s" % ( self.csPath, "SchedulingType" ), "Files" )
self.log.info( "SchedulingType = %s" % self.schedulingType )
self.acceptableFailureRate = gConfig.getValue( "%s/%s" % ( self.csPath, "AcceptableFailureRate" ), 75 )
self.log.info( "AcceptableFailureRate = %s" % self.acceptableFailureRate )
self.acceptableFailedFiles = gConfig.getValue( "%s/%s" % ( self.csPath, "AcceptableFailedFiles" ), 5 )
self.log.info( "AcceptableFailedFiles = %s" % self.acceptableFailedFiles )
# # chosen strategy
self.chosenStrategy = 0
# dispatcher
self.strategyDispatcher = { "MinimiseTotalWait" : self.minimiseTotalWait,
"DynamicThroughput" : self.dynamicThroughput,
"Simple" : self.simple,
"Swarm" : self.swarm }
self.ftsGraph = FTSGraph( "FTSGraph",
ftsHistoryViews,
self.acceptableFailureRate,
self.acceptableFailedFiles,
self.schedulingType )
# for node in self.ftsGraph.nodes():
# self.log.debug( node )
# for edge in self.ftsGraph.edges():
# self.log.debug( edge )
# # if we land here everything is OK
self.log.info( "%s has been constructed" % self.__class__.__name__ )
def test(self):
""" test case """
graph = FTSGraph("ftsGraph", self.ftsHistoryViews)
self.assertEqual(type(graph), FTSGraph, "c'tor failed")
site = graph.findSiteForSE("CERN-FOO")
self.assertEqual(site["OK"], False,
"findSiteForSE call failed for unknown SE")
sourceSite = graph.findSiteForSE("CERN-USER")
self.assertEqual(sourceSite["OK"], True,
"findSiteForSE call failed for target SE")
targetSite = graph.findSiteForSE("RAL-USER")
self.assertEqual(targetSite["OK"], True,
"findSiteForSE call failed for source SE")
route = graph.findRoute("RAL-USER", "CERN-USER")
self.assertEqual(route["OK"], True,
"findRoute failed for known source and target SEs")
route = graph.findRoute("RAL-FOO", "CERN-BAR")
self.assertEqual(route["OK"], False,
"findRoute failed for unknown source and target SEs")
def resetFTSGraph(self):
""" create fts graph """
log = gLogger.getSubLogger("ftsGraph")
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error("unable to get FTS history: %s" % ftsHistory["Message"])
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph("FTSGraph", ftsHistory)
finally:
self.updateLock().release()
log.debug("FTSSites: %s" % len(self.__ftsGraph.nodes()))
for i, site in enumerate(self.__ftsGraph.nodes()):
log.debug(" [%02d] FTSSite: %-25s FTSServer: %s" %
(i, site.name, site.FTSServer))
log.debug("FTSRoutes: %s" % len(self.__ftsGraph.edges()))
for i, route in enumerate(self.__ftsGraph.edges()):
log.debug(
" [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" %
(i, route.routeName, route.ActiveJobs,
route.toNode.MaxActiveJobs))
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now(
) + datetime.timedelta(seconds=self.FTSGRAPH_REFRESH)
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now(
) + datetime.timedelta(seconds=self.RW_REFRESH)
return S_OK()
def resetGraph(cls, ftsHistoryViews):
""" reset graph
:param list ftsHistoryViews: list of FTSHistoryViews
"""
ftsGraph = None
try:
cls.graphLock().acquire()
ftsGraph = FTSGraph("FTSGraph", ftsHistoryViews,
cls.acceptableFailureRate,
cls.acceptableFailedFiles, cls.schedulingType)
if ftsGraph:
cls.ftsGraph = ftsGraph
finally:
cls.graphLock().release()
return S_OK()
def resetFTSGraph( self ):
""" create fts graph """
log = gLogger.getSubLogger( "ftsGraph" )
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error( "unable to get FTS history: %s" % ftsHistory["Message"] )
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph( "FTSGraph", ftsHistory )
finally:
self.updateLock().release()
log.debug( "FTSSites: %s" % len( self.__ftsGraph.nodes() ) )
for i, site in enumerate( self.__ftsGraph.nodes() ):
log.debug( " [%02d] FTSSite: %-25s FTSServer: %s" % ( i, site.name, site.FTSServer ) )
log.debug( "FTSRoutes: %s" % len( self.__ftsGraph.edges() ) )
for i, route in enumerate( self.__ftsGraph.edges() ):
log.debug( " [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" % ( i,
route.routeName,
route.ActiveJobs,
route.toNode.MaxActiveJobs ) )
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.RW_REFRESH )
return S_OK()
def test( self ):
""" test case """
graph = FTSGraph( "ftsGraph", self.ftsHistoryViews )
self.assertEqual( type( graph ), FTSGraph, "c'tor failed" )
site = graph.findSiteForSE( "CERN-FOO" )
self.assertEqual( site["OK"], False, "findSiteForSE call failed for unknown SE" )
sourceSite = graph.findSiteForSE( "CERN-USER" )
self.assertEqual( sourceSite["OK"], True, "findSiteForSE call failed for target SE" )
targetSite = graph.findSiteForSE( "RAL-USER" )
self.assertEqual( targetSite["OK"], True, "findSiteForSE call failed for source SE" )
route = graph.findRoute( "RAL-USER", "CERN-USER" )
self.assertEqual( route["OK"], True, "findRoute failed for known source and target SEs" )
route = graph.findRoute( "RAL-FOO", "CERN-BAR" )
self.assertEqual( route["OK"], False, "findRoute failed for unknown source and target SEs" )
class FTSAgent(AgentModule):
"""
.. class:: FTSAgent
Agent propagating Scheduled request to Done or Failed state in the FTS system.
Requests and associated FTSJobs (and so FTSFiles) are kept in cache.
"""
# # fts graph refresh in seconds
FTSGRAPH_REFRESH = FTSHistoryView.INTERVAL / 2
# # SE R/W access refresh in seconds
RW_REFRESH = 600
# # placeholder for max job per channel
MAX_ACTIVE_JOBS = 50
# # min threads
MIN_THREADS = 1
# # max threads
MAX_THREADS = 10
# # files per job
MAX_FILES_PER_JOB = 100
# # MAX FTS transfer per FTSFile
MAX_ATTEMPT = 256
# # stage flag
STAGE_FILES = False
# # replica manager
__replicaManager = None
# # placeholder for FTS client
__ftsClient = None
# # placeholder for request client
__requestClient = None
# # placeholder for resources helper
__resources = None
# # placeholder for RSS client
__rssClient = None
# # placeholder for FTSGraph
__ftsGraph = None
# # graph regeneration time delta
__ftsGraphValidStamp = None
# # r/w access valid stamp
__rwAccessValidStamp = None
# # placeholder for threadPool
__threadPool = None
# # update lock
__updateLock = None
# # se cache
__seCache = dict()
# # request cache
__reqCache = dict()
def updateLock(self):
""" update lock """
if not self.__updateLock:
self.__updateLock = LockRing().getLock("FTSAgentLock")
return self.__updateLock
@classmethod
def requestClient(cls):
""" request client getter """
if not cls.__requestClient:
cls.__requestClient = ReqClient()
return cls.__requestClient
@classmethod
def ftsClient(cls):
""" FTS client """
if not cls.__ftsClient:
cls.__ftsClient = FTSClient()
return cls.__ftsClient
@classmethod
def replicaManager(cls):
""" replica manager getter """
if not cls.__replicaManager:
cls.__replicaManager = ReplicaManager()
return cls.__replicaManager
@classmethod
def rssClient(cls):
""" RSS client getter """
if not cls.__rssClient:
cls.__rssClient = ResourceStatus()
return cls.__rssClient
@classmethod
def getSE(cls, seName):
""" keep SEs in cache """
if seName not in cls.__seCache:
cls.__seCache[seName] = StorageElement(seName)
return cls.__seCache[seName]
@classmethod
def getRequest(cls, reqName):
""" keep Requests in cache """
if reqName not in cls.__reqCache:
getRequest = cls.requestClient().getRequest(reqName)
if not getRequest["OK"]:
return getRequest
getRequest = getRequest["Value"]
if not getRequest:
return S_ERROR("request of name '%s' not found in ReqDB" %
reqName)
cls.__reqCache[reqName] = getRequest
return S_OK(cls.__reqCache[reqName])
@classmethod
def putRequest(cls, request):
""" put request back to ReqDB
:param Request request: Request instance
also finalize request if status == Done
"""
# # put back request
put = cls.requestClient().putRequest(request)
if not put["OK"]:
return put
# # finalize first is possible
if request.Status == "Done" and request.JobID:
finalizeRequest = cls.requestClient().finalizeRequest(
request.RequestName, request.JobID)
if not finalizeRequest["OK"]:
request.Status = "Scheduled"
# # del request from cache
if request.RequestName in cls.__reqCache:
del cls.__reqCache[request.RequestName]
return S_OK()
@classmethod
def putFTSJobs(cls, ftsJobsList):
""" put back fts jobs to the FTSDB """
for ftsJob in ftsJobsList:
put = cls.ftsClient().putFTSJob(ftsJob)
if not put["OK"]:
return put
return S_OK()
@staticmethod
def updateFTSFileDict(ftsFilesDict, toUpdateDict):
""" update :ftsFilesDict: with FTSFiles in :toUpdateDict: """
for category, ftsFileList in ftsFilesDict.items():
for ftsFile in toUpdateDict.get(category, []):
if ftsFile not in ftsFileList:
ftsFileList.append(ftsFile)
return ftsFilesDict
# def resources( self ):
# """ resource helper getter """
# if not self.__resources:
# self.__resources = Resources()
# return self.__resources
def threadPool(self):
""" thread pool getter """
if not self.__threadPool:
self.__threadPool = ThreadPool(self.MIN_THREADS, self.MAX_THREADS)
self.__threadPool.daemonize()
return self.__threadPool
def resetFTSGraph(self):
""" create fts graph """
log = gLogger.getSubLogger("ftsGraph")
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error("unable to get FTS history: %s" % ftsHistory["Message"])
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph("FTSGraph", ftsHistory)
finally:
self.updateLock().release()
log.debug("FTSSites: %s" % len(self.__ftsGraph.nodes()))
for i, site in enumerate(self.__ftsGraph.nodes()):
log.debug(" [%02d] FTSSite: %-25s FTSServer: %s" %
(i, site.name, site.FTSServer))
log.debug("FTSRoutes: %s" % len(self.__ftsGraph.edges()))
for i, route in enumerate(self.__ftsGraph.edges()):
log.debug(
" [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" %
(i, route.routeName, route.ActiveJobs,
route.toNode.MaxActiveJobs))
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now(
) + datetime.timedelta(seconds=self.FTSGRAPH_REFRESH)
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now(
) + datetime.timedelta(seconds=self.RW_REFRESH)
return S_OK()
def initialize(self):
""" agent's initialization """
log = self.log.getSubLogger("initialize")
self.FTSGRAPH_REFRESH = self.am_getOption("FTSGraphValidityPeriod",
self.FTSGRAPH_REFRESH)
log.info("FTSGraph validity period = %s s" %
self.FTSGRAPH_REFRESH)
self.RW_REFRESH = self.am_getOption("RWAccessValidityPeriod",
self.RW_REFRESH)
log.info("SEs R/W access validity period = %s s" % self.RW_REFRESH)
self.STAGE_FILES = self.am_getOption("StageFiles", self.STAGE_FILES)
log.info("Stage files before submission = %s" % {
True: "yes",
False: "no"
}[bool(self.STAGE_FILES)])
self.MAX_ACTIVE_JOBS = self.am_getOption("MaxActiveJobsPerRoute",
self.MAX_ACTIVE_JOBS)
log.info("Max active FTSJobs/route = %s" % self.MAX_ACTIVE_JOBS)
self.MAX_FILES_PER_JOB = self.am_getOption("MaxFilesPerJob",
self.MAX_FILES_PER_JOB)
log.info("Max FTSFiles/FTSJob = %d" %
self.MAX_FILES_PER_JOB)
self.MAX_ATTEMPT = self.am_getOption("MaxTransferAttempts",
self.MAX_ATTEMPT)
log.info("Max transfer attempts = %s" % self.MAX_ATTEMPT)
# # thread pool
self.MIN_THREADS = self.am_getOption("MinThreads", self.MIN_THREADS)
self.MAX_THREADS = self.am_getOption("MaxThreads", self.MAX_THREADS)
minmax = (abs(self.MIN_THREADS), abs(self.MAX_THREADS))
self.MIN_THREADS, self.MAX_THREADS = min(minmax), max(minmax)
log.info("ThreadPool min threads = %s" % self.MIN_THREADS)
log.info("ThreadPool max threads = %s" % self.MAX_THREADS)
log.info("initialize: creation of FTSGraph...")
createGraph = self.resetFTSGraph()
if not createGraph["OK"]:
log.error("initialize: %s" % createGraph["Message"])
return createGraph
# This sets the Default Proxy to used as that defined under
# /Operations/Shifter/DataManager
# the shifterProxy option in the Configuration can be used to change this default.
self.am_setOption('shifterProxy', 'DataManager')
log.info("will use DataManager proxy")
# # gMonitor stuff here
gMonitor.registerActivity("RequestsAtt",
"Attempted requests executions", "FTSAgent",
"Requests/min", gMonitor.OP_SUM)
gMonitor.registerActivity("RequestsOK",
"Successful requests executions", "FTSAgent",
"Requests/min", gMonitor.OP_SUM)
gMonitor.registerActivity("RequestsFail", "Failed requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsSubAtt", "FTSJobs creation attempts",
"FTSAgent", "Created FTSJobs/min",
gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsSubOK",
"FTSJobs submitted successfully", "FTSAgent",
"Successful FTSJobs submissions/min",
gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsSubFail",
"FTSJobs submissions failed", "FTSAgent",
"Failed FTSJobs submissions/min",
gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsMonAtt", "FTSJobs monitored",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsMonOK",
"FTSJobs monitored successfully", "FTSAgent",
"FTSJobs/min", gMonitor.OP_SUM)
gMonitor.registerActivity("FTSJobsMonFail", "FTSJobs attempts failed",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM)
gMonitor.registerActivity("FTSMonitorFail",
"Failed FTS monitor executions", "FTSAgent",
"Execution/mins", gMonitor.OP_SUM)
pollingTime = self.am_getOption("PollingTime", 60)
for status in list(FTSJob.INITSTATES + FTSJob.TRANSSTATES +
FTSJob.FAILEDSTATES + FTSJob.FINALSTATES):
gMonitor.registerActivity("FTSJobs%s" % status,
"FTSJobs %s" % status, "FTSAgent",
"FTSJobs/cycle", gMonitor.OP_ACUM,
pollingTime)
gMonitor.registerActivity("FtSJobsPerRequest",
"Average FTSJobs per request", "FTSAgent",
"FTSJobs/Request", gMonitor.OP_MEAN)
gMonitor.registerActivity("FTSFilesPerJob", "FTSFiles per FTSJob",
"FTSAgent", "Number of FTSFiles per FTSJob",
gMonitor.OP_MEAN)
gMonitor.registerActivity("FTSSizePerJob",
"Average FTSFiles size per FTSJob",
"FTSAgent",
"Average submitted size per FTSJob",
gMonitor.OP_MEAN)
return S_OK()
def finalize(self):
""" finalize processing """
log = self.log.getSubLogger("finalize")
for request in self.__reqCache.values():
put = self.requestClient().putRequest(request)
if not put["OK"]:
log.error("unable to put back request '%s': %s" %
(request.RequestName, put["Message"]))
return S_OK()
def execute(self):
""" one cycle execution """
log = gLogger.getSubLogger("execute")
# # reset FTSGraph if expired
now = datetime.datetime.now()
if now > self.__ftsGraphValidStamp:
log.info("resetting expired FTS graph...")
resetFTSGraph = self.resetFTSGraph()
if not resetFTSGraph["OK"]:
log.error("FTSGraph recreation error: %s" %
resetFTSGraph["Message"])
return resetFTSGraph
self.__ftsGraphValidStamp = now + datetime.timedelta(
seconds=self.FTSGRAPH_REFRESH)
# # update R/W access in FTSGraph if expired
if now > self.__rwAccessValidStamp:
log.info("updating expired R/W access for SEs...")
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
self.__rwAccessValidStamp = now + datetime.timedelta(
seconds=self.RW_REFRESH)
requestNames = self.requestClient().getRequestNamesList(["Scheduled"])
if not requestNames["OK"]:
log.error("unable to read scheduled request names: %s" %
requestNames["Message"])
return requestNames
if not requestNames["Value"]:
requestNames = self.__reqCache.keys()
else:
requestNames = [req[0] for req in requestNames["Value"]]
requestNames = list(set(requestNames + self.__reqCache.keys()))
if not requestNames:
log.info("no 'Scheduled' requests to process")
return S_OK()
log.info("found %s requests to process:" % len(requestNames))
log.info(" => from internal cache: %s" % (len(self.__reqCache)))
log.info(" => new read from RMS: %s" %
(len(requestNames) - len(self.__reqCache)))
for requestName in requestNames:
request = self.getRequest(requestName)
if not request["OK"]:
log.error(request["Message"])
continue
request = request["Value"]
sTJId = request.RequestName
while True:
queue = self.threadPool().generateJobAndQueueIt(
self.processRequest, args=(request, ), sTJId=sTJId)
if queue["OK"]:
log.info("request '%s' enqueued for execution" % sTJId)
gMonitor.addMark("RequestsAtt", 1)
break
time.sleep(1)
# # process all results
self.threadPool().processAllResults()
return S_OK()
def processRequest(self, request):
""" process one request
:param Request request: ReqDB.Request
"""
log = self.log.getSubLogger(request.RequestName)
operation = request.getWaiting()
if not operation["OK"]:
log.error(
"unable to find 'Scheduled' ReplicateAndRegister operation in request"
)
return self.putRequest(request)
operation = operation["Value"]
if operation.Type != "ReplicateAndRegister":
log.error(
"operation to be executed is not a ReplicateAndRegister but %s"
% operation.Type)
return self.putRequest(request)
if operation.Status != "Scheduled":
log.error(
"operation in a wrong state, expecting 'Scheduled', got %s" %
operation.Status)
return self.putRequest(request)
# # select FTSJobs, by default all in TRANS_STATES and INIT_STATES
ftsJobs = self.ftsClient().getFTSJobsForRequest(request.RequestID)
if not ftsJobs["OK"]:
log.error(ftsJobs["Message"])
return ftsJobs
ftsJobs = ftsJobs["Value"] if ftsJobs["Value"] else []
# # dict keeping info about files to reschedule, submit, fail and register
ftsFilesDict = dict([(k, list())
for k in ("toRegister", "toSubmit", "toFail",
"toReschedule", "toUpdate")])
if ftsJobs:
log.info("==> found %s FTSJobs to monitor" % len(ftsJobs))
# # PHASE 0 = monitor active FTSJobs
for ftsJob in ftsJobs:
monitor = self.__monitorJob(request, ftsJob)
if not monitor["OK"]:
log.error("unable to monitor FTSJob %s: %s" %
(ftsJob.FTSJobID, monitor["Message"]))
ftsJob.Status = "Submitted"
continue
ftsFilesDict = self.updateFTSFileDict(ftsFilesDict,
monitor["Value"])
log.info("monitoring of FTSJobs completed")
for key, ftsFiles in ftsFilesDict.items():
if ftsFiles:
log.debug(" => %s FTSFiles to %s" %
(len(ftsFiles), key[2:].lower()))
# # PHASE ONE - check ready replicas
missingReplicas = self.__checkReadyReplicas(request, operation)
if not missingReplicas["OK"]:
log.error(missingReplicas["Message"])
else:
missingReplicas = missingReplicas["Value"]
for opFile in operation:
# Actually the condition below should never happen... Change printout for checking
if opFile.LFN not in missingReplicas and opFile.Status != 'Done':
log.warn("Should be set! %s is replicated at all targets" %
opFile.LFN)
opFile.Status = "Done"
toFail = ftsFilesDict.get("toFail", [])
toReschedule = ftsFilesDict.get("toReschedule", [])
toSubmit = ftsFilesDict.get("toSubmit", [])
toRegister = ftsFilesDict.get("toRegister", [])
toUpdate = ftsFilesDict.get("toUpdate", [])
# # PHASE TWO = Failed files? -> make request Failed and return
if toFail:
log.error(
"==> found %s 'Failed' FTSFiles, request execution cannot proceed..."
% len(toFail))
for opFile in operation:
for ftsFile in toFail:
if opFile.FileID == ftsFile.FileID:
opFile.Error = ftsFile.Error
opFile.Status = "Failed"
operation.Error = "%s files are missing any replicas" % len(toFail)
# # requets.Status should be Failed at this stage "Failed"
if request.Status == "Failed":
request.Error = "ReplicateAndRegister %s failed" % operation.Order
log.error("request is set to 'Failed'")
return self.putRequest(request)
# # PHASE THREE - update Waiting#SourceSE FTSFiles
if toUpdate:
log.info("==> found %s possible FTSFiles to update..." %
(len(toUpdate)))
byTarget = {}
for ftsFile in toUpdate:
if ftsFile.TargetSE not in byTarget:
byTarget.setdefault(ftsFile.TargetSE, [])
byTarget[ftsFile.TargetSE].append(ftsFile.FileID)
for targetSE, fileIDList in byTarget.items():
update = self.ftsClient().setFTSFilesWaiting(
operation.OperationID, targetSE, fileIDList)
if not update["OK"]:
log.error("update FTSFiles failed: %s" % update["Message"])
continue
# # PHASE FOUR - add 'RegisterReplica' Operations
if toRegister:
log.info(
"==> found %s Files waiting for registration, adding 'RegisterReplica' operations"
)
registerFiles = self.__register(request, operation, toRegister)
if not registerFiles["OK"]:
log.error("unable to create 'RegisterReplica' operations: %s" %
registerFiles["Message"])
if request.Status == "Waiting":
log.info(
"request is in 'Waiting' state, will put it back to RMS")
return self.putRequest(request)
# # PHASE FIVE - reschedule operation files
if toReschedule:
log.info("==> found %s Files to reschedule" % len(toReschedule))
rescheduleFiles = self.__reschedule(request, operation,
toReschedule)
if not rescheduleFiles["OK"]:
log.error(rescheduleFiles["Message"])
if request.Status == "Waiting":
log.info(
"request is in 'Waiting' state, will put it back to ReqDB")
return self.putRequest(request)
# # PHASE SIX - read Waiting ftsFiles and submit new FTSJobs
ftsFiles = self.ftsClient().getFTSFilesForRequest(
request.RequestID, ["Waiting"])
if not ftsFiles["OK"]:
log.error(ftsFiles["Message"])
else:
retryIds = list(set([ftsFile.FTSFileID for ftsFile in toSubmit]))
for ftsFile in ftsFiles["Value"]:
if ftsFile.FTSFileID not in retryIds:
toSubmit.append(ftsFile)
retryIds.append(ftsFile.FTSFileID)
# # submit new ftsJobs
if operation.Status == "Scheduled" and toSubmit:
log.info("==> found %s FTSFiles to submit" % len(toSubmit))
submit = self.__submit(request, operation, toSubmit)
if not submit["OK"]:
log.error(submit["Message"])
else:
ftsJobs += submit["Value"]
# # status change? - put back request
if request.Status != "Scheduled":
put = self.putRequest(request)
if not put["OK"]:
log.error("unable to put back request: %s" % put["Message"])
return put
# # put back jobs
if ftsJobs:
putJobs = self.putFTSJobs(ftsJobs)
if not putJobs["OK"]:
log.error("unable to put back FTSJobs: %s" %
putJobs["Message"])
return putJobs
return S_OK()
def __reschedule(self, request, operation, toReschedule):
""" reschedule list of :toReschedule: files in request for operation :operation:
:param Request request:
:param Operation operation:
:param list toReschedule: list of FTSFiles
"""
log = self.log.getSubLogger("%s/reschedule" % request.RequestName)
log.info("found %s files to reschedule" % len(toReschedule))
for opFile in operation:
for ftsFile in toReschedule:
if opFile.FileID == ftsFile.FileID:
opFile.Status = "Waiting"
toSchedule = []
# # filter files
for opFile in operation.getWaitingFilesList():
replicas = self.__filterReplicas(opFile)
if not replicas["OK"]:
continue
replicas = replicas["Value"]
if not replicas["Valid"] and replicas["Banned"]:
log.warn(
"unable to schedule '%s', replicas only at banned SEs" %
opFile.LFN)
continue
validReplicas = replicas["Valid"]
bannedReplicas = replicas["Banned"]
if not validReplicas and bannedReplicas:
log.warn(
"unable to schedule '%s', replicas only at banned SEs" %
opFile.LFN)
continue
if validReplicas:
validTargets = list(
set(operation.targetSEList) - set(validReplicas))
if not validTargets:
log.info("file %s is already present at all targets" %
opFile.LFN)
opFile.Status = "Done"
continue
toSchedule.append(
(opFile.toJSON()["Value"], validReplicas, validTargets))
# # do real schedule here
if toSchedule:
ftsSchedule = self.ftsClient().ftsSchedule(request.RequestID,
operation.OperationID,
toSchedule)
if not ftsSchedule["OK"]:
self.log.error(ftsSchedule["Message"])
return ftsSchedule
ftsSchedule = ftsSchedule["Value"]
for fileID in ftsSchedule["Successful"]:
for opFile in operation:
if fileID == opFile.FileID:
opFile.Status = "Scheduled"
for fileID, reason in ftsSchedule["Failed"]:
for opFile in operation:
if fileID == opFile.FileID:
opFile.Error = reason
return S_OK()
def __submit(self, request, operation, toSubmit):
""" create and submit new FTSJobs using list of FTSFiles
:param Request request: ReqDB.Request instance
:param list ftsFiles: list of FTSFile instances
:return: [ FTSJob, FTSJob, ...]
"""
log = self.log.getSubLogger("%s/submit" % request.RequestName)
bySourceAndTarget = {}
for ftsFile in toSubmit:
if ftsFile.SourceSE not in bySourceAndTarget:
bySourceAndTarget.setdefault(ftsFile.SourceSE, {})
if ftsFile.TargetSE not in bySourceAndTarget[ftsFile.SourceSE]:
bySourceAndTarget[ftsFile.SourceSE].setdefault(
ftsFile.TargetSE, [])
bySourceAndTarget[ftsFile.SourceSE][ftsFile.TargetSE].append(
ftsFile)
ftsJobs = []
for source, targetDict in bySourceAndTarget.items():
for target, ftsFileList in targetDict.items():
log.info("found %s files to submit from %s to %s" %
(len(ftsFileList), source, target))
route = self.__ftsGraph.findRoute(source, target)
if not route["OK"]:
log.error(route["Message"])
continue
route = route["Value"]
sourceRead = route.fromNode.SEs[source]["read"]
if not sourceRead:
log.error("SourceSE %s is banned for reading right now" %
source)
continue
targetWrite = route.toNode.SEs[target]["write"]
if not targetWrite:
log.error("TargetSE %s is banned for writing right now" %
target)
continue
if route.ActiveJobs > route.toNode.MaxActiveJobs:
log.warn(
"unable to submit new FTS job, max active jobs reached"
)
continue
# # create FTSJob
ftsJob = FTSJob()
ftsJob.RequestID = request.RequestID
ftsJob.OperationID = operation.OperationID
ftsJob.SourceSE = source
ftsJob.TargetSE = target
sourceSE = self.getSE(source)
sourceToken = sourceSE.getStorageParameters("SRM2")
if not sourceToken["OK"]:
log.error("unable to get sourceSE '%s' parameters: %s" %
(source, sourceToken["Message"]))
continue
ftsJob.SourceToken = sourceToken["Value"].get("SpaceToken", "")
targetSE = self.getSE(target)
targetToken = targetSE.getStorageParameters("SRM2")
if not targetToken["OK"]:
log.error("unable to get targetSE '%s' parameters: %s" %
(target, targetToken["Message"]))
continue
ftsJob.TargetToken = targetToken["Value"].get("SpaceToken", "")
ftsJob.FTSServer = route.toNode.FTSServer
for ftsFile in ftsFileList:
ftsFile.Attempt += 1
ftsFile.Error = ""
ftsJob.addFile(ftsFile)
submit = ftsJob.submitFTS2(self.STAGE_FILES)
if not submit["OK"]:
log.error("unable to submit FTSJob: %s" %
submit["Message"])
continue
log.info("FTSJob '%s'@'%s' has been submitted" %
(ftsJob.FTSGUID, ftsJob.FTSServer))
# # update statuses for job files
for ftsFile in ftsJob:
ftsFile.FTSGUID = ftsJob.FTSGUID
ftsFile.Status = "Submitted"
ftsFile.Attempt += 1
# # update graph route
try:
self.updateLock().acquire()
route.ActiveJobs += 1
finally:
self.updateLock().release()
ftsJobs.append(ftsJob)
log.info("%s new FTSJobs have been submitted" % len(ftsJobs))
return S_OK(ftsJobs)
def __monitorJob(self, request, ftsJob):
""" execute FTSJob.monitorFTS2 for a given :ftsJob:
if ftsJob is in a final state, finalize it
:param Request request: ReqDB.Request instance
:param FTSJob ftsJob: FTSDB.FTSJob instance
"""
log = self.log.getSubLogger("%s/monitor/%s" %
(request.RequestName, ftsJob.FTSGUID))
log.info("FTSJob '%s'@'%s'" % (ftsJob.FTSGUID, ftsJob.FTSServer))
# # this will be returned
ftsFilesDict = dict([(k, list())
for k in ("toRegister", "toSubmit", "toFail",
"toReschedule", "toUpdate")])
monitor = ftsJob.monitorFTS2()
if not monitor["OK"]:
gMonitor.addMark("FTSMonitorFail", 1)
log.error(monitor["Message"])
if "getTransferJobSummary2: Not authorised to query request" in monitor[
"Message"]:
log.error("FTSJob not known (expired on server?)")
for ftsFile in ftsJob:
ftsFile.Status = "Waiting"
ftsFilesDict["toSubmit"] = ftsFile
return S_OK(ftsFilesDict)
return monitor
monitor = monitor["Value"]
log.info("FTSJob Status = %s Completeness = %s" %
(ftsJob.Status, ftsJob.Completeness))
# # monitor status change
gMonitor.addMark("FTSJobs%s" % ftsJob.Status, 1)
if ftsJob.Status in FTSJob.FINALSTATES:
finalizeFTSJob = self.__finalizeFTSJob(request, ftsJob)
if not finalizeFTSJob["OK"]:
log.error(finalizeFTSJob["Message"])
return finalizeFTSJob
ftsFilesDict = self.updateFTSFileDict(ftsFilesDict,
finalizeFTSJob["Value"])
return S_OK(ftsFilesDict)
def __finalizeFTSJob(self, request, ftsJob):
""" finalize FTSJob
:param Request request: ReqDB.Request instance
:param FTSJob ftsJob: FTSDB.FTSJob instance
"""
log = self.log.getSubLogger("%s/monitor/%s/finalize" %
(request.RequestName, ftsJob.FTSJobID))
log.info("finalizing FTSJob %s@%s" %
(ftsJob.FTSGUID, ftsJob.FTSServer))
# # this will be returned
ftsFilesDict = dict([(k, list())
for k in ("toRegister", "toSubmit", "toFail",
"toReschedule", "toUpdate")])
monitor = ftsJob.monitorFTS2(full=True)
if not monitor["OK"]:
log.error(monitor["Message"])
return monitor
# # split FTSFiles to different categories
processFiles = self.__filterFiles(ftsJob)
if not processFiles["OK"]:
log.error(processFiles["Message"])
return processFiles
ftsFilesDict = self.updateFTSFileDict(ftsFilesDict,
processFiles["Value"])
# # send accounting record for this job
self.__sendAccounting(ftsJob, request.OwnerDN)
# # update graph - remove this job from graph
route = self.__ftsGraph.findRoute(ftsJob.SourceSE, ftsJob.TargetSE)
if route["OK"]:
try:
self.updateLock().acquire()
route["Value"].ActiveJobs -= 1
finally:
self.updateLock().release()
log.info("FTSJob is finalized")
return S_OK(ftsFilesDict)
def __filterFiles(self, ftsJob):
""" process ftsFiles from finished ftsJob
:param FTSJob ftsJob: monitored FTSJob instance
"""
# # lists for different categories
toUpdate = []
toReschedule = []
toRegister = []
toSubmit = []
toFail = []
# # loop over files in fts job
for ftsFile in ftsJob:
# # successful files
if ftsFile.Status == "Finished":
if ftsFile.Error == "AddCatalogReplicaFailed":
toRegister.append(ftsFile)
toUpdate.append(ftsFile)
continue
if ftsFile.Status == "Failed":
if ftsFile.Error == "MissingSource":
toReschedule.append(ftsFile)
else:
if ftsFile.Attempt < self.MAX_ATTEMPT:
toSubmit.append(ftsFile)
else:
toFail.append(ftsFile)
ftsFile.Error = "Max attempts reached"
return S_OK({
"toUpdate": toUpdate,
"toSubmit": toSubmit,
"toRegister": toRegister,
"toReschedule": toReschedule,
"toFail": toFail
})
def __register(self, request, operation, toRegister):
""" add RegisterReplica operation
:param Request request: request instance
:param Operation transferOp: 'ReplicateAndRegister' operation for this FTSJob
:param list toRegister: [ FTSDB.FTSFile, ... ] - files that failed to register
"""
log = self.log.getSubLogger("%s/registerFiles" % request.RequestName)
byTarget = {}
for ftsFile in toRegister:
if ftsFile.TargetSE not in byTarget:
byTarget.setdefault(ftsFile.TargetSE, [])
byTarget[ftsFile.TargetSE].append(ftsFile)
log.info("will create %s 'RegisterReplica' operations" % len(byTarget))
for target, ftsFileList in byTarget.items():
log.info(
"creating 'RegisterReplica' operation for targetSE %s with %s files..."
% (target, len(ftsFileList)))
registerOperation = Operation()
registerOperation.Type = "RegisterReplica"
registerOperation.Status = "Waiting"
registerOperation.TargetSE = target
targetSE = self.getSE(target)
for ftsFile in ftsFileList:
opFile = File()
opFile.LFN = ftsFile.LFN
pfn = targetSE.getPfnForProtocol(ftsFile.TargetSURL,
"SRM2",
withPort=False)
if not pfn["OK"]:
continue
opFile.PFN = pfn["Value"]
registerOperation.addFile(opFile)
request.insertBefore(registerOperation, operation)
return S_OK()
@staticmethod
def __sendAccounting(ftsJob, ownerDN):
""" prepare and send DataOperation to AccouringDB """
dataOp = DataOperation()
dataOp.setStartTime(fromString(ftsJob.SubmitTime))
dataOp.setEndTime(fromString(ftsJob.LastUpdate))
accountingDict = dict()
accountingDict["OperationType"] = "ReplicateAndRegister"
username = getUsernameForDN(ownerDN)
if not username["OK"]:
username = ownerDN
else:
username = username["Value"]
accountingDict["User"] = username
accountingDict[
"Protocol"] = "FTS3" if 'fts3' in ftsJob.FTSServer.lower(
) else 'FTS'
accountingDict['ExecutionSite'] = ftsJob.FTSServer
accountingDict['RegistrationTime'] = ftsJob._regTime
accountingDict['RegistrationOK'] = ftsJob._regSuccess
accountingDict['RegistrationTotal'] = ftsJob._regTotal
accountingDict["TransferOK"] = len(
[f for f in ftsJob if f.Status in FTSFile.SUCCESS_STATES])
accountingDict["TransferTotal"] = len(ftsJob)
accountingDict["TransferSize"] = ftsJob.Size - ftsJob.FailedSize
accountingDict["FinalStatus"] = ftsJob.Status
accountingDict["Source"] = ftsJob.SourceSE
accountingDict["Destination"] = ftsJob.TargetSE
dt = ftsJob.LastUpdate - ftsJob.SubmitTime
transferTime = dt.days * 86400 + dt.seconds
accountingDict["TransferTime"] = transferTime
# accountingDict['TransferTime'] = sum( [f._duration for f in ftsJob])
dataOp.setValuesFromDict(accountingDict)
dataOp.commit()
def __checkReadyReplicas(self, request, operation):
""" check ready replicas for transferOperation """
log = self.log.getSubLogger("%s/checkReadyReplicas" %
request.RequestName)
targetSESet = set(operation.targetSEList)
# # { LFN: [ targetSE, ... ] }
missingReplicas = {}
scheduledFiles = dict([(opFile.LFN, opFile) for opFile in operation
if opFile.Status in ("Scheduled", "Waiting")])
# # get replicas
replicas = self.replicaManager().getCatalogReplicas(
scheduledFiles.keys())
if not replicas["OK"]:
self.log.error(replicas["Message"])
return replicas
replicas = replicas["Value"]
fullyReplicated = 0
missingSEs = {}
for successfulLFN in replicas["Successful"]:
reps = set(replicas['Successful'][successfulLFN])
if targetSESet.issubset(reps):
log.info("%s has been replicated to all targets" %
successfulLFN)
fullyReplicated += 1
scheduledFiles[successfulLFN].Status = "Done"
else:
missingReplicas[successfulLFN] = sorted(targetSESet - reps)
ses = ",".join(missingReplicas[successfulLFN])
missingSEs[ses] = missingSEs.setdefault(ses, 0) + 1
log.verbose("%s is still missing at %s" % (successfulLFN, ses))
if fullyReplicated:
log.info("%d new files have been replicated to all targets" %
fullyReplicated)
if missingSEs:
for ses in missingSEs:
log.info("%d replicas still missing at %s" %
(missingSEs[ses], ses))
reMissing = re.compile("no such file or directory")
for failedLFN, errStr in replicas["Failed"].items():
scheduledFiles[failedLFN].Error = errStr
if reMissing.search(errStr.lower()):
log.error("%s is missing, setting its status to 'Failed'" %
failedLFN)
scheduledFiles[failedLFN].Status = "Failed"
else:
log.warn("unable to read replicas for %s: %s" %
(failedLFN, errStr))
return S_OK(missingReplicas)
def __filterReplicas(self, opFile):
""" filter out banned/invalid source SEs """
log = self.log.getSubLogger("filterReplicas")
ret = {"Valid": [], "Banned": [], "Bad": []}
replicas = self.replicaManager().getActiveReplicas(opFile.LFN)
if not replicas["OK"]:
log.error(replicas["Message"])
reNotExists = re.compile("not such file or directory")
replicas = replicas["Value"]
failed = replicas["Failed"].get(opFile.LFN, "")
if reNotExists.match(failed.lower()):
opFile.Status = "Failed"
opFile.Error = failed
return S_ERROR(failed)
replicas = replicas["Successful"][
opFile.LFN] if opFile.LFN in replicas["Successful"] else {}
for repSEName in replicas:
repSE = self.getSE(repSEName)
pfn = repSE.getPfnForLfn(opFile.LFN)
if not pfn["OK"]:
log.warn("unable to create pfn for %s lfn: %s" %
(opFile.LFN, pfn["Message"]))
ret["Banned"].append(repSEName)
continue
pfn = pfn["Value"]
repSEMetadata = repSE.getFileMetadata(pfn, singleFile=True)
if not repSEMetadata["OK"]:
self.log.warn(repSEMetadata["Message"])
ret["Banned"].append(repSEName)
continue
repSEMetadata = repSEMetadata["Value"]
seChecksum = repSEMetadata["Checksum"].replace(
"x", "0").zfill(8) if "Checksum" in repSEMetadata else None
if opFile.Checksum and opFile.Checksum != seChecksum:
self.log.warn(" %s checksum mismatch: %s %s:%s" %
(opFile.LFN, opFile.Checksum, repSE, seChecksum))
ret["Bad"].append(repSEName)
continue
# # if we're here repSE is OK
ret["Valid"].append(repSEName)
return S_OK(ret)
class FTSStrategy( object ):
"""
.. class:: FTSStrategy
helper class to create replication forest for a given file and it's replicas using
several different strategies
"""
# # make it singleton
__metaclass__ = DIRACSingleton
# # list of supported strategies
__supportedStrategies = [ 'Simple', 'DynamicThroughput', 'Swarm', 'MinimiseTotalWait' ]
# # FTS graph
__ftsGraph = None
# # lock
__graphLock = None
# # resources
__resources = None
# # rss client
__rssClient = None
# # acceptable failure rate
acceptableFailureRate = 75
# # acceptable failed files
acceptableFailedFiles = 5
# # scheduling type
schedulingType = "Files"
def __init__( self, csPath = None, ftsSites = None, ftsHistoryViews = None ):
"""c'tor
:param self: self reference
:param str csPath: CS path
:param list ftsSites: list of FTSSites
:param list ftsHistoryViews: list of FTSHistoryViews
"""
# ## config path
self.csPath = csPath
# # fts sites
ftsSites = ftsSites if ftsSites else []
# # history views
ftsHistoryViews = ftsHistoryViews if ftsHistoryViews else []
# # own sub logger
self.log = gLogger.getSubLogger( "FTSStrategy", child = True )
self.log.setLevel( gConfig.getValue( self.csPath + "/LogLevel", "DEBUG" ) )
for ftsSite in ftsSites:
self.log.info( "FTSSite: %-16s FTSServer=%s" % ( ftsSite.Name, ftsSite.FTSServer ) )
# # CS options
self.log.info( "Supported strategies = %s" % ", ".join( self.supportedStrategies ) )
self.activeStrategies = gConfig.getValue( "%s/%s" % ( self.csPath, "ActiveStrategies" ), ["MinimiseTotalWait"] )
self.log.info( "ActiveStrategies = %s" % ", ".join( self.activeStrategies ) )
self.numberOfStrategies = len( self.activeStrategies )
self.log.info( "Number of active strategies = %s" % self.numberOfStrategies )
self.sigma = gConfig.getValue( "%s/%s" % ( self.csPath, "HopSigma" ), 5 )
self.log.info( "HopSigma = %s" % self.sigma )
self.schedulingType = gConfig.getValue( "%s/%s" % ( self.csPath, "SchedulingType" ), "Files" )
self.log.info( "SchedulingType = %s" % self.schedulingType )
self.acceptableFailureRate = gConfig.getValue( "%s/%s" % ( self.csPath, "AcceptableFailureRate" ), 75 )
self.log.info( "AcceptableFailureRate = %s" % self.acceptableFailureRate )
self.acceptableFailedFiles = gConfig.getValue( "%s/%s" % ( self.csPath, "AcceptableFailedFiles" ), 5 )
self.log.info( "AcceptableFailedFiles = %s" % self.acceptableFailedFiles )
# # chosen strategy
self.chosenStrategy = 0
# dispatcher
self.strategyDispatcher = { "MinimiseTotalWait" : self.minimiseTotalWait,
"DynamicThroughput" : self.dynamicThroughput,
"Simple" : self.simple,
"Swarm" : self.swarm }
self.ftsGraph = FTSGraph( "FTSGraph",
ftsHistoryViews,
self.acceptableFailureRate,
self.acceptableFailedFiles,
self.schedulingType )
# for node in self.ftsGraph.nodes():
# self.log.debug( node )
# for edge in self.ftsGraph.edges():
# self.log.debug( edge )
# # if we land here everything is OK
self.log.info( "%s has been constructed" % self.__class__.__name__ )
@classmethod
def graphLock( cls ):
""" get graph lock """
if not cls.__graphLock:
cls.__graphLock = LockRing().getLock( "FTSGraphLock" )
return cls.__graphLock
@classmethod
def resetGraph( cls, ftsHistoryViews ):
""" reset graph
:param list ftsHistoryViews: list of FTSHistoryViews
"""
ftsGraph = None
try:
cls.graphLock().acquire()
ftsGraph = FTSGraph( "FTSGraph",
ftsHistoryViews,
cls.acceptableFailureRate,
cls.acceptableFailedFiles,
cls.schedulingType )
if ftsGraph:
cls.ftsGraph = ftsGraph
finally:
cls.graphLock().release()
return S_OK()
def updateRWAccess( self ):
""" update RW access in FTS graph """
updateRWAccess = S_OK()
try:
self.graphLock().acquire()
updateRWAccess = self.ftsGraph.updateRWAccess()
if not updateRWAccess["OK"]:
self.log.error( updateRWAccess["Message"] )
finally:
self.graphLock().release()
return updateRWAccess
def addTreeToGraph( self, replicationTree = None, size = 0.0 ):
""" update rw access for nodes (sites) and size anf files for edges (channels) """
replicationTree = replicationTree if replicationTree else {}
size = size if size else 0.0
if replicationTree:
try:
self.graphLock().acquire()
for route in self.ftsGraph.edges():
if route.routeName in replicationTree:
route.WaitingSize += size
route.WaitingFiles += 1
finally:
self.graphLock().release()
return S_OK()
def simple( self, sourceSEs, targetSEs ):
""" simple strategy - one source, many targets
:param list sourceSEs: list with only one sourceSE name
:param list targetSEs: list with target SE names
:param str lfn: logical file name
"""
# # make targetSEs list unique
if len( sourceSEs ) != 1:
return S_ERROR( "simple: wrong argument supplied for sourceSEs, only one sourceSE allowed" )
sourceSE = sourceSEs[0]
tree = {}
for targetSE in targetSEs:
route = self.ftsGraph.findRoute( sourceSE, targetSE )
if not route["OK"]:
return S_ERROR( route["Message"] )
route = route["Value"]
if not route.fromNode.SEs[sourceSE]["read"]:
return S_ERROR( "simple: sourceSE '%s' in banned for reading right now" % sourceSE )
if not route.toNode.SEs[targetSE]["write"]:
return S_ERROR( "simple: targetSE '%s' is banned for writing right now" % targetSE )
if route.name in tree:
return S_ERROR( "simple: unable to create replication tree, route '%s' cannot be used twice" % \
route.name )
tree[route.name] = { "Ancestor" : False, "SourceSE" : sourceSE,
"TargetSE" : targetSE, "Strategy" : "Simple" }
return S_OK( tree )
def swarm( self, sourceSEs, targetSEs ):
""" swarm strategy - one target, many sources, pick up the fastest
:param list sourceSEs: list of source SE
:param str targetSEs: on element list with name of target SE
:param str lfn: logical file name
"""
tree = {}
routes = []
if len( targetSEs ) > 1:
return S_ERROR( "swarm: wrong argument supplied for targetSEs, only one targetSE allowed" )
targetSE = targetSEs[0]
# # find channels
for sourceSE in sourceSEs:
route = self.ftsGraph.findRoute( sourceSE, targetSE )
if not route["OK"]:
self.log.warn( "swarm: %s" % route["Message"] )
continue
routes.append( ( sourceSE, route["Value"] ) )
# # exit - no channels
if not routes:
return S_ERROR( "swarm: unable to find FTS routes between '%s' and '%s'" % ( ",".join( sourceSEs ), targetSE ) )
# # filter out non active channels
routes = [ ( sourceSE, route ) for sourceSE, route in routes
if route.fromNode.SEs[sourceSE]["read"] and route.toNode.SEs[targetSE]["write"] and
route.timeToStart < float( "inf" ) ]
# # exit - no active channels
if not routes:
return S_ERROR( "swarm: no active routes found between %s and %s" % ( sourceSEs, targetSE ) )
# # find min timeToStart
minTimeToStart = float( "inf" )
selSourceSE = selRoute = None
for sourceSE, route in routes:
if route.timeToStart < minTimeToStart:
minTimeToStart = route.timeToStart
selSourceSE = sourceSE
selRoute = route
if not selSourceSE:
return S_ERROR( "swarm: no active routes found between %s and %s" % ( sourceSEs, targetSE ) )
tree[selRoute.name] = { "Ancestor" : False, "SourceSE" : selSourceSE,
"TargetSE" : targetSE, "Strategy" : "Swarm" }
return S_OK( tree )
def minimiseTotalWait( self, sourceSEs, targetSEs ):
""" find dag minimizing start time
:param list sourceSEs: list of avialable source SEs
:param list targetSEs: list of target SEs
:param str lfn: logical file name
"""
tree = {}
primarySources = sourceSEs
while targetSEs:
minTimeToStart = float( "inf" )
channels = []
self.log.verbose( "minimiseTotalWait: searching routes between %s and %s" % ( ",".join( sourceSEs ),
",".join( targetSEs ) ) )
for targetSE in targetSEs:
for sourceSE in sourceSEs:
ftsChannel = self.ftsGraph.findRoute( sourceSE, targetSE )
if not ftsChannel["OK"]:
self.log.warn( "minimiseTotalWait: %s" % ftsChannel["Message"] )
continue
ftsChannel = ftsChannel["Value"]
channels.append( ( ftsChannel, sourceSE, targetSE ) )
if not channels:
msg = "minimiseTotalWait: FTS route between these SEs are not defined"
self.log.error( msg )
return S_ERROR( msg )
self.log.verbose( "minimiseTotalWait: found %s candidate routes, checking RSS status" % len( channels ) )
for ch, s, t in channels:
self.log.verbose( "%s %s %s" % ( ch.routeName, ch.fromNode.SEs[s]["read"], ch.toNode.SEs[t]["write"] ) )
channels = [ ( channel, sourceSE, targetSE ) for channel, sourceSE, targetSE in channels
if channel.fromNode.SEs[sourceSE]["read"] and channel.toNode.SEs[targetSE]["write"]
and channel.timeToStart < float( "inf" ) ]
if not channels:
self.log.error( "minimiseTotalWait: no active FTS routes found" )
return S_ERROR( "minimiseTotalWait: no active FTS routes found" )
candidates = []
for channel, sourceSE, targetSE in channels:
timeToStart = channel.timeToStart
if sourceSE not in primarySources:
timeToStart += self.sigma
# # local found
if channel.fromNode == channel.toNode:
self.log.debug( "minimiseTotalWait: found local route '%s'" % channel.routeName )
candidates = [ ( channel, sourceSE, targetSE ) ]
break
if timeToStart <= minTimeToStart:
minTimeToStart = timeToStart
candidates = [ ( channel, sourceSE, targetSE ) ]
elif timeToStart == minTimeToStart:
candidates.append( ( channel, sourceSE, targetSE ) )
if not candidates:
return S_ERROR( "minimiseTotalWait: unable to find candidate FTS routes minimizing total wait time" )
random.shuffle( candidates )
selChannel, selSourceSE, selTargetSE = candidates[0]
ancestor = False
for routeName, treeItem in tree.items():
if selSourceSE in treeItem["TargetSE"]:
ancestor = treeItem["TargetSE"]
tree[selChannel.routeName] = { "Ancestor" : ancestor, "SourceSE" : selSourceSE,
"TargetSE" : selTargetSE, "Strategy" : "MinimiseTotalWait" }
sourceSEs.append( selTargetSE )
targetSEs.remove( selTargetSE )
return S_OK( tree )
def dynamicThroughput( self, sourceSEs, targetSEs ):
""" dynamic throughput - many sources, many targets - find dag minimizing overall throughput
:param list sourceSEs: list of available source SE names
:param list targetSE: list of target SE names
:param str lfn: logical file name
"""
tree = {}
primarySources = sourceSEs
timeToSite = {}
while targetSEs:
minTimeToStart = float( "inf" )
channels = []
for targetSE in targetSEs:
for sourceSE in sourceSEs:
ftsChannel = self.ftsGraph.findRoute( sourceSE, targetSE )
if not ftsChannel["OK"]:
self.log.warn( "dynamicThroughput: %s" % ftsChannel["Message"] )
continue
ftsChannel = ftsChannel["Value"]
channels.append( ( ftsChannel, sourceSE, targetSE ) )
# # no candidate channels found
if not channels:
msg = "dynamicThroughput: FTS routes between %s and %s are not defined" % ( ",".join( sourceSEs ),
",".join( targetSEs ) )
self.log.error( msg )
return S_ERROR( msg )
# # filter out already used channels
channels = [ ( channel, sourceSE, targetSE ) for channel, sourceSE, targetSE in channels
if channel.routeName not in tree ]
if not channels:
msg = "dynamicThroughput: all FTS routes between %s and %s are already used in tree" % ( ",".join( sourceSEs ),
",".join( targetSEs ) )
self.log.error( msg )
return S_ERROR( msg )
# # filter out non-active channels
self.log.debug( "dynamicThroughput: found %s candidate routes, checking activity" % len( channels ) )
channels = [ ( channel, sourceSE, targetSE ) for channel, sourceSE, targetSE in channels
if channel.fromNode.SEs[sourceSE]["read"] and channel.toNode.SEs[targetSE]["write"]
and channel.timeToStart < float( "inf" ) ]
if not channels:
self.log.warn( "dynamicThroughput: active candidate routes not found" )
return S_ERROR( "dynamicThroughput: no active candidate FTS routes found" )
candidates = []
selTimeToStart = None
for channel, sourceSE, targetSE in channels:
timeToStart = channel.timeToStart
if sourceSE not in primarySources:
timeToStart += self.sigma
if sourceSE in timeToSite:
timeToStart += timeToSite[sourceSE]
# # local found
if channel.fromNode == channel.toNode:
self.log.debug( "dynamicThroughput: found local route '%s'" % channel.channelName )
candidates = [ ( channel, sourceSE, targetSE ) ]
selTimeToStart = timeToStart
break
if timeToStart <= minTimeToStart:
selTimeToStart = timeToStart
minTimeToStart = timeToStart
candidates = [ ( channel, sourceSE, targetSE ) ]
elif timeToStart == minTimeToStart:
candidates.append( ( channel, sourceSE, targetSE ) )
if not candidates:
return S_ERROR( "dynamicThroughput: unable to find candidate FTS routes" )
random.shuffle( candidates )
selChannel, selSourceSE, selTargetSE = candidates[0]
ancestor = False
for routeName, treeItem in tree.items():
if selSourceSE in treeItem["TargetSE"]:
ancestor = treeItem["TargetSE"]
tree[selChannel.name] = { "Ancestor": ancestor, "SourceSE": selSourceSE,
"TargetSE": selTargetSE, "Strategy": "DynamicThroughput" }
timeToSite[selTargetSE] = selTimeToStart
sourceSEs.append( selTargetSE )
targetSEs.remove( selTargetSE )
return S_OK( tree )
def reset( self ):
""" reset :chosenStrategy:
:param self: self reference
"""
self.chosenStrategy = 0
@property
def supportedStrategies( self ):
""" Get supported strategies.
:param self: self reference
"""
return self.__supportedStrategies
def replicationTree( self, sourceSEs, targetSEs, size, strategy = None ):
""" get replication tree
:param str lfn: LFN
:param list sourceSEs: list of sources SE names to use
:param list targetSEs: list of target SE names to use
:param long size: file size
:param str strategy: strategy name
"""
# # get strategy
strategy = strategy if strategy else self.__selectStrategy()
if strategy not in self.activeStrategies:
return S_ERROR( "replicationTree: inactive or unsupported strategy '%s'" % strategy )
self.log.verbose( "replicationTree: strategy=%s sourceSEs=%s targetSEs=%s size=%s" % \
( strategy, sourceSEs, targetSEs, size ) )
# # fire action from dispatcher
tree = self.strategyDispatcher[strategy]( sourceSEs, targetSEs )
if not tree["OK"]:
self.log.error( "replicationTree: %s" % tree["Message"] )
return tree
# # update graph edges
self.log.verbose( "replicationTree: %s" % tree["Value"] )
update = self.addTreeToGraph( replicationTree = tree["Value"], size = size )
if not update["OK"]:
self.log.error( "replicationTree: unable to update FTS graph: %s" % update["Message"] )
return update
return tree
def __selectStrategy( self ):
""" If more than one active strategy use one after the other.
:param self: self reference
"""
chosenStrategy = self.activeStrategies[self.chosenStrategy]
self.chosenStrategy += 1
if self.chosenStrategy == self.numberOfStrategies:
self.chosenStrategy = 0
return chosenStrategy
class FTSAgent( AgentModule ):
"""
.. class:: FTSAgent
Agent propagating Scheduled request to Done or Failed state in the FTS system.
Requests and associated FTSJobs (and so FTSFiles) are kept in cache.
"""
# # fts graph refresh in seconds
FTSGRAPH_REFRESH = FTSHistoryView.INTERVAL / 2
# # SE R/W access refresh in seconds
RW_REFRESH = 600
# # placeholder for max job per channel
MAX_ACTIVE_JOBS = 50
# # min threads
MIN_THREADS = 1
# # max threads
MAX_THREADS = 10
# # files per job
MAX_FILES_PER_JOB = 100
# # MAX FTS transfer per FTSFile
MAX_ATTEMPT = 256
# # stage flag
PIN_TIME = 0
# # FTS submission command
SUBMIT_COMMAND = 'glite-transfer-submit'
# # FTS monitoring command
MONITOR_COMMAND = 'glite-transfer-status'
# # placeholder for FTS client
__ftsClient = None
# # placeholder for request client
__requestClient = None
# # placeholder for resources helper
__resources = None
# # placeholder for RSS client
__rssClient = None
# # placeholder for FTSGraph
__ftsGraph = None
# # graph regeneration time delta
__ftsGraphValidStamp = None
# # r/w access valid stamp
__rwAccessValidStamp = None
# # placeholder for threadPool
__threadPool = None
# # update lock
__updateLock = None
# # se cache
__seCache = dict()
# # request cache
__reqCache = dict()
def updateLock( self ):
""" update lock """
if not self.__updateLock:
self.__updateLock = LockRing().getLock( "FTSAgentLock" )
return self.__updateLock
@classmethod
def requestClient( cls ):
""" request client getter """
if not cls.__requestClient:
cls.__requestClient = ReqClient()
return cls.__requestClient
@classmethod
def ftsClient( cls ):
""" FTS client """
if not cls.__ftsClient:
cls.__ftsClient = FTSClient()
return cls.__ftsClient
@classmethod
def rssClient( cls ):
""" RSS client getter """
if not cls.__rssClient:
cls.__rssClient = ResourceStatus()
return cls.__rssClient
@classmethod
def getSE( cls, seName ):
""" keep SEs in cache """
if seName not in cls.__seCache:
cls.__seCache[seName] = StorageElement( seName )
return cls.__seCache[seName]
@classmethod
def getSECache( cls ):
return cls.__seCache
@classmethod
def getRequest( cls, reqName ):
""" get Requests systematically and refresh cache """
getRequest = cls.requestClient().getRequest( reqName )
if not getRequest["OK"]:
cls.__reqCache.pop( reqName, None )
return getRequest
getRequest = getRequest["Value"]
if not getRequest:
cls.__reqCache.pop( reqName, None )
return S_ERROR( "request of name '%s' not found in ReqDB" % reqName )
cls.__reqCache[reqName] = getRequest
return S_OK( cls.__reqCache[reqName] )
@classmethod
def putRequest( cls, request, clearCache = True ):
""" put request back to ReqDB
:param Request request: Request instance
:param bool clearCache: clear the cache?
also finalize request if status == Done
"""
# # put back request
if request.RequestName not in cls.__reqCache:
return S_OK()
put = cls.requestClient().putRequest( request )
if not put["OK"]:
return put
# # finalize first if possible
if request.Status == "Done" and request.JobID:
finalizeRequest = cls.requestClient().finalizeRequest( request.RequestName, request.JobID )
if not finalizeRequest["OK"]:
request.Status = "Scheduled"
# # del request from cache if needed
if clearCache:
cls.__reqCache.pop( request.RequestName, None )
return S_OK()
@classmethod
def putFTSJobs( cls, ftsJobsList ):
""" put back fts jobs to the FTSDB """
for ftsJob in ftsJobsList:
put = cls.ftsClient().putFTSJob( ftsJob )
if not put["OK"]:
return put
return S_OK()
@staticmethod
def updateFTSFileDict( ftsFilesDict, toUpdateDict ):
""" update :ftsFilesDict: with FTSFiles in :toUpdateDict: """
for category, ftsFileList in ftsFilesDict.items():
for ftsFile in toUpdateDict.get( category, [] ):
if ftsFile not in ftsFileList:
ftsFileList.append( ftsFile )
return ftsFilesDict
# def resources( self ):
# """ resource helper getter """
# if not self.__resources:
# self.__resources = Resources()
# return self.__resources
def threadPool( self ):
""" thread pool getter """
if not self.__threadPool:
self.__threadPool = ThreadPool( self.MIN_THREADS, self.MAX_THREADS )
self.__threadPool.daemonize()
return self.__threadPool
def resetFTSGraph( self ):
""" create fts graph """
log = gLogger.getSubLogger( "ftsGraph" )
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error( "unable to get FTS history: %s" % ftsHistory["Message"] )
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph( "FTSGraph", ftsHistory )
finally:
self.updateLock().release()
log.debug( "FTSSites: %s" % len( self.__ftsGraph.nodes() ) )
for i, site in enumerate( self.__ftsGraph.nodes() ):
log.debug( " [%02d] FTSSite: %-25s FTSServer: %s" % ( i, site.name, site.FTSServer ) )
log.debug( "FTSRoutes: %s" % len( self.__ftsGraph.edges() ) )
for i, route in enumerate( self.__ftsGraph.edges() ):
log.debug( " [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" % ( i,
route.routeName,
route.ActiveJobs,
route.toNode.MaxActiveJobs ) )
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.RW_REFRESH )
return S_OK()
def initialize( self ):
""" agent's initialization """
# # data manager
self.dataManager = DataManager()
log = self.log.getSubLogger( "initialize" )
self.FTSGRAPH_REFRESH = self.am_getOption( "FTSGraphValidityPeriod", self.FTSGRAPH_REFRESH )
log.info( "FTSGraph validity period = %s s" % self.FTSGRAPH_REFRESH )
self.RW_REFRESH = self.am_getOption( "RWAccessValidityPeriod", self.RW_REFRESH )
log.info( "SEs R/W access validity period = %s s" % self.RW_REFRESH )
self.SUBMIT_COMMAND = self.am_getOption( "SubmitCommand", self.SUBMIT_COMMAND )
log.info( "FTS submit command = %s" % self.SUBMIT_COMMAND )
self.MONITOR_COMMAND = self.am_getOption( "MonitorCommand", self.MONITOR_COMMAND )
log.info( "FTS commands: submit = %s monitor %s" % ( self.SUBMIT_COMMAND, self.MONITOR_COMMAND ) )
self.PIN_TIME = self.am_getOption( "PinTime", self.PIN_TIME )
log.info( "Stage files before submission = %s" % {True: "yes", False: "no"}[bool( self.PIN_TIME )] )
self.MAX_ACTIVE_JOBS = self.am_getOption( "MaxActiveJobsPerRoute", self.MAX_ACTIVE_JOBS )
log.info( "Max active FTSJobs/route = %s" % self.MAX_ACTIVE_JOBS )
self.MAX_FILES_PER_JOB = self.am_getOption( "MaxFilesPerJob", self.MAX_FILES_PER_JOB )
log.info( "Max FTSFiles/FTSJob = %d" % self.MAX_FILES_PER_JOB )
self.MAX_ATTEMPT = self.am_getOption( "MaxTransferAttempts", self.MAX_ATTEMPT )
log.info( "Max transfer attempts = %s" % self.MAX_ATTEMPT )
# # thread pool
self.MIN_THREADS = self.am_getOption( "MinThreads", self.MIN_THREADS )
self.MAX_THREADS = self.am_getOption( "MaxThreads", self.MAX_THREADS )
minmax = ( abs( self.MIN_THREADS ), abs( self.MAX_THREADS ) )
self.MIN_THREADS, self.MAX_THREADS = min( minmax ), max( minmax )
log.info( "ThreadPool min threads = %s" % self.MIN_THREADS )
log.info( "ThreadPool max threads = %s" % self.MAX_THREADS )
log.info( "initialize: creation of FTSGraph..." )
createGraph = self.resetFTSGraph()
if not createGraph["OK"]:
log.error( "initialize: %s" % createGraph["Message"] )
return createGraph
# This sets the Default Proxy to used as that defined under
# /Operations/Shifter/DataManager
# the shifterProxy option in the Configuration can be used to change this default.
self.am_setOption( 'shifterProxy', 'DataManager' )
log.info( "will use DataManager proxy" )
# # gMonitor stuff here
gMonitor.registerActivity( "RequestsAtt", "Attempted requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsOK", "Successful requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsFail", "Failed requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubAtt", "FTSJobs creation attempts",
"FTSAgent", "Created FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubOK", "FTSJobs submitted successfully",
"FTSAgent", "Successful FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubFail", "FTSJobs submissions failed",
"FTSAgent", "Failed FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonAtt", "FTSJobs monitored",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonOK", "FTSJobs monitored successfully",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonFail", "FTSJobs attempts failed",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSMonitorFail", "Failed FTS monitor executions",
"FTSAgent", "Execution/mins", gMonitor.OP_SUM )
pollingTime = self.am_getOption( "PollingTime", 60 )
for status in list( FTSJob.INITSTATES + FTSJob.TRANSSTATES + FTSJob.FAILEDSTATES + FTSJob.FINALSTATES ):
gMonitor.registerActivity( "FTSJobs%s" % status, "FTSJobs %s" % status ,
"FTSAgent", "FTSJobs/cycle", gMonitor.OP_ACUM, pollingTime )
gMonitor.registerActivity( "FtSJobsPerRequest", "Average FTSJobs per request",
"FTSAgent", "FTSJobs/Request", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSFilesPerJob", "FTSFiles per FTSJob",
"FTSAgent", "Number of FTSFiles per FTSJob", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSSizePerJob", "Average FTSFiles size per FTSJob",
"FTSAgent", "Average submitted size per FTSJob", gMonitor.OP_MEAN )
return S_OK()
def finalize( self ):
""" finalize processing """
# log = self.log.getSubLogger( "finalize" )
# if self.__reqCache:
# log.info( 'putting back %d requests from cache' % len( self.__reqCache ) )
# else:
# log.info( 'no requests to put back' )
# for request in self.__reqCache.values():
# put = self.requestClient().putRequest( request )
# if not put["OK"]:
# log.error( "unable to put back request '%s': %s" % ( request.RequestName, put["Message"] ) )
return S_OK()
def execute( self ):
""" one cycle execution """
log = gLogger.getSubLogger( "execute" )
# # reset FTSGraph if expired
now = datetime.datetime.now()
if now > self.__ftsGraphValidStamp:
log.info( "resetting expired FTS graph..." )
resetFTSGraph = self.resetFTSGraph()
if not resetFTSGraph["OK"]:
log.error( "FTSGraph recreation error: %s" % resetFTSGraph["Message"] )
return resetFTSGraph
self.__ftsGraphValidStamp = now + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # update R/W access in FTSGraph if expired
if now > self.__rwAccessValidStamp:
log.info( "updating expired R/W access for SEs..." )
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
self.__rwAccessValidStamp = now + datetime.timedelta( seconds = self.RW_REFRESH )
requestNames = self.requestClient().getRequestNamesList( [ "Scheduled" ] )
if not requestNames["OK"]:
log.error( "unable to read scheduled request names: %s" % requestNames["Message"] )
return requestNames
if not requestNames["Value"]:
requestNames = self.__reqCache.keys()
else:
requestNames = [ req[0] for req in requestNames["Value"] ]
requestNames = list( set ( requestNames + self.__reqCache.keys() ) )
if not requestNames:
log.info( "no 'Scheduled' requests to process" )
return S_OK()
log.info( "found %s requests to process:" % len( requestNames ) )
log.info( " => from internal cache: %s" % ( len( self.__reqCache ) ) )
log.info( " => new read from RMS: %s" % ( len( requestNames ) - len( self.__reqCache ) ) )
for requestName in requestNames:
request = self.getRequest( requestName )
if not request["OK"]:
log.error( request["Message"] )
continue
request = request["Value"]
sTJId = request.RequestName
while True:
queue = self.threadPool().generateJobAndQueueIt( self.processRequest,
args = ( request, ),
sTJId = sTJId )
if queue["OK"]:
log.info( "request '%s' enqueued for execution" % sTJId )
gMonitor.addMark( "RequestsAtt", 1 )
break
time.sleep( 1 )
# # process all results
self.threadPool().processAllResults()
return S_OK()
def processRequest( self, request ):
""" process one request
:param Request request: ReqDB.Request
"""
log = self.log.getSubLogger( request.RequestName )
operation = request.getWaiting()
if not operation["OK"]:
log.error( "unable to find 'Scheduled' ReplicateAndRegister operation in request" )
return self.putRequest( request )
operation = operation["Value"]
if not isinstance( operation, Operation ):
log.error( "waiting returned operation is not an operation: %s" % type( operation ) )
return self.putRequest( request )
if operation.Type != "ReplicateAndRegister":
log.error( "operation to be executed is not a ReplicateAndRegister but %s" % operation.Type )
return self.putRequest( request )
if operation.Status != "Scheduled":
log.error( "operation in a wrong state, expecting 'Scheduled', got %s" % operation.Status )
return self.putRequest( request )
log.info( 'start processRequest' )
# # select FTSJobs, by default all in TRANS_STATES and INIT_STATES
ftsJobs = self.ftsClient().getFTSJobsForRequest( request.RequestID )
if not ftsJobs["OK"]:
log.error( ftsJobs["Message"] )
return ftsJobs
ftsJobs = [ftsJob for ftsJob in ftsJobs.get( "Value", [] ) if ftsJob.Status not in FTSJob.FINALSTATES]
# # Use a try: finally: for making sure FTS jobs are put back before returnin
try:
# # dict keeping info about files to reschedule, submit, fail and register
ftsFilesDict = dict( [ ( k, list() ) for k in ( "toRegister", "toSubmit", "toFail", "toReschedule", "toUpdate" ) ] )
if ftsJobs:
log.info( "==> found %s FTSJobs to monitor" % len( ftsJobs ) )
# # PHASE 0 = monitor active FTSJobs
for ftsJob in ftsJobs:
monitor = self.__monitorJob( request, ftsJob )
if not monitor["OK"]:
log.error( "unable to monitor FTSJob %s: %s" % ( ftsJob.FTSJobID, monitor["Message"] ) )
ftsJob.Status = "Submitted"
else:
ftsFilesDict = self.updateFTSFileDict( ftsFilesDict, monitor["Value"] )
log.info( "monitoring of FTSJobs completed" )
for key, ftsFiles in ftsFilesDict.items():
if ftsFiles:
log.debug( " => %s FTSFiles to %s" % ( len( ftsFiles ), key[2:].lower() ) )
# # PHASE ONE - check ready replicas
missingReplicas = self.__checkReadyReplicas( request, operation )
if not missingReplicas["OK"]:
log.error( missingReplicas["Message"] )
else:
missingReplicas = missingReplicas["Value"]
for opFile in operation:
# Actually the condition below should never happen... Change printout for checking
if opFile.LFN not in missingReplicas and opFile.Status not in ( 'Done', 'Failed' ):
log.warn( "File should be set Done! %s is replicated at all targets" % opFile.LFN )
opFile.Status = "Done"
if missingReplicas:
# Check if these files are in the FTSDB
ftsFiles = self.ftsClient().getAllFTSFilesForRequest( request.RequestID )
if not ftsFiles['OK']:
log.error( ftsFiles['Message'] )
else:
ftsFiles = ftsFiles['Value']
ftsLfns = set( [ftsFile.LFN for ftsFile in ftsFiles] )
# Recover files not in FTSDB
toSchedule = set( missingReplicas ) - ftsLfns
if toSchedule:
log.warn( '%d files in operation are not in FTSDB, reset them Waiting' % len( toSchedule ) )
for opFile in operation:
if opFile.LFN in toSchedule and opFile.Status == 'Scheduled':
opFile.Status = 'Waiting'
# Recover files with target not in FTSDB
toSchedule = set( [missing for missing, missingSEs in missingReplicas.items()
if not [ftsFile for ftsFile in ftsFiles
if ftsFile.LFN == missing and ftsFile.TargetSE in missingSEs]] )
if toSchedule:
log.warn( '%d targets in operation are not in FTSDB, reset files Waiting' % len( toSchedule ) )
for opFile in operation:
if opFile.LFN in toSchedule and opFile.Status == 'Scheduled':
opFile.Status = 'Waiting'
# identify missing LFNs that are waiting for a replication which is finished
for ftsFile in [f for f in ftsFiles if f.LFN in missingReplicas and f.Status.startswith( 'Waiting#' )]:
targetSE = ftsFile.Status.split( '#' )[1]
finishedFiles = [f for f in ftsFiles if
f.LFN == ftsFile.LFN and
f.Status == 'Finished' and
f.TargetSE == targetSE and
f not in ftsFilesDict['toUpdate']]
if finishedFiles:
log.warn( "%s is %s while replication was Finished to %s, update" % ( ftsFile.LFN, ftsFile.Status, targetSE ) )
ftsFilesDict['toUpdate'] += finishedFiles
# identify Finished transfer for which the replica is still missing
for ftsFile in [f for f in ftsFiles if f.Status == 'Finished' and f.TargetSE in missingReplicas.get( f.LFN, [] ) and f not in ftsFilesDict['toRegister'] ]:
# Check if there is a registration operation for that file and that target
regOp = [op for op in request if
op.Type == 'RegisterReplica' and
op.TargetSE == ftsFile.TargetSE and
[f for f in op if f.LFN == ftsFile.LFN]]
if not regOp:
ftsFilesDict['toReschedule'].append( ftsFile )
toFail = ftsFilesDict.get( "toFail", [] )
toReschedule = ftsFilesDict.get( "toReschedule", [] )
toSubmit = ftsFilesDict.get( "toSubmit", [] )
toRegister = ftsFilesDict.get( "toRegister", [] )
toUpdate = ftsFilesDict.get( "toUpdate", [] )
# # PHASE TWO = Failed files? -> make request Failed and return
if toFail:
log.error( "==> found %s 'Failed' FTSFiles, but maybe other files can be processed..." % len( toFail ) )
for opFile in operation:
for ftsFile in toFail:
if opFile.FileID == ftsFile.FileID:
opFile.Error = ftsFile.Error
opFile.Status = "Failed"
operation.Error = "%s files are missing any replicas" % len( toFail )
# # requets.Status should be Failed if all files in the operation "Failed"
if request.Status == "Failed":
request.Error = "ReplicateAndRegister %s failed" % operation.Order
log.error( "request is set to 'Failed'" )
# # putRequest is done by the finally: clause... Not good to do it twice
raise escapeTry
# # PHASE THREE - update Waiting#TargetSE FTSFiles
if toUpdate:
log.info( "==> found %s possible FTSFiles to update..." % ( len( toUpdate ) ) )
byTarget = {}
for ftsFile in toUpdate:
byTarget.setdefault( ftsFile.TargetSE, [] ).append( ftsFile.FileID )
for targetSE, fileIDList in byTarget.items():
update = self.ftsClient().setFTSFilesWaiting( operation.OperationID, targetSE, fileIDList )
if not update["OK"]:
log.error( "update FTSFiles failed: %s" % update["Message"] )
# # PHASE FOUR - add 'RegisterReplica' Operations
if toRegister:
log.info( "==> found %d Files waiting for registration, adding 'RegisterReplica' operations" % len( toRegister ) )
registerFiles = self.__insertRegisterOperation( request, operation, toRegister )
if not registerFiles["OK"]:
log.error( "unable to create 'RegisterReplica' operations: %s" % registerFiles["Message"] )
# if request.Status == "Waiting":
# log.info( "request is in 'Waiting' state, will put it back to RMS" )
# return self.putRequest( request )
# # PHASE FIVE - reschedule operation files
if toReschedule:
log.info( "==> found %s Files to reschedule" % len( toReschedule ) )
rescheduleFiles = self.__reschedule( request, operation, toReschedule )
if not rescheduleFiles["OK"]:
log.error( rescheduleFiles["Message"] )
# # PHASE SIX - read Waiting ftsFiles and submit new FTSJobs. We get also Failed files to recover them if needed
ftsFiles = self.ftsClient().getFTSFilesForRequest( request.RequestID, [ "Waiting", "Failed", 'Submitted', 'Canceled' ] )
if not ftsFiles["OK"]:
log.error( ftsFiles["Message"] )
else:
retryIds = set ( [ ftsFile.FTSFileID for ftsFile in toSubmit ] )
for ftsFile in ftsFiles["Value"]:
if ftsFile.FTSFileID not in retryIds:
if ftsFile.Status in ( 'Failed', 'Canceled' ):
# If the file was not unrecoverable failed and is not yet set toSubmit
_reschedule, submit, _fail = self.__checkFailed( ftsFile )
elif ftsFile.Status == 'Submitted':
if ftsFile.FTSGUID not in [job.FTSGUID for job in ftsJobs]:
log.warn( 'FTS GUID %s not found in FTS jobs, resubmit file transfer' % ftsFile.FTSGUID )
ftsFile.Status = 'Waiting'
submit = True
else:
submit = False
else:
submit = True
if submit:
toSubmit.append( ftsFile )
retryIds.add( ftsFile.FTSFileID )
# # submit new ftsJobs
if toSubmit:
if request.Status != 'Scheduled':
log.info( "Found %d FTSFiles to submit while request is no longer in Scheduled status (%s)" \
% ( len( toSubmit ), request.Status ) )
else:
self.__checkDuplicates( request.RequestName, toSubmit )
log.info( "==> found %s FTSFiles to submit" % len( toSubmit ) )
submit = self.__submit( request, operation, toSubmit )
if not submit["OK"]:
log.error( submit["Message"] )
else:
ftsJobs += submit["Value"]
# # status change? - put back request
if request.Status != "Scheduled":
log.info( "request no longer in 'Scheduled' state (%s), will put it back to RMS" % request.Status )
except escapeTry:
# This clause is raised when one wants to return from within the try: clause
pass
except Exception, exceptMessage:
log.exception( "Exception in processRequest", lException = exceptMessage )
finally:
class FTSAgent( AgentModule ):
"""
.. class:: FTSAgent
Agent propagating Scheduled request to Done or Failed state in the FTS system.
Requests and associated FTSJobs (and so FTSFiles) are kept in cache.
"""
# # fts graph refresh in seconds
FTSGRAPH_REFRESH = FTSHistoryView.INTERVAL / 2
# # SE R/W access refresh in seconds
RW_REFRESH = 600
# # placeholder for max job per channel
MAX_ACTIVE_JOBS = 50
# # min threads
MIN_THREADS = 1
# # max threads
MAX_THREADS = 10
# # files per job
MAX_FILES_PER_JOB = 100
# # MAX FTS transfer per FTSFile
MAX_ATTEMPT = 256
# # stage flag
PIN_TIME = 0
# # FTS submission command
SUBMIT_COMMAND = 'glite-transfer-submit'
# # FTS monitoring command
MONITOR_COMMAND = 'glite-transfer-status'
# # placeholder for FTS client
__ftsClient = None
# # placeholder for request client
__requestClient = None
# # placeholder for resources helper
__resources = None
# # placeholder for RSS client
__rssClient = None
# # placeholder for FTSGraph
__ftsGraph = None
# # graph regeneration time delta
__ftsGraphValidStamp = None
# # r/w access valid stamp
__rwAccessValidStamp = None
# # placeholder for threadPool
__threadPool = None
# # update lock
__updateLock = None
# # se cache
__seCache = dict()
# # request cache
__reqCache = dict()
def updateLock( self ):
""" update lock """
if not self.__updateLock:
self.__updateLock = LockRing().getLock( "FTSAgentLock" )
return self.__updateLock
@classmethod
def requestClient( cls ):
""" request client getter """
if not cls.__requestClient:
cls.__requestClient = ReqClient()
return cls.__requestClient
@classmethod
def ftsClient( cls ):
""" FTS client """
if not cls.__ftsClient:
cls.__ftsClient = FTSClient()
return cls.__ftsClient
@classmethod
def rssClient( cls ):
""" RSS client getter """
if not cls.__rssClient:
cls.__rssClient = ResourceStatus()
return cls.__rssClient
@classmethod
def getSE( cls, seName ):
""" keep SEs in cache """
if seName not in cls.__seCache:
cls.__seCache[seName] = StorageElement( seName )
return cls.__seCache[seName]
@classmethod
def getSECache( cls ):
return cls.__seCache
@classmethod
def getRequest( cls, reqName ):
""" get Requests systematically and refresh cache """
getRequest = cls.requestClient().getRequest( reqName )
if not getRequest["OK"]:
cls.__reqCache.pop( reqName, None )
return getRequest
getRequest = getRequest["Value"]
if not getRequest:
cls.__reqCache.pop( reqName, None )
return S_ERROR( "request of name '%s' not found in ReqDB" % reqName )
cls.__reqCache[reqName] = getRequest
return S_OK( cls.__reqCache[reqName] )
@classmethod
def putRequest( cls, request, clearCache = True ):
""" put request back to ReqDB
:param Request request: Request instance
:param bool clearCache: clear the cache?
also finalize request if status == Done
"""
# # put back request
if request.RequestName not in cls.__reqCache:
return S_OK()
put = cls.requestClient().putRequest( request )
if not put["OK"]:
return put
# # finalize first if possible
if request.Status == "Done" and request.JobID:
finalizeRequest = cls.requestClient().finalizeRequest( request.RequestName, request.JobID )
if not finalizeRequest["OK"]:
request.Status = "Scheduled"
# # del request from cache if needed
if clearCache:
cls.__reqCache.pop( request.RequestName, None )
return S_OK()
@classmethod
def putFTSJobs( cls, ftsJobsList ):
""" put back fts jobs to the FTSDB """
for ftsJob in ftsJobsList:
put = cls.ftsClient().putFTSJob( ftsJob )
if not put["OK"]:
return put
return S_OK()
@staticmethod
def updateFTSFileDict( ftsFilesDict, toUpdateDict ):
""" update :ftsFilesDict: with FTSFiles in :toUpdateDict: """
for category, ftsFileList in ftsFilesDict.items():
for ftsFile in toUpdateDict.get( category, [] ):
if ftsFile not in ftsFileList:
ftsFileList.append( ftsFile )
return ftsFilesDict
# def resources( self ):
# """ resource helper getter """
# if not self.__resources:
# self.__resources = Resources()
# return self.__resources
def threadPool( self ):
""" thread pool getter """
if not self.__threadPool:
self.__threadPool = ThreadPool( self.MIN_THREADS, self.MAX_THREADS )
self.__threadPool.daemonize()
return self.__threadPool
def resetFTSGraph( self ):
""" create fts graph """
log = gLogger.getSubLogger( "ftsGraph" )
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error( "unable to get FTS history:", ftsHistory["Message"] )
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph( "FTSGraph", ftsHistory, maxActiveJobs = self.MAX_ACTIVE_JOBS )
finally:
self.updateLock().release()
log.debug( "FTSSites:", len( self.__ftsGraph.nodes() ) )
for i, site in enumerate( self.__ftsGraph.nodes() ):
log.debug( " [%02d] FTSSite: %-25s FTSServer: %s" % ( i, site.name, site.FTSServer ) )
log.debug( "FTSRoutes: %s" % len( self.__ftsGraph.edges() ) )
for i, route in enumerate( self.__ftsGraph.edges() ):
log.debug( " [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" % ( i,
route.routeName,
route.ActiveJobs,
route.toNode.MaxActiveJobs ) )
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.RW_REFRESH )
return S_OK()
def initialize( self ):
""" agent's initialization """
# # data manager
self.dataManager = DataManager()
log = self.log.getSubLogger( "initialize" )
self.FTSGRAPH_REFRESH = self.am_getOption( "FTSGraphValidityPeriod", self.FTSGRAPH_REFRESH )
log.info( "FTSGraph validity period = %s s" % self.FTSGRAPH_REFRESH )
self.RW_REFRESH = self.am_getOption( "RWAccessValidityPeriod", self.RW_REFRESH )
log.info( "SEs R/W access validity period = %s s" % self.RW_REFRESH )
self.SUBMIT_COMMAND = self.am_getOption( "SubmitCommand", self.SUBMIT_COMMAND )
log.info( "FTS submit command = %s" % self.SUBMIT_COMMAND )
self.MONITOR_COMMAND = self.am_getOption( "MonitorCommand", self.MONITOR_COMMAND )
log.info( "FTS commands: submit = %s monitor %s" % ( self.SUBMIT_COMMAND, self.MONITOR_COMMAND ) )
self.PIN_TIME = self.am_getOption( "PinTime", self.PIN_TIME )
log.info( "Stage files before submission = ", {True: "yes", False: "no"}[bool( self.PIN_TIME )] )
self.MAX_ACTIVE_JOBS = self.am_getOption( "MaxActiveJobsPerRoute", self.MAX_ACTIVE_JOBS )
log.info( "Max active FTSJobs/route = ", str( self.MAX_ACTIVE_JOBS ) )
self.MAX_FILES_PER_JOB = self.am_getOption( "MaxFilesPerJob", self.MAX_FILES_PER_JOB )
log.info( "Max FTSFiles/FTSJob = ", str( self.MAX_FILES_PER_JOB ) )
self.MAX_ATTEMPT = self.am_getOption( "MaxTransferAttempts", self.MAX_ATTEMPT )
log.info( "Max transfer attempts = ", str( self.MAX_ATTEMPT ) )
# # thread pool
self.MIN_THREADS = self.am_getOption( "MinThreads", self.MIN_THREADS )
self.MAX_THREADS = self.am_getOption( "MaxThreads", self.MAX_THREADS )
minmax = ( abs( self.MIN_THREADS ), abs( self.MAX_THREADS ) )
self.MIN_THREADS, self.MAX_THREADS = min( minmax ), max( minmax )
log.info( "ThreadPool min threads = ", str( self.MIN_THREADS ) )
log.info( "ThreadPool max threads = ", str( self.MAX_THREADS ) )
log.info( "initialize: creation of FTSGraph..." )
createGraph = self.resetFTSGraph()
if not createGraph["OK"]:
log.error( "initialize: ", createGraph["Message"] )
return createGraph
# This sets the Default Proxy to used as that defined under
# /Operations/Shifter/DataManager
# the shifterProxy option in the Configuration can be used to change this default.
self.am_setOption( 'shifterProxy', 'DataManager' )
log.info( "will use DataManager proxy" )
# # gMonitor stuff here
gMonitor.registerActivity( "RequestsAtt", "Attempted requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsOK", "Successful requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsFail", "Failed requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubAtt", "FTSJobs creation attempts",
"FTSAgent", "Created FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubOK", "FTSJobs submitted successfully",
"FTSAgent", "Successful FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubFail", "FTSJobs submissions failed",
"FTSAgent", "Failed FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonAtt", "FTSJobs monitored",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonOK", "FTSJobs monitored successfully",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonFail", "FTSJobs attempts failed",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSMonitorFail", "Failed FTS monitor executions",
"FTSAgent", "Execution/mins", gMonitor.OP_SUM )
pollingTime = self.am_getOption( "PollingTime", 60 )
for status in list( FTSJob.INITSTATES + FTSJob.TRANSSTATES + FTSJob.FAILEDSTATES + FTSJob.FINALSTATES ):
gMonitor.registerActivity( "FTSJobs%s" % status, "FTSJobs %s" % status ,
"FTSAgent", "FTSJobs/cycle", gMonitor.OP_ACUM, pollingTime )
gMonitor.registerActivity( "FtSJobsPerRequest", "Average FTSJobs per request",
"FTSAgent", "FTSJobs/Request", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSFilesPerJob", "FTSFiles per FTSJob",
"FTSAgent", "Number of FTSFiles per FTSJob", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSSizePerJob", "Average FTSFiles size per FTSJob",
"FTSAgent", "Average submitted size per FTSJob", gMonitor.OP_MEAN )
return S_OK()
def finalize( self ):
""" finalize processing """
# log = self.log.getSubLogger( "finalize" )
# if self.__reqCache:
# log.info( 'putting back %d requests from cache' % len( self.__reqCache ) )
# else:
# log.info( 'no requests to put back' )
# for request in self.__reqCache.values():
# put = self.requestClient().putRequest( request )
# if not put["OK"]:
# log.error( "unable to put back request '%s': %s" % ( request.RequestName, put["Message"] ) )
return S_OK()
def execute( self ):
""" one cycle execution """
log = gLogger.getSubLogger( "execute" )
# # reset FTSGraph if expired
now = datetime.datetime.now()
if now > self.__ftsGraphValidStamp:
log.info( "resetting expired FTS graph..." )
resetFTSGraph = self.resetFTSGraph()
if not resetFTSGraph["OK"]:
log.error( "FTSGraph recreation error: ", resetFTSGraph["Message"] )
return resetFTSGraph
self.__ftsGraphValidStamp = now + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # update R/W access in FTSGraph if expired
if now > self.__rwAccessValidStamp:
log.info( "updating expired R/W access for SEs..." )
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
self.__rwAccessValidStamp = now + datetime.timedelta( seconds = self.RW_REFRESH )
requestNames = self.requestClient().getRequestNamesList( [ "Scheduled" ] )
if not requestNames["OK"]:
log.error( "Unable to read scheduled request names: ", requestNames["Message"] )
return requestNames
if not requestNames["Value"]:
requestNames = self.__reqCache.keys()
else:
requestNames = [ req[0] for req in requestNames["Value"] ]
requestNames = list( set ( requestNames + self.__reqCache.keys() ) )
if not requestNames:
log.info( "No 'Scheduled' requests to process" )
return S_OK()
log.info( "Found requests to process:", str( len( requestNames ) ) )
log.info( " => from internal cache:", str( ( len( self.__reqCache ) ) ) )
log.info( " => new read from RMS:", str( ( len( requestNames ) - len( self.__reqCache ) ) ) )
for requestName in requestNames:
request = self.getRequest( requestName )
if not request["OK"]:
log.error( "Error getting request", "%s: %s" % ( requestName, request["Message"] ) )
continue
request = request["Value"]
sTJId = request.RequestName
while True:
queue = self.threadPool().generateJobAndQueueIt( self.processRequest,
args = ( request, ),
sTJId = sTJId )
if queue["OK"]:
log.info( "Request enqueued for execution", sTJId )
gMonitor.addMark( "RequestsAtt", 1 )
break
time.sleep( 1 )
# # process all results
self.threadPool().processAllResults()
return S_OK()
def processRequest( self, request ):
""" process one request
:param Request request: ReqDB.Request
"""
log = self.log.getSubLogger( request.RequestName )
operation = request.getWaiting()
if not operation["OK"]:
log.error( "Unable to find 'Scheduled' ReplicateAndRegister operation in request" )
return self.putRequest( request )
operation = operation["Value"]
if not isinstance( operation, Operation ):
log.error( "Waiting returned operation is not an operation:", type( operation ) )
return self.putRequest( request )
if operation.Type != "ReplicateAndRegister":
log.error( "operation to be executed is not a ReplicateAndRegister but", operation.Type )
return self.putRequest( request )
if operation.Status != "Scheduled":
log.error( "operation in a wrong state, expecting 'Scheduled', got", operation.Status )
return self.putRequest( request )
log.info( 'start processRequest' )
# # select FTSJobs, by default all in TRANS_STATES and INIT_STATES
ftsJobs = self.ftsClient().getFTSJobsForRequest( request.RequestID )
if not ftsJobs["OK"]:
log.error( ftsJobs["Message"] )
return ftsJobs
ftsJobs = [ftsJob for ftsJob in ftsJobs.get( "Value", [] ) if ftsJob.Status not in FTSJob.FINALSTATES]
# # Use a try: finally: for making sure FTS jobs are put back before returnin
try:
# # dict keeping info about files to reschedule, submit, fail and register
ftsFilesDict = dict( [ ( k, list() ) for k in ( "toRegister", "toSubmit", "toFail", "toReschedule", "toUpdate" ) ] )
if ftsJobs:
log.info( "==> found %s FTSJobs to monitor" % len( ftsJobs ) )
# # PHASE 0 = monitor active FTSJobs
for ftsJob in ftsJobs:
monitor = self.__monitorJob( request, ftsJob )
if not monitor["OK"]:
log.error( "unable to monitor FTSJob %s: %s" % ( ftsJob.FTSJobID, monitor["Message"] ) )
ftsJob.Status = "Submitted"
else:
ftsFilesDict = self.updateFTSFileDict( ftsFilesDict, monitor["Value"] )
log.info( "monitoring of FTSJobs completed" )
for key, ftsFiles in ftsFilesDict.items():
if ftsFiles:
log.debug( " => %s FTSFiles to %s" % ( len( ftsFiles ), key[2:].lower() ) )
# # PHASE ONE - check ready replicas
missingReplicas = self.__checkReadyReplicas( request, operation )
if not missingReplicas["OK"]:
log.error( missingReplicas["Message"] )
else:
missingReplicas = missingReplicas["Value"]
for opFile in operation:
# Actually the condition below should never happen... Change printout for checking
if opFile.LFN not in missingReplicas and opFile.Status not in ( 'Done', 'Failed' ):
log.warn( "File should be set Done! %s is replicated at all targets" % opFile.LFN )
opFile.Status = "Done"
if missingReplicas:
# Check if these files are in the FTSDB
ftsFiles = self.ftsClient().getAllFTSFilesForRequest( request.RequestID )
if not ftsFiles['OK']:
log.error( ftsFiles['Message'] )
else:
ftsFiles = ftsFiles['Value']
ftsLfns = set( [ftsFile.LFN for ftsFile in ftsFiles] )
# Recover files not in FTSDB
toSchedule = set( missingReplicas ) - ftsLfns
if toSchedule:
log.warn( '%d files in operation are not in FTSDB, reset them Waiting' % len( toSchedule ) )
for opFile in operation:
if opFile.LFN in toSchedule and opFile.Status == 'Scheduled':
opFile.Status = 'Waiting'
# Recover files with target not in FTSDB
toSchedule = set( [missing for missing, missingSEs in missingReplicas.items()
if not [ftsFile for ftsFile in ftsFiles
if ftsFile.LFN == missing and ftsFile.TargetSE in missingSEs]] )
if toSchedule:
log.warn( '%d targets in operation are not in FTSDB, reset files Waiting' % len( toSchedule ) )
for opFile in operation:
if opFile.LFN in toSchedule and opFile.Status == 'Scheduled':
opFile.Status = 'Waiting'
# identify missing LFNs that are waiting for a replication which is finished
for ftsFile in [f for f in ftsFiles if f.LFN in missingReplicas and f.Status.startswith( 'Waiting#' )]:
targetSE = ftsFile.Status.split( '#' )[1]
finishedFiles = [f for f in ftsFiles if
f.LFN == ftsFile.LFN and
f.Status == 'Finished' and
f.TargetSE == targetSE and
f not in ftsFilesDict['toUpdate']]
if finishedFiles:
log.warn( "%s is %s while replication was Finished to %s, update" % ( ftsFile.LFN, ftsFile.Status, targetSE ) )
ftsFilesDict['toUpdate'] += finishedFiles
# identify Finished transfer for which the replica is still missing
for ftsFile in [f for f in ftsFiles if f.Status == 'Finished' and f.TargetSE in missingReplicas.get( f.LFN, [] ) and f not in ftsFilesDict['toRegister'] ]:
# Check if there is a registration operation for that file and that target
regOp = [op for op in request if
op.Type == 'RegisterReplica' and
op.TargetSE == ftsFile.TargetSE and
[f for f in op if f.LFN == ftsFile.LFN]]
if not regOp:
ftsFilesDict['toReschedule'].append( ftsFile )
toFail = ftsFilesDict.get( "toFail", [] )
toReschedule = ftsFilesDict.get( "toReschedule", [] )
toSubmit = ftsFilesDict.get( "toSubmit", [] )
toRegister = ftsFilesDict.get( "toRegister", [] )
toUpdate = ftsFilesDict.get( "toUpdate", [] )
# # PHASE TWO = Failed files? -> make request Failed and return
if toFail:
log.error( "==> found %s 'Failed' FTSFiles, but maybe other files can be processed..." % len( toFail ) )
for opFile in operation:
for ftsFile in toFail:
if opFile.FileID == ftsFile.FileID:
opFile.Error = ftsFile.Error
opFile.Status = "Failed"
operation.Error = "%s files are missing any replicas" % len( toFail )
# # requets.Status should be Failed if all files in the operation "Failed"
if request.Status == "Failed":
request.Error = "ReplicateAndRegister %s failed" % operation.Order
log.error( "request is set to 'Failed'" )
# # putRequest is done by the finally: clause... Not good to do it twice
raise escapeTry
# # PHASE THREE - update Waiting#TargetSE FTSFiles
if toUpdate:
log.info( "==> found %s possible FTSFiles to update..." % ( len( toUpdate ) ) )
byTarget = {}
for ftsFile in toUpdate:
byTarget.setdefault( ftsFile.TargetSE, [] ).append( ftsFile.FileID )
for targetSE, fileIDList in byTarget.items():
update = self.ftsClient().setFTSFilesWaiting( operation.OperationID, targetSE, fileIDList )
if not update["OK"]:
log.error( "update FTSFiles failed:", update["Message"] )
# # PHASE FOUR - add 'RegisterReplica' Operations
if toRegister:
log.info( "==> found %d Files waiting for registration, adding 'RegisterReplica' operations" % len( toRegister ) )
registerFiles = self.__insertRegisterOperation( request, operation, toRegister )
if not registerFiles["OK"]:
log.error( "unable to create 'RegisterReplica' operations:", registerFiles["Message"] )
# if request.Status == "Waiting":
# log.info( "request is in 'Waiting' state, will put it back to RMS" )
# return self.putRequest( request )
# # PHASE FIVE - reschedule operation files
if toReschedule:
log.info( "==> found %s Files to reschedule" % len( toReschedule ) )
rescheduleFiles = self.__reschedule( request, operation, toReschedule )
if not rescheduleFiles["OK"]:
log.error( rescheduleFiles["Message"] )
# # PHASE SIX - read Waiting ftsFiles and submit new FTSJobs. We get also Failed files to recover them if needed
ftsFiles = self.ftsClient().getFTSFilesForRequest( request.RequestID, [ "Waiting", "Failed", 'Submitted', 'Canceled' ] )
if not ftsFiles["OK"]:
log.error( ftsFiles["Message"] )
else:
retryIds = set ( [ ftsFile.FTSFileID for ftsFile in toSubmit ] )
for ftsFile in ftsFiles["Value"]:
if ftsFile.FTSFileID not in retryIds:
if ftsFile.Status in ( 'Failed', 'Canceled' ):
# If the file was not unrecoverable failed and is not yet set toSubmit
_reschedule, submit, _fail = self.__checkFailed( ftsFile )
elif ftsFile.Status == 'Submitted':
if ftsFile.FTSGUID not in [job.FTSGUID for job in ftsJobs]:
log.warn( 'FTS GUID %s not found in FTS jobs, resubmit file transfer' % ftsFile.FTSGUID )
ftsFile.Status = 'Waiting'
submit = True
else:
submit = False
else:
submit = True
if submit:
toSubmit.append( ftsFile )
retryIds.add( ftsFile.FTSFileID )
# # submit new ftsJobs
if toSubmit:
if request.Status != 'Scheduled':
log.info( "Found %d FTSFiles to submit while request is no longer in Scheduled status (%s)" \
% ( len( toSubmit ), request.Status ) )
else:
self.__checkDuplicates( request.RequestName, toSubmit )
log.info( "==> found %s FTSFiles to submit" % len( toSubmit ) )
submit = self.__submit( request, operation, toSubmit )
if not submit["OK"]:
log.error( submit["Message"] )
else:
ftsJobs += submit["Value"]
# # status change? - put back request
if request.Status != "Scheduled":
log.info( "request no longer in 'Scheduled' state (%s), will put it back to RMS" % request.Status )
except escapeTry:
# This clause is raised when one wants to return from within the try: clause
pass
except Exception, exceptMessage:
log.exception( "Exception in processRequest", lException = exceptMessage )
finally:
class FTSAgent( AgentModule ):
"""
.. class:: FTSAgent
Agent propagating Scheduled request to Done or Failed state in the FTS system.
Requests and associated FTSJobs (and so FTSFiles) are kept in cache.
"""
# # fts graph refresh in seconds
FTSGRAPH_REFRESH = FTSHistoryView.INTERVAL / 2
# # SE R/W access refresh in seconds
RW_REFRESH = 600
# # placeholder for max job per channel
MAX_ACTIVE_JOBS = 50
# # min threads
MIN_THREADS = 1
# # max threads
MAX_THREADS = 10
# # files per job
MAX_FILES_PER_JOB = 100
# # MAX FTS transfer per FTSFile
MAX_ATTEMPT = 256
# # stage flag
STAGE_FILES = False
# # replica manager
__replicaManager = None
# # placeholder for FTS client
__ftsClient = None
# # placeholder for request client
__requestClient = None
# # placeholder for resources helper
__resources = None
# # placeholder for RSS client
__rssClient = None
# # placeholder for FTSGraph
__ftsGraph = None
# # graph regeneration time delta
__ftsGraphValidStamp = None
# # r/w access valid stamp
__rwAccessValidStamp = None
# # placeholder for threadPool
__threadPool = None
# # update lock
__updateLock = None
# # se cache
__seCache = dict()
# # request cache
__reqCache = dict()
def updateLock( self ):
""" update lock """
if not self.__updateLock:
self.__updateLock = LockRing().getLock( "FTSAgentLock" )
return self.__updateLock
@classmethod
def requestClient( cls ):
""" request client getter """
if not cls.__requestClient:
cls.__requestClient = ReqClient()
return cls.__requestClient
@classmethod
def ftsClient( cls ):
""" FTS client """
if not cls.__ftsClient:
cls.__ftsClient = FTSClient()
return cls.__ftsClient
@classmethod
def replicaManager( cls ):
""" replica manager getter """
if not cls.__replicaManager:
cls.__replicaManager = ReplicaManager()
return cls.__replicaManager
@classmethod
def rssClient( cls ):
""" RSS client getter """
if not cls.__rssClient:
cls.__rssClient = ResourceStatus()
return cls.__rssClient
@classmethod
def getSE( cls, seName ):
""" keep SEs in cache """
if seName not in cls.__seCache:
cls.__seCache[seName] = StorageElement( seName )
return cls.__seCache[seName]
@classmethod
def getRequest( cls, reqName ):
""" keep Requests in cache """
if reqName not in cls.__reqCache:
getRequest = cls.requestClient().getRequest( reqName )
if not getRequest["OK"]:
return getRequest
getRequest = getRequest["Value"]
if not getRequest:
return S_ERROR( "request of name '%s' not found in ReqDB" % reqName )
cls.__reqCache[reqName] = getRequest
return S_OK( cls.__reqCache[reqName] )
@classmethod
def putRequest( cls, request ):
""" put request back to ReqDB
:param Request request: Request instance
also finalize request if status == Done
"""
# # put back request
put = cls.requestClient().putRequest( request )
if not put["OK"]:
return put
# # finalize first is possible
if request.Status == "Done" and request.JobID:
finalizeRequest = cls.requestClient().finalizeRequest( request.RequestName, request.JobID )
if not finalizeRequest["OK"]:
request.Status = "Scheduled"
# # del request from cache
if request.RequestName in cls.__reqCache:
del cls.__reqCache[ request.RequestName ]
return S_OK()
@classmethod
def putFTSJobs( cls, ftsJobsList ):
""" put back fts jobs to the FTSDB """
for ftsJob in ftsJobsList:
put = cls.ftsClient().putFTSJob( ftsJob )
if not put["OK"]:
return put
return S_OK()
@staticmethod
def updateFTSFileDict( ftsFilesDict, toUpdateDict ):
""" update :ftsFilesDict: with FTSFiles in :toUpdateDict: """
for category, ftsFileList in ftsFilesDict.items():
for ftsFile in toUpdateDict.get( category, [] ):
if ftsFile not in ftsFileList:
ftsFileList.append( ftsFile )
return ftsFilesDict
# def resources( self ):
# """ resource helper getter """
# if not self.__resources:
# self.__resources = Resources()
# return self.__resources
def threadPool( self ):
""" thread pool getter """
if not self.__threadPool:
self.__threadPool = ThreadPool( self.MIN_THREADS, self.MAX_THREADS )
self.__threadPool.daemonize()
return self.__threadPool
def resetFTSGraph( self ):
""" create fts graph """
log = gLogger.getSubLogger( "ftsGraph" )
ftsHistory = self.ftsClient().getFTSHistory()
if not ftsHistory["OK"]:
log.error( "unable to get FTS history: %s" % ftsHistory["Message"] )
return ftsHistory
ftsHistory = ftsHistory["Value"]
try:
self.updateLock().acquire()
self.__ftsGraph = FTSGraph( "FTSGraph", ftsHistory )
finally:
self.updateLock().release()
log.debug( "FTSSites: %s" % len( self.__ftsGraph.nodes() ) )
for i, site in enumerate( self.__ftsGraph.nodes() ):
log.debug( " [%02d] FTSSite: %-25s FTSServer: %s" % ( i, site.name, site.FTSServer ) )
log.debug( "FTSRoutes: %s" % len( self.__ftsGraph.edges() ) )
for i, route in enumerate( self.__ftsGraph.edges() ):
log.debug( " [%02d] FTSRoute: %-25s Active FTSJobs (Max) = %s (%s)" % ( i,
route.routeName,
route.ActiveJobs,
route.toNode.MaxActiveJobs ) )
# # save graph stamp
self.__ftsGraphValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # refresh SE R/W access
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
# # save rw access stamp
self.__rwAccessValidStamp = datetime.datetime.now() + datetime.timedelta( seconds = self.RW_REFRESH )
return S_OK()
def initialize( self ):
""" agent's initialization """
log = self.log.getSubLogger( "initialize" )
self.FTSGRAPH_REFRESH = self.am_getOption( "FTSGraphValidityPeriod", self.FTSGRAPH_REFRESH )
log.info( "FTSGraph validity period = %s s" % self.FTSGRAPH_REFRESH )
self.RW_REFRESH = self.am_getOption( "RWAccessValidityPeriod", self.RW_REFRESH )
log.info( "SEs R/W access validity period = %s s" % self.RW_REFRESH )
self.STAGE_FILES = self.am_getOption( "StageFiles", self.STAGE_FILES )
log.info( "Stage files before submission = %s" % {True: "yes", False: "no"}[bool( self.STAGE_FILES )] )
self.MAX_ACTIVE_JOBS = self.am_getOption( "MaxActiveJobsPerRoute", self.MAX_ACTIVE_JOBS )
log.info( "Max active FTSJobs/route = %s" % self.MAX_ACTIVE_JOBS )
self.MAX_FILES_PER_JOB = self.am_getOption( "MaxFilesPerJob", self.MAX_FILES_PER_JOB )
log.info( "Max FTSFiles/FTSJob = %d" % self.MAX_FILES_PER_JOB )
self.MAX_ATTEMPT = self.am_getOption( "MaxTransferAttempts", self.MAX_ATTEMPT )
log.info( "Max transfer attempts = %s" % self.MAX_ATTEMPT )
# # thread pool
self.MIN_THREADS = self.am_getOption( "MinThreads", self.MIN_THREADS )
self.MAX_THREADS = self.am_getOption( "MaxThreads", self.MAX_THREADS )
minmax = ( abs( self.MIN_THREADS ), abs( self.MAX_THREADS ) )
self.MIN_THREADS, self.MAX_THREADS = min( minmax ), max( minmax )
log.info( "ThreadPool min threads = %s" % self.MIN_THREADS )
log.info( "ThreadPool max threads = %s" % self.MAX_THREADS )
log.info( "initialize: creation of FTSGraph..." )
createGraph = self.resetFTSGraph()
if not createGraph["OK"]:
log.error( "initialize: %s" % createGraph["Message"] )
return createGraph
# This sets the Default Proxy to used as that defined under
# /Operations/Shifter/DataManager
# the shifterProxy option in the Configuration can be used to change this default.
self.am_setOption( 'shifterProxy', 'DataManager' )
log.info( "will use DataManager proxy" )
# # gMonitor stuff here
gMonitor.registerActivity( "RequestsAtt", "Attempted requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsOK", "Successful requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "RequestsFail", "Failed requests executions",
"FTSAgent", "Requests/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubAtt", "FTSJobs creation attempts",
"FTSAgent", "Created FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubOK", "FTSJobs submitted successfully",
"FTSAgent", "Successful FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsSubFail", "FTSJobs submissions failed",
"FTSAgent", "Failed FTSJobs submissions/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonAtt", "FTSJobs monitored",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonOK", "FTSJobs monitored successfully",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSJobsMonFail", "FTSJobs attempts failed",
"FTSAgent", "FTSJobs/min", gMonitor.OP_SUM )
gMonitor.registerActivity( "FTSMonitorFail", "Failed FTS monitor executions",
"FTSAgent", "Execution/mins", gMonitor.OP_SUM )
pollingTime = self.am_getOption( "PollingTime", 60 )
for status in list( FTSJob.INITSTATES + FTSJob.TRANSSTATES + FTSJob.FAILEDSTATES + FTSJob.FINALSTATES ):
gMonitor.registerActivity( "FTSJobs%s" % status, "FTSJobs %s" % status ,
"FTSAgent", "FTSJobs/cycle", gMonitor.OP_ACUM, pollingTime )
gMonitor.registerActivity( "FtSJobsPerRequest", "Average FTSJobs per request",
"FTSAgent", "FTSJobs/Request", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSFilesPerJob", "FTSFiles per FTSJob",
"FTSAgent", "Number of FTSFiles per FTSJob", gMonitor.OP_MEAN )
gMonitor.registerActivity( "FTSSizePerJob", "Average FTSFiles size per FTSJob",
"FTSAgent", "Average submitted size per FTSJob", gMonitor.OP_MEAN )
return S_OK()
def finalize( self ):
""" finalize processing """
log = self.log.getSubLogger( "finalize" )
for request in self.__reqCache.values():
put = self.requestClient().putRequest( request )
if not put["OK"]:
log.error( "unable to put back request '%s': %s" % ( request.RequestName, put["Message"] ) )
return S_OK()
def execute( self ):
""" one cycle execution """
log = gLogger.getSubLogger( "execute" )
# # reset FTSGraph if expired
now = datetime.datetime.now()
if now > self.__ftsGraphValidStamp:
log.info( "resetting expired FTS graph..." )
resetFTSGraph = self.resetFTSGraph()
if not resetFTSGraph["OK"]:
log.error( "FTSGraph recreation error: %s" % resetFTSGraph["Message"] )
return resetFTSGraph
self.__ftsGraphValidStamp = now + datetime.timedelta( seconds = self.FTSGRAPH_REFRESH )
# # update R/W access in FTSGraph if expired
if now > self.__rwAccessValidStamp:
log.info( "updating expired R/W access for SEs..." )
try:
self.updateLock().acquire()
self.__ftsGraph.updateRWAccess()
finally:
self.updateLock().release()
self.__rwAccessValidStamp = now + datetime.timedelta( seconds = self.RW_REFRESH )
requestNames = self.requestClient().getRequestNamesList( [ "Scheduled" ] )
if not requestNames["OK"]:
log.error( "unable to read scheduled request names: %s" % requestNames["Message"] )
return requestNames
if not requestNames["Value"]:
requestNames = self.__reqCache.keys()
else:
requestNames = [ req[0] for req in requestNames["Value"] ]
requestNames = list( set ( requestNames + self.__reqCache.keys() ) )
if not requestNames:
log.info( "no 'Scheduled' requests to process" )
return S_OK()
log.info( "found %s requests to process:" % len( requestNames ) )
log.info( " => from internal cache: %s" % ( len( self.__reqCache ) ) )
log.info( " => new read from RMS: %s" % ( len( requestNames ) - len( self.__reqCache ) ) )
for requestName in requestNames:
request = self.getRequest( requestName )
if not request["OK"]:
log.error( request["Message"] )
continue
request = request["Value"]
sTJId = request.RequestName
while True:
queue = self.threadPool().generateJobAndQueueIt( self.processRequest,
args = ( request, ),
sTJId = sTJId )
if queue["OK"]:
log.info( "request '%s' enqueued for execution" % sTJId )
gMonitor.addMark( "RequestsAtt", 1 )
break
time.sleep( 1 )
# # process all results
self.threadPool().processAllResults()
return S_OK()
def processRequest( self, request ):
""" process one request
:param Request request: ReqDB.Request
"""
log = self.log.getSubLogger( request.RequestName )
operation = request.getWaiting()
if not operation["OK"]:
log.error( "unable to find 'Scheduled' ReplicateAndRegister operation in request" )
return self.putRequest( request )
operation = operation["Value"]
if operation.Type != "ReplicateAndRegister":
log.error( "operation to be executed is not a ReplicateAndRegister but %s" % operation.Type )
return self.putRequest( request )
if operation.Status != "Scheduled":
log.error( "operation in a wrong state, expecting 'Scheduled', got %s" % operation.Status )
return self.putRequest( request )
# # select FTSJobs, by default all in TRANS_STATES and INIT_STATES
ftsJobs = self.ftsClient().getFTSJobsForRequest( request.RequestID )
if not ftsJobs["OK"]:
log.error( ftsJobs["Message"] )
return ftsJobs
ftsJobs = ftsJobs["Value"] if ftsJobs["Value"] else []
# # dict keeping info about files to reschedule, submit, fail and register
ftsFilesDict = dict( [ ( k, list() ) for k in ( "toRegister", "toSubmit", "toFail", "toReschedule", "toUpdate" ) ] )
if ftsJobs:
log.info( "==> found %s FTSJobs to monitor" % len( ftsJobs ) )
# # PHASE 0 = monitor active FTSJobs
for ftsJob in ftsJobs:
monitor = self.__monitorJob( request, ftsJob )
if not monitor["OK"]:
log.error( "unable to monitor FTSJob %s: %s" % ( ftsJob.FTSJobID, monitor["Message"] ) )
ftsJob.Status = "Submitted"
continue
ftsFilesDict = self.updateFTSFileDict( ftsFilesDict, monitor["Value"] )
log.info( "monitoring of FTSJobs completed" )
for key, ftsFiles in ftsFilesDict.items():
if ftsFiles:
log.debug( " => %s FTSFiles to %s" % ( len( ftsFiles ), key[2:].lower() ) )
# # PHASE ONE - check ready replicas
missingReplicas = self.__checkReadyReplicas( request, operation )
if not missingReplicas["OK"]:
log.error( missingReplicas["Message"] )
else:
missingReplicas = missingReplicas["Value"]
for opFile in operation:
# Actually the condition below should never happen... Change printout for checking
if opFile.LFN not in missingReplicas and opFile.Status != 'Done':
log.warn( "Should be set! %s is replicated at all targets" % opFile.LFN )
opFile.Status = "Done"
toFail = ftsFilesDict.get( "toFail", [] )
toReschedule = ftsFilesDict.get( "toReschedule", [] )
toSubmit = ftsFilesDict.get( "toSubmit", [] )
toRegister = ftsFilesDict.get( "toRegister", [] )
toUpdate = ftsFilesDict.get( "toUpdate", [] )
# # PHASE TWO = Failed files? -> make request Failed and return
if toFail:
log.error( "==> found %s 'Failed' FTSFiles, request execution cannot proceed..." % len( toFail ) )
for opFile in operation:
for ftsFile in toFail:
if opFile.FileID == ftsFile.FileID:
opFile.Error = ftsFile.Error
opFile.Status = "Failed"
operation.Error = "%s files are missing any replicas" % len( toFail )
# # requets.Status should be Failed at this stage "Failed"
if request.Status == "Failed":
request.Error = "ReplicateAndRegister %s failed" % operation.Order
log.error( "request is set to 'Failed'" )
return self.putRequest( request )
# # PHASE THREE - update Waiting#SourceSE FTSFiles
if toUpdate:
log.info( "==> found %s possible FTSFiles to update..." % ( len( toUpdate ) ) )
byTarget = {}
for ftsFile in toUpdate:
if ftsFile.TargetSE not in byTarget:
byTarget.setdefault( ftsFile.TargetSE, [] )
byTarget[ftsFile.TargetSE].append( ftsFile.FileID )
for targetSE, fileIDList in byTarget.items():
update = self.ftsClient().setFTSFilesWaiting( operation.OperationID, targetSE, fileIDList )
if not update["OK"]:
log.error( "update FTSFiles failed: %s" % update["Message"] )
continue
# # PHASE FOUR - add 'RegisterReplica' Operations
if toRegister:
log.info( "==> found %s Files waiting for registration, adding 'RegisterReplica' operations" )
registerFiles = self.__register( request, operation, toRegister )
if not registerFiles["OK"]:
log.error( "unable to create 'RegisterReplica' operations: %s" % registerFiles["Message"] )
if request.Status == "Waiting":
log.info( "request is in 'Waiting' state, will put it back to RMS" )
return self.putRequest( request )
# # PHASE FIVE - reschedule operation files
if toReschedule:
log.info( "==> found %s Files to reschedule" % len( toReschedule ) )
rescheduleFiles = self.__reschedule( request, operation, toReschedule )
if not rescheduleFiles["OK"]:
log.error( rescheduleFiles["Message"] )
if request.Status == "Waiting":
log.info( "request is in 'Waiting' state, will put it back to ReqDB" )
return self.putRequest( request )
# # PHASE SIX - read Waiting ftsFiles and submit new FTSJobs
ftsFiles = self.ftsClient().getFTSFilesForRequest( request.RequestID, [ "Waiting" ] )
if not ftsFiles["OK"]:
log.error( ftsFiles["Message"] )
else:
retryIds = list( set ( [ ftsFile.FTSFileID for ftsFile in toSubmit ] ) )
for ftsFile in ftsFiles["Value"]:
if ftsFile.FTSFileID not in retryIds:
toSubmit.append( ftsFile )
retryIds.append( ftsFile.FTSFileID )
# # submit new ftsJobs
if operation.Status == "Scheduled" and toSubmit:
log.info( "==> found %s FTSFiles to submit" % len( toSubmit ) )
submit = self.__submit( request, operation, toSubmit )
if not submit["OK"]:
log.error( submit["Message"] )
else:
ftsJobs += submit["Value"]
# # status change? - put back request
if request.Status != "Scheduled":
put = self.putRequest( request )
if not put["OK"]:
log.error( "unable to put back request: %s" % put["Message"] )
return put
# # put back jobs
if ftsJobs:
putJobs = self.putFTSJobs( ftsJobs )
if not putJobs["OK"]:
log.error( "unable to put back FTSJobs: %s" % putJobs["Message"] )
return putJobs
return S_OK()
def __reschedule( self, request, operation, toReschedule ):
""" reschedule list of :toReschedule: files in request for operation :operation:
:param Request request:
:param Operation operation:
:param list toReschedule: list of FTSFiles
"""
log = self.log.getSubLogger( "%s/reschedule" % request.RequestName )
log.info( "found %s files to reschedule" % len( toReschedule ) )
for opFile in operation:
for ftsFile in toReschedule:
if opFile.FileID == ftsFile.FileID:
opFile.Status = "Waiting"
toSchedule = []
# # filter files
for opFile in operation.getWaitingFilesList():
replicas = self.__filterReplicas( opFile )
if not replicas["OK"]:
continue
replicas = replicas["Value"]
if not replicas["Valid"] and replicas["Banned"]:
log.warn( "unable to schedule '%s', replicas only at banned SEs" % opFile.LFN )
continue
validReplicas = replicas["Valid"]
bannedReplicas = replicas["Banned"]
if not validReplicas and bannedReplicas:
log.warn( "unable to schedule '%s', replicas only at banned SEs" % opFile.LFN )
continue
if validReplicas:
validTargets = list( set( operation.targetSEList ) - set( validReplicas ) )
if not validTargets:
log.info( "file %s is already present at all targets" % opFile.LFN )
opFile.Status = "Done"
continue
toSchedule.append( ( opFile.toJSON()["Value"], validReplicas, validTargets ) )
# # do real schedule here
if toSchedule:
ftsSchedule = self.ftsClient().ftsSchedule( request.RequestID,
operation.OperationID,
toSchedule )
if not ftsSchedule["OK"]:
self.log.error( ftsSchedule["Message"] )
return ftsSchedule
ftsSchedule = ftsSchedule["Value"]
for fileID in ftsSchedule["Successful"]:
for opFile in operation:
if fileID == opFile.FileID:
opFile.Status = "Scheduled"
for fileID, reason in ftsSchedule["Failed"]:
for opFile in operation:
if fileID == opFile.FileID:
opFile.Error = reason
return S_OK()
def __submit( self, request, operation, toSubmit ):
""" create and submit new FTSJobs using list of FTSFiles
:param Request request: ReqDB.Request instance
:param list ftsFiles: list of FTSFile instances
:return: [ FTSJob, FTSJob, ...]
"""
log = self.log.getSubLogger( "%s/submit" % request.RequestName )
bySourceAndTarget = {}
for ftsFile in toSubmit:
if ftsFile.SourceSE not in bySourceAndTarget:
bySourceAndTarget.setdefault( ftsFile.SourceSE, {} )
if ftsFile.TargetSE not in bySourceAndTarget[ftsFile.SourceSE]:
bySourceAndTarget[ftsFile.SourceSE].setdefault( ftsFile.TargetSE, [] )
bySourceAndTarget[ftsFile.SourceSE][ftsFile.TargetSE].append( ftsFile )
ftsJobs = []
for source, targetDict in bySourceAndTarget.items():
for target, ftsFileList in targetDict.items():
log.info( "found %s files to submit from %s to %s" % ( len( ftsFileList ), source, target ) )
route = self.__ftsGraph.findRoute( source, target )
if not route["OK"]:
log.error( route["Message"] )
continue
route = route["Value"]
sourceRead = route.fromNode.SEs[source]["read"]
if not sourceRead:
log.error( "SourceSE %s is banned for reading right now" % source )
continue
targetWrite = route.toNode.SEs[target]["write"]
if not targetWrite:
log.error( "TargetSE %s is banned for writing right now" % target )
continue
if route.ActiveJobs > route.toNode.MaxActiveJobs:
log.warn( "unable to submit new FTS job, max active jobs reached" )
continue
# # create FTSJob
ftsJob = FTSJob()
ftsJob.RequestID = request.RequestID
ftsJob.OperationID = operation.OperationID
ftsJob.SourceSE = source
ftsJob.TargetSE = target
sourceSE = self.getSE( source )
sourceToken = sourceSE.getStorageParameters( "SRM2" )
if not sourceToken["OK"]:
log.error( "unable to get sourceSE '%s' parameters: %s" % ( source, sourceToken["Message"] ) )
continue
ftsJob.SourceToken = sourceToken["Value"].get( "SpaceToken", "" )
targetSE = self.getSE( target )
targetToken = targetSE.getStorageParameters( "SRM2" )
if not targetToken["OK"]:
log.error( "unable to get targetSE '%s' parameters: %s" % ( target, targetToken["Message"] ) )
continue
ftsJob.TargetToken = targetToken["Value"].get( "SpaceToken", "" )
ftsJob.FTSServer = route.toNode.FTSServer
for ftsFile in ftsFileList:
ftsFile.Attempt += 1
ftsFile.Error = ""
ftsJob.addFile( ftsFile )
submit = ftsJob.submitFTS2( self.STAGE_FILES )
if not submit["OK"]:
log.error( "unable to submit FTSJob: %s" % submit["Message"] )
continue
log.info( "FTSJob '%s'@'%s' has been submitted" % ( ftsJob.FTSGUID, ftsJob.FTSServer ) )
# # update statuses for job files
for ftsFile in ftsJob:
ftsFile.FTSGUID = ftsJob.FTSGUID
ftsFile.Status = "Submitted"
ftsFile.Attempt += 1
# # update graph route
try:
self.updateLock().acquire()
route.ActiveJobs += 1
finally:
self.updateLock().release()
ftsJobs.append( ftsJob )
log.info( "%s new FTSJobs have been submitted" % len( ftsJobs ) )
return S_OK( ftsJobs )
def __monitorJob( self, request, ftsJob ):
""" execute FTSJob.monitorFTS2 for a given :ftsJob:
if ftsJob is in a final state, finalize it
:param Request request: ReqDB.Request instance
:param FTSJob ftsJob: FTSDB.FTSJob instance
"""
log = self.log.getSubLogger( "%s/monitor/%s" % ( request.RequestName, ftsJob.FTSGUID ) )
log.info( "FTSJob '%s'@'%s'" % ( ftsJob.FTSGUID, ftsJob.FTSServer ) )
# # this will be returned
ftsFilesDict = dict( [ ( k, list() ) for k in ( "toRegister", "toSubmit", "toFail", "toReschedule", "toUpdate" ) ] )
monitor = ftsJob.monitorFTS2()
if not monitor["OK"]:
gMonitor.addMark( "FTSMonitorFail", 1 )
log.error( monitor["Message"] )
if "getTransferJobSummary2: Not authorised to query request" in monitor["Message"]:
log.error( "FTSJob not known (expired on server?)" )
for ftsFile in ftsJob:
ftsFile.Status = "Waiting"
ftsFilesDict["toSubmit"] = ftsFile
return S_OK( ftsFilesDict )
return monitor
monitor = monitor["Value"]
log.info( "FTSJob Status = %s Completeness = %s" % ( ftsJob.Status, ftsJob.Completeness ) )
# # monitor status change
gMonitor.addMark( "FTSJobs%s" % ftsJob.Status, 1 )
if ftsJob.Status in FTSJob.FINALSTATES:
finalizeFTSJob = self.__finalizeFTSJob( request, ftsJob )
if not finalizeFTSJob["OK"]:
log.error( finalizeFTSJob["Message"] )
return finalizeFTSJob
ftsFilesDict = self.updateFTSFileDict( ftsFilesDict, finalizeFTSJob["Value"] )
return S_OK( ftsFilesDict )
def __finalizeFTSJob( self, request, ftsJob ):
""" finalize FTSJob
:param Request request: ReqDB.Request instance
:param FTSJob ftsJob: FTSDB.FTSJob instance
"""
log = self.log.getSubLogger( "%s/monitor/%s/finalize" % ( request.RequestName, ftsJob.FTSJobID ) )
log.info( "finalizing FTSJob %s@%s" % ( ftsJob.FTSGUID, ftsJob.FTSServer ) )
# # this will be returned
ftsFilesDict = dict( [ ( k, list() ) for k in ( "toRegister", "toSubmit", "toFail", "toReschedule", "toUpdate" ) ] )
monitor = ftsJob.monitorFTS2( full = True )
if not monitor["OK"]:
log.error( monitor["Message"] )
return monitor
# # split FTSFiles to different categories
processFiles = self.__filterFiles( ftsJob )
if not processFiles["OK"]:
log.error( processFiles["Message"] )
return processFiles
ftsFilesDict = self.updateFTSFileDict( ftsFilesDict, processFiles["Value"] )
# # send accounting record for this job
self.__sendAccounting( ftsJob, request.OwnerDN )
# # update graph - remove this job from graph
route = self.__ftsGraph.findRoute( ftsJob.SourceSE, ftsJob.TargetSE )
if route["OK"]:
try:
self.updateLock().acquire()
route["Value"].ActiveJobs -= 1
finally:
self.updateLock().release()
log.info( "FTSJob is finalized" )
return S_OK( ftsFilesDict )
def __filterFiles( self, ftsJob ):
""" process ftsFiles from finished ftsJob
:param FTSJob ftsJob: monitored FTSJob instance
"""
# # lists for different categories
toUpdate = []
toReschedule = []
toRegister = []
toSubmit = []
toFail = []
# # loop over files in fts job
for ftsFile in ftsJob:
# # successful files
if ftsFile.Status == "Finished":
if ftsFile.Error == "AddCatalogReplicaFailed":
toRegister.append( ftsFile )
toUpdate.append( ftsFile )
continue
if ftsFile.Status == "Failed":
if ftsFile.Error == "MissingSource":
toReschedule.append( ftsFile )
else:
if ftsFile.Attempt < self.MAX_ATTEMPT:
toSubmit.append( ftsFile )
else:
toFail.append( ftsFile )
ftsFile.Error = "Max attempts reached"
return S_OK( { "toUpdate": toUpdate,
"toSubmit": toSubmit,
"toRegister": toRegister,
"toReschedule": toReschedule,
"toFail": toFail } )
def __register( self, request, operation, toRegister ):
""" add RegisterReplica operation
:param Request request: request instance
:param Operation transferOp: 'ReplicateAndRegister' operation for this FTSJob
:param list toRegister: [ FTSDB.FTSFile, ... ] - files that failed to register
"""
log = self.log.getSubLogger( "%s/registerFiles" % request.RequestName )
byTarget = {}
for ftsFile in toRegister:
if ftsFile.TargetSE not in byTarget:
byTarget.setdefault( ftsFile.TargetSE, [] )
byTarget[ftsFile.TargetSE].append( ftsFile )
log.info( "will create %s 'RegisterReplica' operations" % len( byTarget ) )
for target, ftsFileList in byTarget.items():
log.info( "creating 'RegisterReplica' operation for targetSE %s with %s files..." % ( target,
len( ftsFileList ) ) )
registerOperation = Operation()
registerOperation.Type = "RegisterReplica"
registerOperation.Status = "Waiting"
registerOperation.TargetSE = target
targetSE = self.getSE( target )
for ftsFile in ftsFileList:
opFile = File()
opFile.LFN = ftsFile.LFN
pfn = targetSE.getPfnForProtocol( ftsFile.TargetSURL, "SRM2", withPort = False )
if not pfn["OK"]:
continue
opFile.PFN = pfn["Value"]
registerOperation.addFile( opFile )
request.insertBefore( registerOperation, operation )
return S_OK()
@staticmethod
def __sendAccounting( ftsJob, ownerDN ):
""" prepare and send DataOperation to AccouringDB """
dataOp = DataOperation()
dataOp.setStartTime( fromString( ftsJob.SubmitTime ) )
dataOp.setEndTime( fromString( ftsJob.LastUpdate ) )
accountingDict = dict()
accountingDict["OperationType"] = "ReplicateAndRegister"
username = getUsernameForDN( ownerDN )
if not username["OK"]:
username = ownerDN
else:
username = username["Value"]
accountingDict["User"] = username
accountingDict["Protocol"] = "FTS3" if 'fts3' in ftsJob.FTSServer.lower() else 'FTS'
accountingDict['ExecutionSite'] = ftsJob.FTSServer
accountingDict['RegistrationTime'] = ftsJob._regTime
accountingDict['RegistrationOK'] = ftsJob._regSuccess
accountingDict['RegistrationTotal'] = ftsJob._regTotal
accountingDict["TransferOK"] = len( [ f for f in ftsJob if f.Status in FTSFile.SUCCESS_STATES ] )
accountingDict["TransferTotal"] = len( ftsJob )
accountingDict["TransferSize"] = ftsJob.Size - ftsJob.FailedSize
accountingDict["FinalStatus"] = ftsJob.Status
accountingDict["Source"] = ftsJob.SourceSE
accountingDict["Destination"] = ftsJob.TargetSE
dt = ftsJob.LastUpdate - ftsJob.SubmitTime
transferTime = dt.days * 86400 + dt.seconds
accountingDict["TransferTime"] = transferTime
# accountingDict['TransferTime'] = sum( [f._duration for f in ftsJob])
dataOp.setValuesFromDict( accountingDict )
dataOp.commit()
def __checkReadyReplicas( self, request, operation ):
""" check ready replicas for transferOperation """
log = self.log.getSubLogger( "%s/checkReadyReplicas" % request.RequestName )
targetSESet = set( operation.targetSEList )
# # { LFN: [ targetSE, ... ] }
missingReplicas = {}
scheduledFiles = dict( [ ( opFile.LFN, opFile ) for opFile in operation
if opFile.Status in ( "Scheduled", "Waiting" ) ] )
# # get replicas
replicas = self.replicaManager().getCatalogReplicas( scheduledFiles.keys() )
if not replicas["OK"]:
self.log.error( replicas["Message"] )
return replicas
replicas = replicas["Value"]
fullyReplicated = 0
missingSEs = {}
for successfulLFN in replicas["Successful"]:
reps = set( replicas['Successful'][successfulLFN] )
if targetSESet.issubset( reps ):
log.info( "%s has been replicated to all targets" % successfulLFN )
fullyReplicated += 1
scheduledFiles[successfulLFN].Status = "Done"
else:
missingReplicas[successfulLFN] = sorted( targetSESet - reps )
ses = ",".join( missingReplicas[ successfulLFN ] )
missingSEs[ses] = missingSEs.setdefault( ses, 0 ) + 1
log.verbose( "%s is still missing at %s" % ( successfulLFN, ses ) )
if fullyReplicated:
log.info( "%d new files have been replicated to all targets" % fullyReplicated )
if missingSEs:
for ses in missingSEs:
log.info( "%d replicas still missing at %s" % ( missingSEs[ses], ses ) )
reMissing = re.compile( "no such file or directory" )
for failedLFN, errStr in replicas["Failed"].items():
scheduledFiles[failedLFN].Error = errStr
if reMissing.search( errStr.lower() ):
log.error( "%s is missing, setting its status to 'Failed'" % failedLFN )
scheduledFiles[failedLFN].Status = "Failed"
else:
log.warn( "unable to read replicas for %s: %s" % ( failedLFN, errStr ) )
return S_OK( missingReplicas )
def __filterReplicas( self, opFile ):
""" filter out banned/invalid source SEs """
log = self.log.getSubLogger( "filterReplicas" )
ret = { "Valid" : [], "Banned" : [], "Bad" : [] }
replicas = self.replicaManager().getActiveReplicas( opFile.LFN )
if not replicas["OK"]:
log.error( replicas["Message"] )
reNotExists = re.compile( "not such file or directory" )
replicas = replicas["Value"]
failed = replicas["Failed"].get( opFile.LFN , "" )
if reNotExists.match( failed.lower() ):
opFile.Status = "Failed"
opFile.Error = failed
return S_ERROR( failed )
replicas = replicas["Successful"][opFile.LFN] if opFile.LFN in replicas["Successful"] else {}
for repSEName in replicas:
repSE = self.getSE( repSEName )
pfn = repSE.getPfnForLfn( opFile.LFN )
if not pfn["OK"]:
log.warn( "unable to create pfn for %s lfn: %s" % ( opFile.LFN, pfn["Message"] ) )
ret["Banned"].append( repSEName )
continue
pfn = pfn["Value"]
repSEMetadata = repSE.getFileMetadata( pfn, singleFile = True )
if not repSEMetadata["OK"]:
self.log.warn( repSEMetadata["Message"] )
ret["Banned"].append( repSEName )
continue
repSEMetadata = repSEMetadata["Value"]
seChecksum = repSEMetadata["Checksum"].replace( "x", "0" ).zfill( 8 ) if "Checksum" in repSEMetadata else None
if opFile.Checksum and opFile.Checksum != seChecksum:
self.log.warn( " %s checksum mismatch: %s %s:%s" % ( opFile.LFN,
opFile.Checksum,
repSE,
seChecksum ) )
ret["Bad"].append( repSEName )
continue
# # if we're here repSE is OK
ret["Valid"].append( repSEName )
return S_OK( ret )