class MonitorRanger:
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
def register_inotify_path(self,path,mask):
self.inotifyWrapper.registerPath(path,mask)
def start_inotify(self):
self.inotifyWrapper.start()
def stop_inotify(self):
logging.info("MonitorRanger: Stop inotify wrapper")
self.inotifyWrapper.stop()
logging.info("MonitorRanger: Join inotify wrapper")
self.inotifyWrapper.join()
logging.info("MonitorRanger: Inotify wrapper returned")
def process_default(self, event):
self.logger.debug("event: %s on: %s" %(str(event.mask),event.fullpath))
if self.eventQueue:
self.eventQueue.put(event)
def setEventQueue(self,queue):
self.eventQueue = queue
def __init__(self, recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self, recursiveMode)
self.queueStatusPath = None
self.queueStatusPathMon = None
self.queueStatusPathDir = None
self.queuedLumiList = []
self.maxQueuedLumi = -1
#max seen/closed by anelastic thread
self.maxReceivedEoLS = -1
self.maxClosedLumi = -1
self.numOpenLumis = -1
self.lock = threading.Lock()
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
self.queueStatusPath = None
self.queueStatusPathMon = None
self.queueStatusPathDir = None
self.queuedLumiList = []
self.maxQueuedLumi=-1
#max seen/closed by anelastic thread
self.maxReceivedEoLS=-1
self.maxClosedLumi=-1
self.numOpenLumis=-1
self.maxCMSSWLumi=-1
self.maxLSWithOutput=-1
self.lock = threading.Lock()
self.output_bw=0
self.lumi_bw=0
self.data_size_ls_num = 0
self.data_size_val = 0
self.data_size_last_update = 0
self.statsCollectorThread = None
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
self.queueStatusPath = None
self.queueStatusPathMon = None
self.queueStatusPathDir = None
self.queuedLumiList = []
self.maxQueuedLumi=-1
#max seen/closed by anelastic thread
self.maxReceivedEoLS=-1
self.maxClosedLumi=-1
self.numOpenLumis=-1
self.maxCMSSWLumi=-1
self.lock = threading.Lock()
def __init__(self,confClass,stateInfo,resInfo,runList,mountMgr,boxInfo,monitor,resource_lock):
self.inotifyWrapper = InotifyWrapper(self)
self.logger = logging.getLogger(self.__class__.__name__)
self.state = stateInfo
self.resInfo = resInfo
self.runList = runList
self.managed_monitor = monitor
self.managed_monitor.preStart()
self.managed_monitor.start()
self.regpath = []
self.mm = mountMgr
self.boxInfo = boxInfo
self.resource_lock = resource_lock
self.hostname = os.uname()[1]
global conf
conf = confClass
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
class MonitorRanger:
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
self.queueStatusPath = None
self.queueStatusPathMon = None
self.queueStatusPathDir = None
self.queuedLumiList = []
self.maxQueuedLumi=-1
#max seen/closed by anelastic thread
self.maxReceivedEoLS=-1
self.maxClosedLumi=-1
self.numOpenLumis=-1
self.maxCMSSWLumi=-1
self.lock = threading.Lock()
def register_inotify_path(self,path,mask):
self.inotifyWrapper.registerPath(path,mask)
def start_inotify(self):
self.inotifyWrapper.start()
def stop_inotifyTimeout(self,timeout):
self.logger.info("MonitorRanger: Stop inotify wrapper")
self.inotifyWrapper.stop()
self.logger.info("MonitorRanger: Join inotify wrapper")
self.inotifyWrapper.join(timeout)
if self.inotifyWrapper.isAlive():
self.logger.info("MonitorRanger: Inotify wrapper join timeout ("+str(timeout)+")")
return False
else:
self.logger.info("MonitorRanger: Inotify wrapper returned")
return True
def stop_inotify(self):
self.logger.info("MonitorRanger: Stop inotify wrapper")
self.inotifyWrapper.stop()
self.logger.info("MonitorRanger: Join inotify wrapper")
self.inotifyWrapper.join()
self.logger.info("MonitorRanger: Inotify wrapper returned")
def process_default(self, event):
self.logger.debug("event: %s on: %s" %(str(event.mask),event.fullpath))
if self.eventQueue:
if self.queueStatusPath!=None:
if self.checkNewLumi(event):
self.eventQueue.put(event)
else:
self.eventQueue.put(event)
def setEventQueue(self,queue):
self.eventQueue = queue
def checkNewLumi(self,event):
if event.fullpath.endswith("_EoLS.jsn"):
try:
queuedLumi = int(os.path.basename(event.fullpath).split('_')[1][2:])
self.lock.acquire()
if queuedLumi not in self.queuedLumiList:
if queuedLumi>self.maxQueuedLumi:
self.maxQueuedLumi=queuedLumi
self.queuedLumiList.append(queuedLumi)
self.lock.release()
self.updateQueueStatusFile()
else:
self.lock.release()
#skip if EoL for LS in queue has already been written once (e.g. double file create race)
return False
except:
self.logger.warning("Problem checking new EoLS filename: "+str(os.path.basename(event.fullpath)) + " error:"+str(ex))
try:self.lock.release()
except:pass
#delete associated BoLS file
try:
os.unlink(event.fullpath[:event.fullpath.rfind("_EoLS.jsn")]+"_BoLS.jsn")
except:
pass
elif event.fullpath.endswith("_BoLS.jsn"):
try:
queuedLumi = int(os.path.basename(event.fullpath).split('_')[1][2:])
if queuedLumi>self.maxCMSSWLumi:
self.maxCMSSWLumi = queuedLumi
self.updateQueueStatusFile()
except:
pass
#not passed to the queue
return False
return True
def notifyLumi(self,ls,maxReceivedEoLS,maxClosedLumi,numOpenLumis):
if self.queueStatusPath==None:return
self.lock.acquire()
if ls!=None and ls in self.queuedLumiList:
self.queuedLumiList.remove(ls)
self.maxReceivedEoLS=maxReceivedEoLS
self.maxClosedLumi=maxClosedLumi
self.numOpenLumis=numOpenLumis
self.lock.release()
self.updateQueueStatusFile()
def setQueueStatusPath(self,path,monpath):
self.queueStatusPath = path
self.queueStatusPathMon = monpath
self.queueStatusPathDir = path[:path.rfind('/')]
def updateQueueStatusFile(self):
if self.queueStatusPath==None:return
num_queued_lumis = len(self.queuedLumiList)
if not os.path.exists(self.queueStatusPathDir):
self.logger.error("No directory to write queueStatusFile: "+str(self.queueStatusPathDir))
else:
self.logger.info("Update status file - queued lumis:"+str(num_queued_lumis)+ " EoLS:: max queued:"+str(self.maxQueuedLumi) \
+" un-queued:"+str(self.maxReceivedEoLS)+" Lumis:: last closed:"+str(self.maxClosedLumi) \
+ " num open:"+str(self.numOpenLumis) + " max LS in cmssw:"+str(self.maxCMSSWLumi))
#write json
doc = {"numQueuedLS":num_queued_lumis,
"maxQueuedLS":self.maxQueuedLumi,
"numReadFromQueueLS:":self.maxReceivedEoLS,
"maxClosedLS":self.maxClosedLumi,
"numReadOpenLS":self.numOpenLumis,
"CMSSWMaxLS":self.maxCMSSWLumi
}
try:
if self.queueStatusPath!=None:
attempts=3
while attempts>0:
try:
with open(self.queueStatusPath+TEMPEXT,"w") as fp:
#fcntl.flock(fp, fcntl.LOCK_EX)
json.dump(doc,fp)
os.rename(self.queueStatusPath+TEMPEXT,self.queueStatusPath)
break
except Exception as ex:
attempts-=1
if attempts==0:
raise ex
self.logger.warning("Unable to write status file, with error:" + str(ex)+".retrying...")
time.sleep(0.05)
try:
shutil.copyfile(self.queueStatusPath,self.queueStatusPathMon)
except:
pass
except Exception as ex:
self.logger.error("Unable to open/write " + self.queueStatusPath)
self.logger.exception(ex)
class MonitorRanger:
def __init__(self,recursiveMode=False):
self.logger = logging.getLogger(self.__class__.__name__)
self.eventQueue = False
self.inotifyWrapper = InotifyWrapper(self,recursiveMode)
self.queueStatusPath = None
self.queueStatusPathMon = None
self.queueStatusPathDir = None
self.queuedLumiList = []
self.maxQueuedLumi=-1
#max seen/closed by anelastic thread
self.maxReceivedEoLS=-1
self.maxClosedLumi=-1
self.numOpenLumis=-1
self.maxCMSSWLumi=-1
self.maxLSWithOutput=-1
self.lock = threading.Lock()
self.output_bw=0
self.lumi_bw=0
self.data_size_ls_num = 0
self.data_size_val = 0
self.data_size_last_update = 0
self.statsCollectorThread = None
def startStatsCollector(self):
self.statsCollectorThread = threading.Thread(target=self.statsCollector)
self.statsCollectorThread.daemon=True #set as daemon thread (not blocking process termination)
self.statsCollectorThread.start()
def statsCollector(self):
global bw_cnt
bw_cnt_time=None
while True:
new_time = time.time()
if bw_cnt_time is not None:
d_t = new_time-bw_cnt_time
if d_t!=0:
self.output_bw=bw_cnt/d_t
bw_cnt=0
bw_cnt_time=new_time
#refresh last completed lumi BW
if self.data_size_ls_num>0:
if new_time - self.data_size_last_update < 60:
self.lumi_bw=self.data_size_val/23.31
else:
self.lumi_bw=0.
if self.queueStatusPathDir and not os.path.exists(self.queueStatusPathDir):
self.logger.info('no queue status dir yet.')
else:
self.updateQueueStatusFile(".statsCollector")
time.sleep(23.4)
def register_inotify_path(self,path,mask):
self.inotifyWrapper.registerPath(path,mask)
def start_inotify(self):
self.inotifyWrapper.start()
def stop_inotifyTimeout(self,timeout):
self.logger.info("MonitorRanger: Stop inotify wrapper")
self.inotifyWrapper.stop()
self.logger.info("MonitorRanger: Join inotify wrapper")
self.inotifyWrapper.join(timeout)
if self.inotifyWrapper.isAlive():
self.logger.info("MonitorRanger: Inotify wrapper join timeout ("+str(timeout)+")")
return False
else:
self.logger.info("MonitorRanger: Inotify wrapper returned")
return True
def stop_inotify(self):
self.logger.info("MonitorRanger: Stop inotify wrapper")
self.inotifyWrapper.stop()
self.logger.info("MonitorRanger: Join inotify wrapper")
self.inotifyWrapper.join()
self.logger.info("MonitorRanger: Inotify wrapper returned")
def process_default(self, event):
self.logger.debug("event: %s on: %s" %(str(event.mask),event.fullpath))
if self.eventQueue:
if self.queueStatusPath!=None:
if self.checkNewLumi(event):
self.eventQueue.put(event)
else:
self.eventQueue.put(event)
def setEventQueue(self,queue):
self.eventQueue = queue
def checkNewLumi(self,event):
if event.fullpath.endswith("_EoLS.jsn"):
try:
queuedLumi = int(os.path.basename(event.fullpath).split('_')[1][2:])
self.lock.acquire()
if queuedLumi not in self.queuedLumiList:
if queuedLumi>self.maxQueuedLumi:
self.maxQueuedLumi=queuedLumi
self.queuedLumiList.append(queuedLumi)
self.lock.release()
self.updateQueueStatusFile(".checkNewLumi")
else:
self.lock.release()
#skip if EoL for LS in queue has already been written once (e.g. double file create race)
return False
except Exception as ex:
self.logger.warning("Problem checking new EoLS filename: "+str(os.path.basename(event.fullpath)) + " error:"+str(ex))
try:self.lock.release()
except:pass
#delete associated BoLS file
try:
os.unlink(event.fullpath[:event.fullpath.rfind("_EoLS.jsn")]+"_BoLS.jsn")
except:
pass
elif event.fullpath.endswith("_BoLS.jsn"):
try:
queuedLumi = int(os.path.basename(event.fullpath).split('_')[1][2:])
if queuedLumi>self.maxCMSSWLumi:
self.maxCMSSWLumi = queuedLumi
self.updateQueueStatusFile(".checkNewLumi")
except:
pass
#not passed to the queue
return False
return True
def notifyLumi(self,ls,maxReceivedEoLS,maxClosedLumi,numOpenLumis):
if self.queueStatusPath==None:return
self.lock.acquire()
if ls!=None and ls in self.queuedLumiList:
self.queuedLumiList.remove(ls)
self.maxReceivedEoLS=maxReceivedEoLS
self.maxClosedLumi=maxClosedLumi
self.numOpenLumis=numOpenLumis
self.lock.release()
self.updateQueueStatusFile(".notifyLumi")
def notifyMaxLsWithOutput(self,ls):
self.maxLSWithOutput=max(ls,self.maxLSWithOutput)
def setQueueStatusPath(self,path,monpath):
self.queueStatusPath = path
self.queueStatusPathMon = monpath
self.queueStatusPathDir = path[:path.rfind('/')]
def updateQueueStatusFile(self,tmpsuffix):
if self.queueStatusPath==None:return
num_queued_lumis = len(self.queuedLumiList)
if not os.path.exists(self.queueStatusPathDir):
self.logger.error("No directory to write queueStatusFile: "+str(self.queueStatusPathDir))
else:
self.logger.info("Update status file - queued lumis:"+str(num_queued_lumis)+ " EoLS:: max queued:"+str(self.maxQueuedLumi) \
+" un-queued:"+str(self.maxReceivedEoLS)+" Lumis:: last closed:"+str(self.maxClosedLumi) \
+ " num open:"+str(self.numOpenLumis) + " max LS in cmssw:"+str(self.maxCMSSWLumi))
#write json
doc = {"numQueuedLS":num_queued_lumis,
"maxQueuedLS":self.maxQueuedLumi,
"numReadFromQueueLS:":self.maxReceivedEoLS,
"maxClosedLS":self.maxClosedLumi,
"numReadOpenLS":self.numOpenLumis,
"CMSSWMaxLS":self.maxCMSSWLumi,
"maxLSWithOutput":self.maxLSWithOutput,
"outputBW": self.output_bw,
"lumiBW": self.lumi_bw
}
if self.queueStatusPath!=None:
attempts=3
while attempts>0:
try:
#copy to main hltd
tmpjson=json.dumps(doc)
with open(self.queueStatusPath+tmpsuffix+TEMPEXT,"w") as fp:
json.dump(doc,fp)
os.rename(self.queueStatusPath+tmpsuffix+TEMPEXT,self.queueStatusPath)
#copy to monitoring directory
try:
shutil.copyfile(self.queueStatusPath,self.queueStatusPathMon)
except:
pass
break
except Exception as ex:
attempts-=1
if attempts==0:
self.logger.error("Unable to open/write " + self.queueStatusPath)
self.logger.exception(ex)
else:
self.logger.warning("Unable to write status file, with error:" + str(ex)+".retrying...")
time.sleep(0.05)
class ResourceRanger:
def __init__(self,confClass,stateInfo,resInfo,runList,mountMgr,boxInfo,monitor,resource_lock):
self.inotifyWrapper = InotifyWrapper(self)
self.logger = logging.getLogger(self.__class__.__name__)
self.state = stateInfo
self.resInfo = resInfo
self.runList = runList
self.managed_monitor = monitor
self.managed_monitor.preStart()
self.managed_monitor.start()
self.regpath = []
self.mm = mountMgr
self.boxInfo = boxInfo
self.resource_lock = resource_lock
self.hostname = os.uname()[1]
global conf
conf = confClass
def register_inotify_path(self,path,mask):
self.inotifyWrapper.registerPath(path,mask)
self.regpath.append(path)
def start_inotify(self):
self.inotifyWrapper.start()
def stop_managed_monitor(self):
self.managed_monitor.stop()
self.managed_monitor.join()
self.logger.info("ResourceRanger: managed monitor shutdown done")
def stop_inotify(self):
self.inotifyWrapper.stop()
self.inotifyWrapper.join()
self.logger.info("ResourceRanger: Inotify wrapper shutdown done")
def process_IN_MOVED_TO(self, event):
self.logger.debug('ResourceRanger-MOVEDTO: event '+event.fullpath)
basename = os.path.basename(event.fullpath)
#closures for stopping resources
def stopResourceMaybe(resourcename,current_run,quarantining):
#detect if this resource belongs to the run and terminate the process if needed
#this will only happen if last run is an ongoing run
if not conf.role=='fu' or not conf.dynamic_resources: return None
activeRuns = self.runList.getActiveRuns()
for checkRun in activeRuns:
#skip if this is part of normal stopping procedure
if current_run:
if not current_run.is_ongoing_run and current_run.runnumber==checkRun.runnumber:continue
for checkRes in checkRun.online_resource_list:
if resourcename in checkRes.cpu and checkRes.processstate==100:
self.logger.info('found matching resource for '+resourcename)
time.sleep(.1)
#allow EoR procedure (no process restarts for that run) in resource is from a previous run
#TODO: check if this is last remaining resource of current run, then EoR action could be taken
eor_allow = current_run.runnumber != checkRun.runnumber and not quarantining
checkRes.Stop(end_run_allow=eor_allow,move_q=quarantining)#stop and release all resources
return checkRes
return None
def waitResource(resource,is_locked):
def resJoin(join_timeout):
if is_locked:
try:self.resource_lock.release()
except:pass
resource.watchdog.join(join_timeout)
if is_locked:self.resource_lock.acquire()
if resource:
try:
resJoin(120)
if resource.isAlive():
self.logger.info('terminating ' + resource.process.pid)
resource.process.terminate()
resJoin(30)
if resource.isAlive():
self.logger.info('killing ' + resource.process.pid)
resource.process.kill()
resJoin(10)
except Exception as ex:
self.logger.info("exception in waitResource: "+str(ex))
if is_locked:
#make sure to return it locked
try:self.resource_lock.release()
except:pass
self.resource_lock.acquire()
return
if basename.startswith('resource_summary'):return
try:
resourcepath=event.fullpath[1:event.fullpath.rfind("/")]
resourcestate=resourcepath[resourcepath.rfind("/")+1:]
resourcename=event.fullpath[event.fullpath.rfind("/")+1:]
self.resource_lock.acquire()
if not (resourcestate == 'online' or resourcestate == 'cloud'
or resourcestate == 'quarantined'):
self.logger.debug('ResourceNotifier: new resource '
+resourcename
+' in '
+resourcepath
+' state '
+resourcestate
)
if self.state.cloud_mode and not \
self.state.entering_cloud_mode and not \
self.state.exiting_cloud_mode and not \
self.state.abort_cloud_mode and not \
self.state.disabled_resource_allocation:
time.sleep(1)
logging.info('detected resource moved to non-cloud resource dir while already switched to cloud mode. Deactivating cloud.')
with open(os.path.join(conf.watch_directory,'include'),'w+') as fobj:
pass
self.resource_lock.release()
time.sleep(1)
return
run = self.runList.getLastOngoingRun()
if run is not None:
self.logger.info("ResourceRanger: found active run "+str(run.runnumber)+ " when received inotify MOVED event for "+event.fullpath)
"""grab resources that become available
#@@EM implement threaded acquisition of resources here
"""
#make sure owner of the process ends or is terminated (in case file was moved by action other than process exit)
if os.path.exists(event.fullpath):
waitResource(stopResourceMaybe(resourcename,run,False),is_locked=True)
if run is not None:
#find all ready cores in same dir where inotify was triggered
try:
reslist = os.listdir('/'+resourcepath)
except Exception as ex:
self.logger.error("RUN:"+str(run.runnumber)+" - exception encountered in looking for resources")
self.logger.exception(ex)
#put inotify-ed resource as the first item
fileFound=False
for resindex,resname in enumerate(reslist):
fileFound=False
if resname == resourcename:
fileFound=True
if resindex != 0:
firstitem = reslist[0]
reslist[0] = resourcename
reslist[resindex] = firstitem
break
if fileFound==False:
#inotified file was already moved earlier
self.resource_lock.release()
return
#acquire sufficient cores for a multithreaded process start
#returns whether it can be matched to existing online resource or not
matchedList = run.MatchResource(reslist)
if matchedList:
#matched with previous resource (restarting process)
acquired_sufficient = True
res = run.AcquireResource(matchedList,resourcestate)
else:
resourcenames = []
for resname in reslist:
if len(resourcenames) < self.resInfo.nstreams:
resourcenames.append(resname)
else:
break
acquired_sufficient = False
if len(resourcenames) == self.resInfo.nstreams:
acquired_sufficient = True
res = run.AcquireResource(resourcenames,resourcestate)
if acquired_sufficient:
self.logger.info("ResourceRanger: acquired resource(s) "+str(res.cpu))
run.StartOnResource(res)
self.logger.info("ResourceRanger: started process on resource "
+str(res.cpu))
else:
#if no run is active, move (x N threads) files from except to idle to be picked up for the next run
#todo: debug,write test for this...
if resourcestate == 'except':
try:
reslist = os.listdir('/'+resourcepath)
#put inotify-ed resource as the first item
fileFound=False
for resindex,resname in enumerate(reslist):
if resname == resourcename:
fileFound=True
if resindex != 0:
firstitem = reslist[0]
reslist[0] = resourcename
reslist[resindex] = firstitem
break
if fileFound==False:
#inotified file was already moved earlier
self.resource_lock.release()
return
resourcenames = []
for resname in reslist:
if len(resourcenames) < self.resInfo.nstreams:
resourcenames.append(resname)
else:
break
if len(resourcenames) == self.resInfo.nstreams:
for resname in resourcenames:
self.resInfo.resmove(self.resInfo.broken,self.resInfo.idles,resname)
#move this except after listdir?
except Exception as ex:
self.logger.info("exception encountered in looking for resources in except")
self.logger.info(ex)
elif resourcestate=="cloud":
#check if cloud mode was initiated, activate if necessary
if conf.role=='fu' and self.state.cloud_mode==False:
time.sleep(1)
logging.info('detected core moved to cloud resources. Triggering cloud activation sequence.')
with open(os.path.join(conf.watch_directory,'exclude'),'w+') as fobj:
pass
time.sleep(1)
elif resourcestate=="quarantined":
#quarantined check, terminate owner if needed
if os.path.exists(event.fullpath):
waitResource(stopResourceMaybe(resourcename,self.runList.getLastOngoingRun(),True),is_locked=True)
except Exception as ex:
self.logger.error("exception in ResourceRanger")
self.logger.exception(ex)
try:
self.resource_lock.release()
except:pass
def process_IN_CREATE(self, event):
self.logger.debug('ResourceRanger-CREATE: event '+event.fullpath)
if conf.dqm_machine:return
basename = os.path.basename(event.fullpath)
if basename.startswith('resource_summary'):return
if basename=='blacklist':return
if basename.startswith('test'):return
if conf.role!='bu' or basename.endswith(self.hostname):
return
#catch stop and hltd restart on FU which might leave state inconsistent
try:
if basename in self.boxInfo.FUMap:
if self.boxInfo.FUMap[basename][0]["cloudState"]=="resourcesMasked":
self.boxInfo.FUMap[basename][0]["cloudState"]="off"
except:
pass
self.findRunAndNotify(basename,event.fullpath,False)
def process_default(self, event):
self.logger.debug('ResourceRanger: event '+event.fullpath +' type '+ str(event.mask))
filename=event.fullpath[event.fullpath.rfind("/")+1:]
def process_IN_CLOSE_WRITE(self, event):
self.logger.debug('ResourceRanger-IN_CLOSE_WRITE: event '+event.fullpath)
resourcepath=event.fullpath[0:event.fullpath.rfind("/")]
basename = os.path.basename(event.fullpath)
if basename.startswith('resource_summary'):return
if conf.role=='fu':return
if basename == os.uname()[1]:return
if basename.startswith('test'):return
if basename == 'blacklist':
with open(os.path.join(conf.watch_directory,'appliance','blacklist'),'r') as fi:
try:
self.boxInfo.machine_blacklist = json.load(fi)
except:
pass
if resourcepath.endswith('boxes'):
if basename in self.boxInfo.machine_blacklist:
try:self.boxInfo.FUMap.pop(basename)
except:pass
else:
current_time = time.time()
current_datetime = datetime.datetime.utcfromtimestamp(current_time)
emptyBox=False
try:
currentBox = self.boxInfo.FUMap[basename]
except:
currentBox = None
try:
infile = fileHandler(event.fullpath)
if infile.data=={}:emptyBox=True
#check which time is later (in case of small clock skew and small difference)
if current_datetime > dateutil.parser.parse(infile.data['fm_date']):
dt = (current_datetime - dateutil.parser.parse(infile.data['fm_date'])).seconds
else:
dt = -(dateutil.parser.parse(infile.data['fm_date'])-current_datetime).seconds
if dt > 5:
self.logger.warning('setting stale flag for resource '+basename + ' which is '+str(dt)+' seconds behind')
#should be << 1s if NFS is responsive, set stale handle flag
infile.data['detectedStaleHandle']=True
elif dt < -5:
self.logger.error('setting stale flag for resource '+basename + ' which is '+str(dt)+' seconds ahead (clock skew)')
infile.data['detectedStaleHandle']=True
self.boxInfo.FUMap[basename] = [infile.data,current_time,True]
#detect flip from cloud to non-cloud (only notify FU if no active runs are found FU)
try:
if currentBox and currentBox[0]["cloudState"]!="off" and infile.data["cloudState"]=="off" and len(infile.data["activeRuns"])==0:
self.logger.info('cloud state flip detected for ' + str(basename) + ':' + str(currentBox[0]["cloudState"]) + ' to ' + str(infile.data["cloudState"]))
self.findRunAndNotify(basename,event.fullpath,True)
except (KeyError,IndexError,TypeError) as ex:
self.logger.warning("cloud flip detection problem: "+str(ex))
except Exception as ex:
if not emptyBox:
self.logger.error("Unable to read of parse boxinfo file "+basename)
self.logger.exception(ex)
else:
self.logger.warning("got empty box file "+basename)
try:
self.boxInfo.FUMap[basename][2]=False
except:
#boxinfo entry doesn't exist yet
self.boxInfo.FUMap[basename]=[None,current_time,False]
def findRunAndNotify(self,basename,fullpath,override):
try:
resourceage = os.path.getmtime(fullpath)
self.resource_lock.acquire()
lrun = self.runList.getLastRun()
newRes = None
if lrun!=None:
is_stale,f_ip = lrun.checkStaleResourceFileAndIP(fullpath)
if is_stale:
self.logger.error("RUN:"+str(lrun.runnumber)+" - notification: skipping resource "+basename+" which is stale")
self.resource_lock.release()
return
self.logger.info('Try attaching FU resource: last run is '+str(lrun.runnumber))
newRes = lrun.maybeNotifyNewRun(basename,resourceage,f_ip,override)
self.resource_lock.release()
if newRes:
newRes.NotifyNewRun(lrun.runnumber)
except Exception as ex:
self.logger.exception(ex)
try:self.resource_lock.release()
except:pass
def checkNotifiedBoxes(self,runNumber):
keys = self.boxInfo.FUMap.keys()
c_time = time.time()
for key in keys:
#if key==self.hostname:continue #checked in inotify thread
try:
edata,etime,lastStatus = self.boxInfo.FUMap[key]
except:
#key deleted
return False,False
if c_time - etime > 20:continue
#parsing or file access, check failed
if lastStatus==False: return False,False
try:
#run is found in at least one box
if runNumber in edata['activeRuns']:return True,True
except:
#invalid boxinfo data
return False,False
#all box data are valid, run not found
return True,False
def checkBoxes(self,runNumber):
checkSuccessful=True
runFound=False
ioErrCount=0
valErrCount=0
files = os.listdir(self.regpath[-1])
c_time = time.time()
for file in files:
if file == self.hostname:continue
#ignore file if it is too old (FU with a problem)
filename = os.path.join(dir,file)
if c_time - os.path.getmtime(filename) > 20:continue
try:
with open(filename,'r') as fp:
doc = json.load(fp)
except IOError as ex:
checkSuccessful=False
break
except ValueError as ex:
checkSuccessful=False
break
except Exception as ex:
self.logger.exception(ex)
checkSuccessful=False
break;
try:
if runNumber in doc['activeRuns']:
runFound=True
break;
except Exception as ex:
self.logger.exception(ex)
checkSuccessful=False
break
return checkSuccessful,runFound
