def checkReplicationStatus(self, continuous = True):
"""
_checkReplicationStatus_
Check if the workqueue replication is ok, if not
then delete the documents so that new replications can be triggered
when appropiate.
It returns True if there is no error, and False otherwise.
"""
if self.parentCouchUrl and self.queueUrl:
# only checks for local queue
couchMonitor = CouchMonitor(self.server)
filter = 'WorkQueue/queueFilter'
query_params = {'childUrl' : self.queueUrl, 'parentUrl' : self.parentCouchUrl}
self.logger.info("set replication from GQ to LQ")
couchMonitor.recoverReplicationErrors(self.parentCouchUrl,
"%s/%s" % (self.hostWithAuth, self.inbox.name),
filter = filter, query_params = query_params,
checkUpdateSeq = False,
continuous = continuous)
self.logger.info("set replication from LQ to GQ")
couchMonitor.recoverReplicationErrors(self.inbox.name, self.parentCouchUrlWithAuth,
filter = filter, query_params = query_params,
checkUpdateSeq = False,
continuous = continuous)
return True
return False
def setUp(self):
"""
_setUp_
Setup the database and logging connection. Try to create all of the
WMBS tables. Also add some dummy locations.
"""
super(WorkQueueTestCase, self).setUp()
self.queueDB = 'workqueue_t'
self.queueInboxDB = 'workqueue_t_inbox'
self.globalQDB = 'workqueue_t_global'
self.globalQInboxDB = 'workqueue_t_global_inbox'
self.localQDB = 'workqueue_t_local'
self.localQInboxDB = 'workqueue_t_local_inbox'
self.localQDB2 = 'workqueue_t_local2'
self.localQInboxDB2 = 'workqueue_t_local2_inbox'
self.configCacheDB = 'workqueue_t_config_cache'
self.logDBName = 'logdb_t'
self.requestDBName = 'workqueue_t_reqmgr_workload_cache'
self.setSchema()
self.testInit = TestInit('WorkQueueTest')
self.testInit.setLogging()
self.testInit.setDatabaseConnection(destroyAllDatabase=True)
self.addCleanup(self.testInit.clearDatabase)
self.addCleanup(logging.debug, 'Cleanup called clearDatabase()')
self.testInit.setSchema(customModules = self.schema,
useDefault = False)
self.testInit.setupCouch(self.queueDB, *self.couchApps)
self.testInit.setupCouch(self.queueInboxDB, *self.couchApps)
self.testInit.setupCouch(self.globalQDB, *self.couchApps)
self.testInit.setupCouch(self.globalQInboxDB , *self.couchApps)
self.testInit.setupCouch(self.localQDB, *self.couchApps)
self.testInit.setupCouch(self.localQInboxDB, *self.couchApps)
self.testInit.setupCouch(self.localQDB2, *self.couchApps)
self.testInit.setupCouch(self.localQInboxDB2, *self.couchApps)
self.testInit.setupCouch(self.configCacheDB, 'ConfigCache')
self.testInit.setupCouch(self.logDBName, 'LogDB')
self.testInit.setupCouch(self.requestDBName, 'ReqMgr')
self.couchURL = os.environ.get("COUCHURL")
couchServer = CouchServer(self.couchURL)
self.configCacheDBInstance = couchServer.connectDatabase(self.configCacheDB)
self.localCouchMonitor = CouchMonitor(self.couchURL)
self.localCouchMonitor.deleteReplicatorDocs()
self.addCleanup(self.localCouchMonitor.deleteReplicatorDocs)
self.addCleanup(logging.debug, 'Cleanup called deleteReplicatorDocs()')
self.addCleanup(self.testInit.tearDownCouch)
self.addCleanup(logging.debug, 'Cleanup called tearDownCouch()')
self.workDir = self.testInit.generateWorkDir()
self.addCleanup(self.testInit.delWorkDir)
self.addCleanup(logging.debug, 'Cleanup called delWorkDir()')
return
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.General.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# set the connection to local queue
self.localQueue = WorkQueueService(self.config.AnalyticsDataCollector.localQueueURL)
# set the connection for local couchDB call
self.localCouchDB = LocalCouchDBData(self.config.AnalyticsDataCollector.localCouchURL,
self.config.JobStateMachine.summaryStatsDBName,
self.summaryLevel)
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
# set the connection for local couchDB call
self.localSummaryCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.localWMStatsURL,
"WMStatsAgent")
if hasattr(self.config, "Tier0Feeder"):
#use local db for tier0
centralRequestCouchDBURL = self.config.AnalyticsDataCollector.localT0RequestDBURL
else:
centralRequestCouchDBURL = self.config.AnalyticsDataCollector.centralRequestDBURL
self.centralRequestCouchDB = RequestDBWriter(centralRequestCouchDBURL,
couchapp = self.config.AnalyticsDataCollector.RequestCouchApp)
#TODO: change the config to hold couch url
self.localCouchServer = CouchMonitor(self.config.JobStateMachine.couchurl)
if self.pluginName != None:
pluginFactory = WMFactory("plugins", "WMComponent.AnalyticsDataCollector.Plugins")
self.plugin = pluginFactory.loadObject(classname = self.pluginName)
def checkReplicationStatus(self, continuous=True):
"""
_checkReplicationStatus_
Check if the workqueue replication is ok, if not
then delete the documents so that new replications can be triggered
when appropiate.
It returns True if there is no error, and False otherwise.
"""
if self.parentCouchUrl and self.queueUrl:
# only checks for local queue
couchMonitor = CouchMonitor(self.server)
filter = 'WorkQueue/queueFilter'
query_params = {
'childUrl': self.queueUrl,
'parentUrl': self.parentCouchUrl
}
self.logger.info("set replication from GQ to LQ")
couchMonitor.recoverReplicationErrors(
self.parentCouchUrl,
"%s/%s" % (self.hostWithAuth, self.inbox.name),
filter=filter,
query_params=query_params,
checkUpdateSeq=False,
continuous=continuous)
self.logger.info("set replication from LQ to GQ")
couchMonitor.recoverReplicationErrors(self.inbox.name,
self.parentCouchUrlWithAuth,
filter=filter,
query_params=query_params,
checkUpdateSeq=False,
continuous=continuous)
return True
return False
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi,
myThread.logger)
if hasattr(self.config, "Tier0Feeder"):
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.localWMStatsURL,
appName="WMStatsAgent")
else:
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(
self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
# set the connection for local couchDB call
#self.localSummaryCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.localWMStatsURL)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchServer = CouchMonitor(self.config.JobStateMachine.couchurl)
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
self.jsonFile = config.AgentStatusWatcher.jsonFile
proxyArgs = {'logger': logging.getLogger()}
self.proxy = Proxy(proxyArgs)
self.proxyFile = self.proxy.getProxyFilename() # X509_USER_PROXY
localWQUrl = config.AnalyticsDataCollector.localQueueURL
self.workqueueDS = WorkQueueDS(localWQUrl)
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({'source': wmstatsSource, 'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"})
# TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({'source': t0Source, 'target': t0Target,
'filter': "T0Request/repfilter"})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams["ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {'childUrl': childURL, 'parentUrl': sanitizeURL(parentQURL)['url']}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({'source': sanitizeURL(parentQURL)['url'], 'target': localQInboxURL,
'filter': wqfilter, 'query_params': query_params})
self.replicatorDocs.append({'source': sanitizeURL(localQInboxURL)['url'], 'target': parentQURL,
'filter': wqfilter, 'query_params': query_params})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(
rp['source'], rp['target'], filter=rp['filter'],
query_params=rp.get('query_params', False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
self.checkProxyLifetime(agentInfo)
timeSpent, wmbsInfo, _ = self.collectWMBSInfo()
wmbsInfo['total_query_time'] = int(timeSpent)
agentInfo["WMBS_INFO"] = wmbsInfo
logging.info("WMBS data collected in: %d secs", timeSpent)
if not hasattr(self.config, "Tier0Feeder"):
# Tier0 Agent doesn't have LQ.
timeSpent, localWQInfo, _ = self.collectWorkQueueInfo()
localWQInfo['total_query_time'] = int(timeSpent)
agentInfo["LocalWQ_INFO"] = localWQInfo
logging.info("Local WorkQueue data collected in: %d secs", timeSpent)
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
# save locally json file as well
with open(self.jsonFile, 'w') as outFile:
json.dump(agentInfo, outFile, indent=2)
except Exception as ex:
logging.exception("Error occurred, will retry later.\nDetails: %s", str(ex))
@timeFunction
def collectWorkQueueInfo(self):
"""
Collect information from local workqueue database
:return:
"""
results = {}
results['workByStatus'] = self.workqueueDS.getJobsByStatus()
results['workByStatusAndPriority'] = self.workqueueDS.getJobsByStatusAndPriority()
elements = self.workqueueDS.getElementsByStatus(['Available', 'Acquired'])
uniSites, posSites = getGlobalSiteStatusSummary(elements, dataLocality=True)
results['uniqueJobsPerSite'] = uniSites
results['possibleJobsPerSite'] = posSites
return results
def collectCouchDBInfo(self):
couchInfo = {'name': 'CouchServer', 'status': 'ok', 'error_message': ""}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(rp['source'],
rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
agentInfo['disk_warning'] = listDiskUsageOverThreshold(self.config, updateDB=True)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['drain_stats'] = DrainStatusPoller.getDrainInfo()
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo()
if lastDataUpload['data_last_update']:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error']:
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode'] or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get('couch_process_warning', 0):
agentInfo['status'] = "error"
logging.info("List of agent components down: %s", agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
# direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC, uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
@timeFunction
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
logging.info("Getting wmbs job info ...")
results = {}
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.debug("Running and pending site thresholds: %s", results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
logging.debug("Total number of jobs in WMBS sorted by status: %s", results['wmbsCountByState'])
logging.debug("Total number of 'created' jobs in WMBS sorted by type: %s", results['wmbsCreatedTypeCount'])
logging.debug("Total number of 'executing' jobs in WMBS sorted by type: %s", results['wmbsExecutingTypeCount'])
logging.debug("Total number of active jobs in BossAir sorted by status: %s", results['activeRunJobByStatus'])
logging.debug("Total number of complete jobs in BossAir sorted by status: %s",
results['completeRunJobByStatus'])
logging.debug("Available slots thresholds to pull work from GQ to LQ: %s", results['thresholdsGQ2LQ'])
logging.debug("List of jobs pending for each site, sorted by priority: %s", results['sitePendCountByPrio'])
return results
def checkProxyLifetime(self, agInfo):
"""
Check the proxy lifetime (usually X509_USER_CERT) and raise either
a warning or an error if the proxy validity is about to expire.
:param agInfo: dictionary with plenty of agent monitoring information in place.
:return: same dictionary object plus additional keys/values if needed.
"""
secsLeft = self.proxy.getTimeLeft(proxy=self.proxyFile)
logging.debug("Proxy '%s' lifetime is %d secs", self.proxyFile, secsLeft)
if secsLeft <= 86400 * 3: # 3 days
proxyWarning = True
agInfo['status'] = "error"
elif secsLeft <= 86400 * 5: # 5 days
proxyWarning = True
if agInfo['status'] == "ok":
agInfo['status'] = "warning"
else:
proxyWarning = False
if proxyWarning:
warnMsg = "Agent proxy '%s' must be renewed ASAP. " % self.proxyFile
warnMsg += "Its time left is: %.2f hours." % (secsLeft / 3600.)
agInfo['proxy_warning'] = warnMsg
return
class WorkQueueTestCase(EmulatedUnitTestCase):
def setSchema(self):
"this can be override if the schema setting is different"
self.schema = ["WMCore.WMBS","WMComponent.DBS3Buffer","WMCore.BossAir"]
self.couchApps = ["WorkQueue"]
def setUp(self):
"""
_setUp_
Setup the database and logging connection. Try to create all of the
WMBS tables. Also add some dummy locations.
"""
super(WorkQueueTestCase, self).setUp()
self.queueDB = 'workqueue_t'
self.queueInboxDB = 'workqueue_t_inbox'
self.globalQDB = 'workqueue_t_global'
self.globalQInboxDB = 'workqueue_t_global_inbox'
self.localQDB = 'workqueue_t_local'
self.localQInboxDB = 'workqueue_t_local_inbox'
self.localQDB2 = 'workqueue_t_local2'
self.localQInboxDB2 = 'workqueue_t_local2_inbox'
self.configCacheDB = 'workqueue_t_config_cache'
self.logDBName = 'logdb_t'
self.requestDBName = 'workqueue_t_reqmgr_workload_cache'
self.setSchema()
self.testInit = TestInit('WorkQueueTest')
self.testInit.setLogging()
self.testInit.setDatabaseConnection(destroyAllDatabase=True)
self.addCleanup(self.testInit.clearDatabase)
self.addCleanup(logging.debug, 'Cleanup called clearDatabase()')
self.testInit.setSchema(customModules = self.schema,
useDefault = False)
self.testInit.setupCouch(self.queueDB, *self.couchApps)
self.testInit.setupCouch(self.queueInboxDB, *self.couchApps)
self.testInit.setupCouch(self.globalQDB, *self.couchApps)
self.testInit.setupCouch(self.globalQInboxDB , *self.couchApps)
self.testInit.setupCouch(self.localQDB, *self.couchApps)
self.testInit.setupCouch(self.localQInboxDB, *self.couchApps)
self.testInit.setupCouch(self.localQDB2, *self.couchApps)
self.testInit.setupCouch(self.localQInboxDB2, *self.couchApps)
self.testInit.setupCouch(self.configCacheDB, 'ConfigCache')
self.testInit.setupCouch(self.logDBName, 'LogDB')
self.testInit.setupCouch(self.requestDBName, 'ReqMgr')
self.couchURL = os.environ.get("COUCHURL")
couchServer = CouchServer(self.couchURL)
self.configCacheDBInstance = couchServer.connectDatabase(self.configCacheDB)
self.localCouchMonitor = CouchMonitor(self.couchURL)
self.localCouchMonitor.deleteReplicatorDocs()
self.addCleanup(self.localCouchMonitor.deleteReplicatorDocs)
self.addCleanup(logging.debug, 'Cleanup called deleteReplicatorDocs()')
self.addCleanup(self.testInit.tearDownCouch)
self.addCleanup(logging.debug, 'Cleanup called tearDownCouch()')
self.workDir = self.testInit.generateWorkDir()
self.addCleanup(self.testInit.delWorkDir)
self.addCleanup(logging.debug, 'Cleanup called delWorkDir()')
return
def tearDown(self):
"""
_tearDown_
Drop all the WMBS tables.
"""
super(WorkQueueTestCase, self).tearDown() # Left here in case it's needed by any of the sub-classes
return
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
self.jsonFile = config.AgentStatusWatcher.jsonFile
# counter for deep agent monitoring. Every 15min (3 cycles of the component)
self.monitorCounter = 0
self.monitorInterval = getattr(config.AgentStatusWatcher,
'monitorPollInterval', 3)
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({
'source': wmstatsSource,
'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"
})
#TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({
'source': t0Source,
'target': t0Target,
'filter': "T0Request/repfilter"
})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams[
"ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {
'childUrl': childURL,
'parentUrl': sanitizeURL(parentQURL)['url']
}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({
'source':
sanitizeURL(parentQURL)['url'],
'target':
localQInboxURL,
'filter':
wqfilter,
'query_params':
query_params
})
self.replicatorDocs.append({
'source':
sanitizeURL(localQInboxURL)['url'],
'target':
parentQURL,
'filter':
wqfilter,
'query_params':
query_params
})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(
rp['source'],
rp['target'],
filter=rp['filter'],
query_params=rp.get('query_params', False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi,
myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(
self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
#set the uploadTime - should be the same for all docs
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
if self.monitorCounter % self.monitorInterval == 0:
monitoring = self.collectWMBSInfo()
monitoring['components'] = agentInfo['down_components']
monitoring['timestamp'] = int(time.time())
with open(self.jsonFile, 'w') as outFile:
json.dump(monitoring, outFile, indent=2)
self.monitorCounter += 1
except Exception as ex:
logging.error("Error occurred, will retry later:")
logging.error(str(ex))
logging.error("Trace back: \n%s" % traceback.format_exc())
def collectCouchDBInfo(self):
couchInfo = {
'name': 'CouchServer',
'status': 'ok',
'error_message': ""
}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(
rp['source'], rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Disk space warning
diskUseList = diskUse()
diskUseThreshold = float(
self.config.AnalyticsDataCollector.diskUseThreshold)
agentInfo['disk_warning'] = []
for disk in diskUseList:
if float(disk['percent'].strip('%')) >= diskUseThreshold and \
disk['mounted'] not in self.config.AnalyticsDataCollector.ignoreDisk:
agentInfo['disk_warning'].append(disk)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo()
if lastDataUpload['data_last_update']:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error']:
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode']
or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get(
'couch_process_warning', 0):
agentInfo['status'] = "error"
if agentInfo['down_components']:
logging.info("List of agent components down: %s" %
agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
#direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC,
uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
results = {}
logging.info("Getting wmbs job info ...")
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.info("Running and pending site thresholds: %s",
results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
logging.info("Total number of jobs in WMBS sorted by status: %s",
results['wmbsCountByState'])
logging.info(
"Total number of 'created' jobs in WMBS sorted by type: %s",
results['wmbsCreatedTypeCount'])
logging.info(
"Total number of 'executing' jobs in WMBS sorted by type: %s",
results['wmbsExecutingTypeCount'])
logging.info(
"Total number of active jobs in BossAir sorted by status: %s",
results['activeRunJobByStatus'])
logging.info(
"Total number of complete jobs in BossAir sorted by status: %s",
results['completeRunJobByStatus'])
logging.info(
"Available slots thresholds to pull work from GQ to LQ: %s",
results['thresholdsGQ2LQ'])
logging.info(
"List of jobs pending for each site, sorted by priority: %s",
results['sitePendCountByPrio'])
return results
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
proxyArgs = {'logger': logging.getLogger(), 'cleanEnvironment': True}
self.proxy = Proxy(proxyArgs)
self.proxyFile = self.proxy.getProxyFilename() # X509_USER_PROXY
self.userCertFile = self.proxy.getUserCertFilename() # X509_USER_CERT
# credential lifetime warning/error thresholds, in days
self.credThresholds = {
'proxy': {
'error': 3,
'warning': 5
},
'certificate': {
'error': 10,
'warning': 20
}
}
# Monitoring setup
self.userAMQ = getattr(config.AgentStatusWatcher, "userAMQ", None)
self.passAMQ = getattr(config.AgentStatusWatcher, "passAMQ", None)
self.postToAMQ = getattr(config.AgentStatusWatcher, "enableAMQ", False)
self.topicAMQ = getattr(config.AgentStatusWatcher, "topicAMQ", None)
self.hostPortAMQ = getattr(config.AgentStatusWatcher, "hostPortAMQ",
[('cms-mb.cern.ch', 61313)])
# T0 doesn't have WorkQueue, so some monitoring/replication code has to be skipped here
if hasattr(self.config, "Tier0Feeder"):
self.isT0agent = True
self.producer = "tier0wmagent"
else:
self.isT0agent = False
self.producer = "wmagent"
localWQUrl = config.AnalyticsDataCollector.localQueueURL
self.workqueueDS = WorkQueueDS(localWQUrl)
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.General.centralWMStatsURL
self.replicatorDocs.append({
'source': wmstatsSource,
'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"
})
if self.isT0agent:
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({
'source': t0Source,
'target': t0Target,
'filter': "T0Request/repfilter"
})
else:
# set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams[
"ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {
'childUrl': childURL,
'parentUrl': sanitizeURL(parentQURL)['url']
}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({
'source': sanitizeURL(parentQURL)['url'],
'target': localQInboxURL,
'filter': wqfilter,
'query_params': query_params
})
self.replicatorDocs.append({
'source':
sanitizeURL(localQInboxURL)['url'],
'target':
parentQURL,
'filter':
wqfilter,
'query_params':
query_params
})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(rp['source'],
rp['target'],
filter=rp['filter'],
query_params=rp.get(
'query_params',
False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi,
myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.General.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(
self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
@timeFunction
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
self.checkCredLifetime(agentInfo, "proxy")
self.checkCredLifetime(agentInfo, "certificate")
timeSpent, wmbsInfo, _ = self.collectWMBSInfo()
wmbsInfo['total_query_time'] = int(timeSpent)
agentInfo["WMBS_INFO"] = wmbsInfo
logging.info("WMBS data collected in: %d secs", timeSpent)
if not self.isT0agent:
timeSpent, localWQInfo, _ = self.collectWorkQueueInfo()
localWQInfo['total_query_time'] = int(timeSpent)
agentInfo["LocalWQ_INFO"] = localWQInfo
logging.info("Local WorkQueue data collected in: %d secs",
timeSpent)
self.uploadAgentInfoToCentralWMStats(agentInfo)
self.buildMonITDocs(agentInfo)
except Exception as ex:
logging.exception("Error occurred, will retry later.\nDetails: %s",
str(ex))
@timeFunction
def collectWorkQueueInfo(self):
"""
Collect information from local workqueue database
:return:
"""
results = {}
wqStates = ['Available', 'Acquired']
results['workByStatus'] = self.workqueueDS.getJobsByStatus()
results[
'workByStatusAndPriority'] = self.workqueueDS.getJobsByStatusAndPriority(
)
elements = self.workqueueDS.getElementsByStatus(wqStates)
uniSites, posSites = getGlobalSiteStatusSummary(elements,
status=wqStates,
dataLocality=True)
results['uniqueJobsPerSite'] = uniSites
results['possibleJobsPerSite'] = posSites
return results
def collectCouchDBInfo(self):
couchInfo = {
'name': 'CouchServer',
'status': 'ok',
'error_message': ""
}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(
rp['source'], rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
agentInfo['disk_warning'] = listDiskUsageOverThreshold(self.config,
updateDB=True)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['drain_stats'] = DrainStatusPoller.getDrainInfo()
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode']
or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get(
'couch_process_warning', 0):
agentInfo['status'] = "error"
logging.info("List of agent components down: %s",
agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo):
"""
Add some required fields to the document before it can get uploaded
to WMStats.
:param agentInfo: dict with agent stats to be posted to couchdb
"""
agentInfo['_id'] = agentInfo["agent_url"]
agentInfo['timestamp'] = int(time.time())
agentInfo['type'] = "agent_info"
# directly upload to the remote to prevent data conflict when agent is cleaned up and redeployed
try:
self.centralWMStatsCouchDB.updateAgentInfo(
agentInfo, propertiesToKeep=["data_last_update", "data_error"])
except Exception as e:
logging.error(
"Failed to upload agent statistics to WMStats. Error: %s",
str(e))
@timeFunction
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
logging.info("Getting wmbs job info ...")
results = {}
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.debug("Running and pending site thresholds: %s",
results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
logging.debug("Total number of jobs in WMBS sorted by status: %s",
results['wmbsCountByState'])
logging.debug(
"Total number of 'created' jobs in WMBS sorted by type: %s",
results['wmbsCreatedTypeCount'])
logging.debug(
"Total number of 'executing' jobs in WMBS sorted by type: %s",
results['wmbsExecutingTypeCount'])
logging.debug(
"Total number of active jobs in BossAir sorted by status: %s",
results['activeRunJobByStatus'])
logging.debug(
"Total number of complete jobs in BossAir sorted by status: %s",
results['completeRunJobByStatus'])
logging.debug(
"Available slots thresholds to pull work from GQ to LQ: %s",
results['thresholdsGQ2LQ'])
logging.debug(
"List of jobs pending for each site, sorted by priority: %s",
results['sitePendCountByPrio'])
return results
def checkCredLifetime(self, agInfo, credType):
"""
Check the credential lifetime. Usually X509_USER_PROXY or X509_USER_CERT
and raise either a warning or an error if the proxy validity is about to expire.
:param agInfo: dictionary with plenty of agent monitoring information in place.
:param credType: credential type, can be: "proxy" or "certificate"
:return: same dictionary object plus additional keys/values if needed.
"""
if credType == "proxy":
credFile = self.proxyFile
secsLeft = self.proxy.getTimeLeft(proxy=credFile)
elif credType == "certificate":
credFile = self.userCertFile
secsLeft = self.proxy.getUserCertTimeLeft(openSSL=True)
else:
logging.error(
"Unknown credential type. Available options are: [proxy, certificate]"
)
return
logging.debug("%s '%s' lifetime is %d seconds", credType, credFile,
secsLeft)
daysLeft = secsLeft / (60 * 60 * 24)
if daysLeft <= self.credThresholds[credType]['error']:
credWarning = True
agInfo['status'] = "error"
elif daysLeft <= self.credThresholds[credType]['warning']:
credWarning = True
if agInfo['status'] == "ok":
agInfo['status'] = "warning"
else:
credWarning = False
if credWarning:
warnMsg = "Agent %s '%s' must be renewed ASAP. " % (credType,
credFile)
warnMsg += "Its time left is: %.2f hours;" % (secsLeft / 3600.)
agInfo['proxy_warning'] = agInfo.get('proxy_warning', "") + warnMsg
logging.warning(warnMsg)
return
def buildMonITDocs(self, dataStats):
"""
Convert agent statistics into MonIT-friendly documents to be posted
to AMQ/ES. It creates 5 different type of documents:
* priority information
* site information
* work information
* agent information
* agent health information
Note that the internal methods are popping some metrics out of dataStats
"""
if not self.postToAMQ:
return
logging.info("Preparing documents to be posted to AMQ/MonIT..")
allDocs = self._buildMonITPrioDocs(dataStats)
allDocs.extend(self._buildMonITSitesDocs(dataStats))
allDocs.extend(self._buildMonITWorkDocs(dataStats))
allDocs.extend(self._buildMonITWMBSDocs(dataStats))
allDocs.extend(self._buildMonITAgentDocs(dataStats))
allDocs.extend(self._buildMonITHealthDocs(dataStats))
allDocs.extend(self._buildMonITSummaryDocs(dataStats))
# and finally post them all to AMQ
logging.info("Found %d documents to post to AMQ", len(allDocs))
self.uploadToAMQ(allDocs, dataStats['agent_url'],
dataStats['timestamp'])
def _buildMonITPrioDocs(self, dataStats):
"""
Uses the `sitePendCountByPrio` metric in order to build documents
reporting the site name, job priority and amount of jobs within that
priority.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_prio_info MonIT docs
"""
docType = "wma_prio_info"
prioDocs = []
sitePendCountByPrio = dataStats['WMBS_INFO'].pop(
'sitePendCountByPrio', [])
for site, item in viewitems(sitePendCountByPrio):
# it seems sites with no jobs are also always here as "Sitename": {0: 0}
if list(item) == [0]:
continue
for prio, jobs in viewitems(item):
prioDoc = {}
prioDoc['site_name'] = site
prioDoc['type'] = docType
prioDoc['priority'] = prio
prioDoc['job_count'] = jobs
prioDocs.append(prioDoc)
return prioDocs
def _buildMonITSitesDocs(self, dataStats):
"""
Uses the site thresholds and job information for each site in order
to build a `site_info` document type for MonIT.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_site_info MonIT docs
"""
docType = "wma_site_info"
siteDocs = []
thresholds = dataStats['WMBS_INFO'].pop('thresholds', {})
thresholdsGQ2LQ = dataStats['WMBS_INFO'].pop('thresholdsGQ2LQ', {})
if self.isT0agent:
possibleJobsPerSite = {}
uniqueJobsPerSite = {}
else:
possibleJobsPerSite = dataStats['LocalWQ_INFO'].pop(
'possibleJobsPerSite', {})
uniqueJobsPerSite = dataStats['LocalWQ_INFO'].pop(
'uniqueJobsPerSite', {})
for site in sorted(thresholds):
siteDoc = {}
siteDoc['site_name'] = site
siteDoc['type'] = docType
siteDoc['thresholds'] = thresholds[site]
siteDoc['state'] = siteDoc['thresholds'].pop('state', 'Unknown')
siteDoc['thresholdsGQ2LQ'] = thresholdsGQ2LQ.get(site, 0)
for status in possibleJobsPerSite:
# make sure these keys are always present in the documents
jobKey = "possible_%s_jobs" % status.lower()
elemKey = "num_%s_elem" % status.lower()
uniJobKey = "unique_%s_jobs" % status.lower()
siteDoc[jobKey], siteDoc[elemKey], siteDoc[uniJobKey] = 0, 0, 0
if site in possibleJobsPerSite[status]:
siteDoc[jobKey] = possibleJobsPerSite[status][site][
'sum_jobs']
siteDoc[elemKey] = possibleJobsPerSite[status][site][
'num_elem']
if site in uniqueJobsPerSite[status]:
siteDoc[uniJobKey] = uniqueJobsPerSite[status][site][
'sum_jobs']
siteDocs.append(siteDoc)
return siteDocs
def _buildMonITWorkDocs(self, dataStats):
"""
Uses the local workqueue information order by WQE status and build
statistics for the workload in terms of workqueue elements and top
level jobs.
Using the WMBS data, also builds documents to show the amount of
work in 'created' and 'executing' WMBS status.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_work_info MonIT docs
"""
workDocs = []
if self.isT0agent:
return workDocs
docType = "wma_work_info"
workByStatus = dataStats['LocalWQ_INFO'].pop('workByStatus', {})
for status, info in viewitems(workByStatus):
workDoc = {}
workDoc['type'] = docType
workDoc['status'] = status
workDoc['num_elem'] = info.get('num_elem', 0)
workDoc['sum_jobs'] = info.get('sum_jobs', 0)
workDocs.append(workDoc)
return workDocs
def _buildMonITWMBSDocs(self, dataStats):
"""
Using the WMBS data, builds documents to show the amount of work in
'created' and 'executing' WMBS status.
It also builds a document for every single wmbs_status in the database.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_wmbs_info and wma_wmbs_state_info docs
"""
docType = "wma_wmbs_info"
wmbsDocs = []
wmbsCreatedTypeCount = dataStats['WMBS_INFO'].pop(
'wmbsCreatedTypeCount', {})
wmbsExecutingTypeCount = dataStats['WMBS_INFO'].pop(
'wmbsExecutingTypeCount', {})
for jobType in wmbsCreatedTypeCount:
wmbsDoc = {}
wmbsDoc['type'] = docType
wmbsDoc['job_type'] = jobType
wmbsDoc['created_jobs'] = wmbsCreatedTypeCount[jobType]
wmbsDoc['executing_jobs'] = wmbsExecutingTypeCount[jobType]
wmbsDocs.append(wmbsDoc)
docType = "wma_wmbs_state_info"
wmbsCountByState = dataStats['WMBS_INFO'].pop('wmbsCountByState', {})
for wmbsStatus in wmbsCountByState:
wmbsDoc = {}
wmbsDoc['type'] = docType
wmbsDoc['wmbs_status'] = wmbsStatus
wmbsDoc['num_jobs'] = wmbsCountByState[wmbsStatus]
wmbsDocs.append(wmbsDoc)
return wmbsDocs
def _buildMonITAgentDocs(self, dataStats):
"""
Uses the BossAir and WMBS table information in order to build a
view of amount of jobs in different statuses.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_agent_info MonIT docs
"""
docType = "wma_agent_info"
agentDocs = []
activeRunJobByStatus = dataStats['WMBS_INFO'].pop(
'activeRunJobByStatus', {})
completeRunJobByStatus = dataStats['WMBS_INFO'].pop(
'completeRunJobByStatus', {})
for schedStatus in activeRunJobByStatus:
agentDoc = {}
agentDoc['type'] = docType
agentDoc['schedd_status'] = schedStatus
agentDoc['active_jobs'] = activeRunJobByStatus[schedStatus]
agentDoc['completed_jobs'] = completeRunJobByStatus[schedStatus]
agentDocs.append(agentDoc)
return agentDocs
def _buildMonITHealthDocs(self, dataStats):
"""
Creates documents with specific agent information, status of
each component and worker thread (similar to what is shown in
wmstats) and also some very basic performance numbers.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_health_info MonIT docs
"""
docType = "wma_health_info"
healthDocs = []
workersStatus = dataStats.pop('workers', {})
for worker in workersStatus:
healthDoc = {}
healthDoc['type'] = docType
healthDoc['worker_name'] = worker['name']
healthDoc['worker_state'] = worker['state']
healthDoc['worker_poll'] = worker['poll_interval']
healthDoc['worker_last_hb'] = worker['last_updated']
healthDoc['worker_cycle_time'] = worker['cycle_time']
healthDocs.append(healthDoc)
return healthDocs
def _buildMonITSummaryDocs(self, dataStats):
"""
Creates a document with the very basic agent info used
in the wmstats monitoring tab.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_health_info MonIT docs
"""
docType = "wma_summary_info"
summaryDocs = []
summaryDoc = {}
summaryDoc['type'] = docType
summaryDoc['agent_team'] = dataStats['agent_team']
summaryDoc['agent_version'] = dataStats['agent_version']
summaryDoc['agent_status'] = dataStats['status']
if not self.isT0agent:
summaryDoc['wq_query_time'] = dataStats['LocalWQ_INFO'][
'total_query_time']
summaryDoc['wmbs_query_time'] = dataStats['WMBS_INFO'][
'total_query_time']
summaryDoc['drain_mode'] = dataStats['drain_mode']
summaryDoc['down_components'] = dataStats['down_components']
summaryDocs.append(summaryDoc)
return summaryDocs
def uploadToAMQ(self, docs, agentUrl, timeS):
"""
_uploadToAMQ_
Sends data to AMQ, which ends up in the MonIT infrastructure.
:param docs: list of documents/dicts to be posted
"""
if not docs:
logging.info("There are no documents to send to AMQ")
return
# add mandatory information for every single document
for doc in docs:
doc['agent_url'] = agentUrl
docType = "cms_%s_info" % self.producer
notifications = []
logging.debug("Sending the following data to AMQ %s", pformat(docs))
try:
stompSvc = StompAMQ(username=self.userAMQ,
password=self.passAMQ,
producer=self.producer,
topic=self.topicAMQ,
validation_schema=None,
host_and_ports=self.hostPortAMQ,
logger=logging)
for doc in docs:
singleNotif, _, _ = stompSvc.make_notification(
payload=doc,
docType=docType,
ts=timeS,
dataSubfield="payload")
notifications.append(singleNotif)
failures = stompSvc.send(notifications)
msg = "%i out of %i documents successfully sent to AMQ" % (
len(notifications) - len(failures), len(notifications))
logging.info(msg)
except Exception as ex:
logging.exception("Failed to send data to StompAMQ. Error %s",
str(ex))
return
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = (config.AnalyticsDataCollector.summaryLevel).lower()
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({'source': wmstatsSource, 'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"})
#TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({'source': t0Source, 'target': t0Target,
'filter': "T0Request/repfilter"})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams["ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {'childUrl' : childURL, 'parentUrl' : sanitizeURL(parentQURL)['url']}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({'source': sanitizeURL(parentQURL)['url'], 'target': localQInboxURL,
'filter': wqfilter, 'query_params': query_params})
self.replicatorDocs.append({'source': sanitizeURL(localQInboxURL)['url'], 'target': parentQURL,
'filter': wqfilter, 'query_params': query_params})
# delete or replicator docs befor setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(
rp['source'], rp['target'], filter = rp['filter'],
query_params = rp.get('query_params', False),
continuous = True, useReplicator = True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
if hasattr(self.config, "Tier0Feeder"):
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.localWMStatsURL,
appName= "WMStatsAgent")
else:
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
logging.info("Getting Agent info ...")
agentInfo = self.collectAgentInfo()
#set the uploadTime - should be the same for all docs
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
logging.info("Agent components down:\n %s" % agentInfo['down_components'])
logging.info("Agent in drain mode:\n %s \nsleep for next WMStats alarm updating cycle"
% agentInfo['drain_mode'])
except Exception as ex:
logging.error("Error occurred, will retry later:")
logging.error(str(ex))
logging.error("Trace back: \n%s" % traceback.format_exc())
def collectCouchDBInfo(self):
couchInfo = {'status': 'ok', 'error_message': ""}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
msg = ""
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(rp['source'],
rp['target'], checkUpdateSeq = False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = msg
return couchInfo
def collectAgentInfo(self):
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['status'] = "warning"
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if (couchInfo['status'] != 'ok'):
agentInfo['down_components'].append("CouchServer")
agentInfo['status'] = couchInfo['status']
couchInfo['name'] = "CouchServer"
agentInfo['down_component_detail'].append(couchInfo)
# Disk space warning
diskUseList = diskUse()
diskUseThreshold = float(self.config.AnalyticsDataCollector.diskUseThreshold)
agentInfo['disk_warning'] = []
for disk in diskUseList:
if float(disk['percent'].strip('%')) >= diskUseThreshold and disk['mounted'] not in self.config.AnalyticsDataCollector.ignoreDisk:
agentInfo['disk_warning'].append(disk)
# Couch process warning
couchProc = numberCouchProcess()
couchProcessThreshold = float(self.config.AnalyticsDataCollector.couchProcessThreshold)
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo(self)
if lastDataUpload['data_last_update']!=0:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error']!="":
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok':
if agentInfo['disk_warning'] != []:
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if ('data_error' in agentInfo and agentInfo['data_error'] != 'ok') or \
('couch_process_warning' in agentInfo and agentInfo['couch_process_warning'] != 0):
agentInfo['status'] = "error"
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
#direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC, uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
proxyArgs = {'logger': logging.getLogger()}
self.proxy = Proxy(proxyArgs)
self.proxyFile = self.proxy.getProxyFilename() # X509_USER_PROXY
self.userCertFile = self.proxy.getUserCertFilename() # X509_USER_CERT
# credential lifetime warning/error thresholds, in days
self.credThresholds = {'proxy': {'error': 3, 'warning': 5},
'certificate': {'error': 10, 'warning': 20}}
# Monitoring setup
self.userAMQ = getattr(config.AgentStatusWatcher, "userAMQ", None)
self.passAMQ = getattr(config.AgentStatusWatcher, "passAMQ", None)
self.postToAMQ = getattr(config.AgentStatusWatcher, "enableAMQ", False)
self.topicAMQ = getattr(config.AgentStatusWatcher, "topicAMQ", None)
self.hostPortAMQ = getattr(config.AgentStatusWatcher, "hostPortAMQ", [('cms-mb.cern.ch', 61313)])
# T0 doesn't have WorkQueue, so some monitoring/replication code has to be skipped here
if hasattr(self.config, "Tier0Feeder"):
self.isT0agent = True
self.producer = "tier0wmagent"
else:
self.isT0agent = False
self.producer = "wmagent"
localWQUrl = config.AnalyticsDataCollector.localQueueURL
self.workqueueDS = WorkQueueDS(localWQUrl)
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.General.centralWMStatsURL
self.replicatorDocs.append({'source': wmstatsSource, 'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"})
if self.isT0agent:
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({'source': t0Source, 'target': t0Target,
'filter': "T0Request/repfilter"})
else:
# set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams["ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {'childUrl': childURL, 'parentUrl': sanitizeURL(parentQURL)['url']}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({'source': sanitizeURL(parentQURL)['url'], 'target': localQInboxURL,
'filter': wqfilter, 'query_params': query_params})
self.replicatorDocs.append({'source': sanitizeURL(localQInboxURL)['url'], 'target': parentQURL,
'filter': wqfilter, 'query_params': query_params})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(
rp['source'], rp['target'], filter=rp['filter'],
query_params=rp.get('query_params', False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.General.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
@timeFunction
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
self.checkCredLifetime(agentInfo, "proxy")
self.checkCredLifetime(agentInfo, "certificate")
timeSpent, wmbsInfo, _ = self.collectWMBSInfo()
wmbsInfo['total_query_time'] = int(timeSpent)
agentInfo["WMBS_INFO"] = wmbsInfo
logging.info("WMBS data collected in: %d secs", timeSpent)
if not self.isT0agent:
timeSpent, localWQInfo, _ = self.collectWorkQueueInfo()
localWQInfo['total_query_time'] = int(timeSpent)
agentInfo["LocalWQ_INFO"] = localWQInfo
logging.info("Local WorkQueue data collected in: %d secs", timeSpent)
self.uploadAgentInfoToCentralWMStats(agentInfo)
self.buildMonITDocs(agentInfo)
except Exception as ex:
logging.exception("Error occurred, will retry later.\nDetails: %s", str(ex))
@timeFunction
def collectWorkQueueInfo(self):
"""
Collect information from local workqueue database
:return:
"""
results = {}
wqStates = ['Available', 'Acquired']
results['workByStatus'] = self.workqueueDS.getJobsByStatus()
results['workByStatusAndPriority'] = self.workqueueDS.getJobsByStatusAndPriority()
elements = self.workqueueDS.getElementsByStatus(wqStates)
uniSites, posSites = getGlobalSiteStatusSummary(elements, status=wqStates, dataLocality=True)
results['uniqueJobsPerSite'] = uniSites
results['possibleJobsPerSite'] = posSites
return results
def collectCouchDBInfo(self):
couchInfo = {'name': 'CouchServer', 'status': 'ok', 'error_message': ""}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(rp['source'],
rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
agentInfo['disk_warning'] = listDiskUsageOverThreshold(self.config, updateDB=True)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['drain_stats'] = DrainStatusPoller.getDrainInfo()
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode'] or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get('couch_process_warning', 0):
agentInfo['status'] = "error"
logging.info("List of agent components down: %s", agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo):
"""
Add some required fields to the document before it can get uploaded
to WMStats.
:param agentInfo: dict with agent stats to be posted to couchdb
"""
agentInfo['_id'] = agentInfo["agent_url"]
agentInfo['timestamp'] = int(time.time())
agentInfo['type'] = "agent_info"
# directly upload to the remote to prevent data conflict when agent is cleaned up and redeployed
try:
self.centralWMStatsCouchDB.updateAgentInfo(agentInfo,
propertiesToKeep=["data_last_update", "data_error"])
except Exception as e:
logging.error("Failed to upload agent statistics to WMStats. Error: %s", str(e))
@timeFunction
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
logging.info("Getting wmbs job info ...")
results = {}
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.debug("Running and pending site thresholds: %s", results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
logging.debug("Total number of jobs in WMBS sorted by status: %s", results['wmbsCountByState'])
logging.debug("Total number of 'created' jobs in WMBS sorted by type: %s", results['wmbsCreatedTypeCount'])
logging.debug("Total number of 'executing' jobs in WMBS sorted by type: %s", results['wmbsExecutingTypeCount'])
logging.debug("Total number of active jobs in BossAir sorted by status: %s", results['activeRunJobByStatus'])
logging.debug("Total number of complete jobs in BossAir sorted by status: %s",
results['completeRunJobByStatus'])
logging.debug("Available slots thresholds to pull work from GQ to LQ: %s", results['thresholdsGQ2LQ'])
logging.debug("List of jobs pending for each site, sorted by priority: %s", results['sitePendCountByPrio'])
return results
def checkCredLifetime(self, agInfo, credType):
"""
Check the credential lifetime. Usually X509_USER_PROXY or X509_USER_CERT
and raise either a warning or an error if the proxy validity is about to expire.
:param agInfo: dictionary with plenty of agent monitoring information in place.
:param credType: credential type, can be: "proxy" or "certificate"
:return: same dictionary object plus additional keys/values if needed.
"""
if credType == "proxy":
credFile = self.proxyFile
secsLeft = self.proxy.getTimeLeft(proxy=credFile)
elif credType == "certificate":
credFile = self.userCertFile
secsLeft = self.proxy.getUserCertTimeLeft(openSSL=True)
else:
logging.error("Unknown credential type. Available options are: [proxy, certificate]")
return
logging.debug("%s '%s' lifetime is %d seconds", credType, credFile, secsLeft)
daysLeft = secsLeft / (60. * 60 * 24)
if daysLeft <= self.credThresholds[credType]['error']:
credWarning = True
agInfo['status'] = "error"
elif daysLeft <= self.credThresholds[credType]['warning']:
credWarning = True
if agInfo['status'] == "ok":
agInfo['status'] = "warning"
else:
credWarning = False
if credWarning:
warnMsg = "Agent %s '%s' must be renewed ASAP. " % (credType, credFile)
warnMsg += "Its time left is: %.2f hours;" % (secsLeft / 3600.)
agInfo['proxy_warning'] = agInfo.get('proxy_warning', "") + warnMsg
logging.warning(warnMsg)
return
def buildMonITDocs(self, dataStats):
"""
Convert agent statistics into MonIT-friendly documents to be posted
to AMQ/ES. It creates 5 different type of documents:
* priority information
* site information
* work information
* agent information
* agent health information
Note that the internal methods are popping some metrics out of dataStats
"""
if not self.postToAMQ:
return
logging.info("Preparing documents to be posted to AMQ/MonIT..")
allDocs = self._buildMonITPrioDocs(dataStats)
allDocs.extend(self._buildMonITSitesDocs(dataStats))
allDocs.extend(self._buildMonITWorkDocs(dataStats))
allDocs.extend(self._buildMonITWMBSDocs(dataStats))
allDocs.extend(self._buildMonITAgentDocs(dataStats))
allDocs.extend(self._buildMonITHealthDocs(dataStats))
allDocs.extend(self._buildMonITSummaryDocs(dataStats))
# and finally post them all to AMQ
logging.info("Found %d documents to post to AMQ", len(allDocs))
self.uploadToAMQ(allDocs, dataStats['agent_url'], dataStats['timestamp'])
def _buildMonITPrioDocs(self, dataStats):
"""
Uses the `sitePendCountByPrio` metric in order to build documents
reporting the site name, job priority and amount of jobs within that
priority.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_prio_info MonIT docs
"""
docType = "wma_prio_info"
prioDocs = []
sitePendCountByPrio = dataStats['WMBS_INFO'].pop('sitePendCountByPrio', [])
for site, item in sitePendCountByPrio.iteritems():
# it seems sites with no jobs are also always here as "Sitename": {0: 0}
if item.keys() == [0]:
continue
for prio, jobs in item.iteritems():
prioDoc = {}
prioDoc['site_name'] = site
prioDoc['type'] = docType
prioDoc['priority'] = prio
prioDoc['job_count'] = jobs
prioDocs.append(prioDoc)
return prioDocs
def _buildMonITSitesDocs(self, dataStats):
"""
Uses the site thresholds and job information for each site in order
to build a `site_info` document type for MonIT.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_site_info MonIT docs
"""
docType = "wma_site_info"
siteDocs = []
thresholds = dataStats['WMBS_INFO'].pop('thresholds', {})
thresholdsGQ2LQ = dataStats['WMBS_INFO'].pop('thresholdsGQ2LQ', {})
if self.isT0agent:
possibleJobsPerSite = {}
uniqueJobsPerSite = {}
else:
possibleJobsPerSite = dataStats['LocalWQ_INFO'].pop('possibleJobsPerSite', {})
uniqueJobsPerSite = dataStats['LocalWQ_INFO'].pop('uniqueJobsPerSite', {})
for site in sorted(thresholds):
siteDoc = {}
siteDoc['site_name'] = site
siteDoc['type'] = docType
siteDoc['thresholds'] = thresholds[site]
siteDoc['state'] = siteDoc['thresholds'].pop('state', 'Unknown')
siteDoc['thresholdsGQ2LQ'] = thresholdsGQ2LQ.get(site, 0)
for status in possibleJobsPerSite.keys():
# make sure these keys are always present in the documents
jobKey = "possible_%s_jobs" % status.lower()
elemKey = "num_%s_elem" % status.lower()
uniJobKey = "unique_%s_jobs" % status.lower()
siteDoc[jobKey], siteDoc[elemKey], siteDoc[uniJobKey] = 0, 0, 0
if site in possibleJobsPerSite[status]:
siteDoc[jobKey] = possibleJobsPerSite[status][site]['sum_jobs']
siteDoc[elemKey] = possibleJobsPerSite[status][site]['num_elem']
if site in uniqueJobsPerSite[status]:
siteDoc[uniJobKey] = uniqueJobsPerSite[status][site]['sum_jobs']
siteDocs.append(siteDoc)
return siteDocs
def _buildMonITWorkDocs(self, dataStats):
"""
Uses the local workqueue information order by WQE status and build
statistics for the workload in terms of workqueue elements and top
level jobs.
Using the WMBS data, also builds documents to show the amount of
work in 'created' and 'executing' WMBS status.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_work_info MonIT docs
"""
workDocs = []
if self.isT0agent:
return workDocs
docType = "wma_work_info"
workByStatus = dataStats['LocalWQ_INFO'].pop('workByStatus', {})
for status, info in workByStatus.items():
workDoc = {}
workDoc['type'] = docType
workDoc['status'] = status
workDoc['num_elem'] = info.get('num_elem', 0)
workDoc['sum_jobs'] = info.get('sum_jobs', 0)
workDocs.append(workDoc)
return workDocs
def _buildMonITWMBSDocs(self, dataStats):
"""
Using the WMBS data, builds documents to show the amount of work in
'created' and 'executing' WMBS status.
It also builds a document for every single wmbs_status in the database.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_wmbs_info and wma_wmbs_state_info docs
"""
docType = "wma_wmbs_info"
wmbsDocs = []
wmbsCreatedTypeCount = dataStats['WMBS_INFO'].pop('wmbsCreatedTypeCount', {})
wmbsExecutingTypeCount = dataStats['WMBS_INFO'].pop('wmbsExecutingTypeCount', {})
for jobType in wmbsCreatedTypeCount:
wmbsDoc = {}
wmbsDoc['type'] = docType
wmbsDoc['job_type'] = jobType
wmbsDoc['created_jobs'] = wmbsCreatedTypeCount[jobType]
wmbsDoc['executing_jobs'] = wmbsExecutingTypeCount[jobType]
wmbsDocs.append(wmbsDoc)
docType = "wma_wmbs_state_info"
wmbsCountByState = dataStats['WMBS_INFO'].pop('wmbsCountByState', {})
for wmbsStatus in wmbsCountByState:
wmbsDoc = {}
wmbsDoc['type'] = docType
wmbsDoc['wmbs_status'] = wmbsStatus
wmbsDoc['num_jobs'] = wmbsCountByState[wmbsStatus]
wmbsDocs.append(wmbsDoc)
return wmbsDocs
def _buildMonITAgentDocs(self, dataStats):
"""
Uses the BossAir and WMBS table information in order to build a
view of amount of jobs in different statuses.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_agent_info MonIT docs
"""
docType = "wma_agent_info"
agentDocs = []
activeRunJobByStatus = dataStats['WMBS_INFO'].pop('activeRunJobByStatus', {})
completeRunJobByStatus = dataStats['WMBS_INFO'].pop('completeRunJobByStatus', {})
for schedStatus in activeRunJobByStatus:
agentDoc = {}
agentDoc['type'] = docType
agentDoc['schedd_status'] = schedStatus
agentDoc['active_jobs'] = activeRunJobByStatus[schedStatus]
agentDoc['completed_jobs'] = completeRunJobByStatus[schedStatus]
agentDocs.append(agentDoc)
return agentDocs
def _buildMonITHealthDocs(self, dataStats):
"""
Creates documents with specific agent information, status of
each component and worker thread (similar to what is shown in
wmstats) and also some very basic performance numbers.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_health_info MonIT docs
"""
docType = "wma_health_info"
healthDocs = []
workersStatus = dataStats.pop('workers', {})
for worker in workersStatus:
healthDoc = {}
healthDoc['type'] = docType
healthDoc['worker_name'] = worker['name']
healthDoc['worker_state'] = worker['state']
healthDoc['worker_poll'] = worker['poll_interval']
healthDoc['worker_last_hb'] = worker['last_updated']
healthDoc['worker_cycle_time'] = worker['cycle_time']
healthDocs.append(healthDoc)
return healthDocs
def _buildMonITSummaryDocs(self, dataStats):
"""
Creates a document with the very basic agent info used
in the wmstats monitoring tab.
:param dataStats: dictionary with metrics previously posted to WMStats
:return: list of dictionaries with the wma_health_info MonIT docs
"""
docType = "wma_summary_info"
summaryDocs = []
summaryDoc = {}
summaryDoc['type'] = docType
summaryDoc['agent_team'] = dataStats['agent_team']
summaryDoc['agent_version'] = dataStats['agent_version']
summaryDoc['agent_status'] = dataStats['status']
if not self.isT0agent:
summaryDoc['wq_query_time'] = dataStats['LocalWQ_INFO']['total_query_time']
summaryDoc['wmbs_query_time'] = dataStats['WMBS_INFO']['total_query_time']
summaryDoc['drain_mode'] = dataStats['drain_mode']
summaryDoc['down_components'] = dataStats['down_components']
summaryDocs.append(summaryDoc)
return summaryDocs
def uploadToAMQ(self, docs, agentUrl, timeS):
"""
_uploadToAMQ_
Sends data to AMQ, which ends up in the MonIT infrastructure.
:param docs: list of documents/dicts to be posted
"""
if not docs:
logging.info("There are no documents to send to AMQ")
return
# add mandatory information for every single document
for doc in docs:
doc['agent_url'] = agentUrl
docType = "cms_%s_info" % self.producer
logging.debug("Sending the following data to AMQ %s", pformat(docs))
try:
stompSvc = StompAMQ(username=self.userAMQ,
password=self.passAMQ,
producer=self.producer,
topic=self.topicAMQ,
host_and_ports=self.hostPortAMQ,
logger=logging)
notifications = [stompSvc.make_notification(payload=doc, docType=docType, ts=timeS,
dataSubfield="payload") for doc in docs]
failures = stompSvc.send(notifications)
logging.info("%i docs successfully sent to AMQ", len(notifications) - len(failures))
except Exception as ex:
logging.exception("Failed to send data to StompAMQ. Error %s", str(ex))
return
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = (
config.AnalyticsDataCollector.summaryLevel).lower()
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({
'source': wmstatsSource,
'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"
})
#TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({
'source': t0Source,
'target': t0Target,
'filter': "T0Request/repfilter"
})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams[
"ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {
'childUrl': childURL,
'parentUrl': sanitizeURL(parentQURL)['url']
}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({
'source': sanitizeURL(parentQURL)['url'],
'target': localQInboxURL,
'filter': wqfilter,
'query_params': query_params
})
self.replicatorDocs.append({
'source':
sanitizeURL(localQInboxURL)['url'],
'target':
parentQURL,
'filter':
wqfilter,
'query_params':
query_params
})
# delete or replicator docs befor setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(rp['source'],
rp['target'],
filter=rp['filter'],
query_params=rp.get(
'query_params',
False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi,
myThread.logger)
if hasattr(self.config, "Tier0Feeder"):
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.localWMStatsURL,
appName="WMStatsAgent")
else:
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(
self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
logging.info("Getting Agent info ...")
agentInfo = self.collectAgentInfo()
#set the uploadTime - should be the same for all docs
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
logging.info("Agent components down:\n %s" %
agentInfo['down_components'])
logging.info(
"Agent in drain mode:\n %s \nsleep for next WMStats alarm updating cycle"
% agentInfo['drain_mode'])
except Exception as ex:
logging.error("Error occurred, will retry later:")
logging.error(str(ex))
logging.error("Trace back: \n%s" % traceback.format_exc())
def collectCouchDBInfo(self):
couchInfo = {'status': 'ok', 'error_message': ""}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
msg = ""
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(
rp['source'], rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = msg
return couchInfo
def collectAgentInfo(self):
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['status'] = "warning"
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if (couchInfo['status'] != 'ok'):
agentInfo['down_components'].append("CouchServer")
agentInfo['status'] = couchInfo['status']
couchInfo['name'] = "CouchServer"
agentInfo['down_component_detail'].append(couchInfo)
# Disk space warning
diskUseList = diskUse()
diskUseThreshold = float(
self.config.AnalyticsDataCollector.diskUseThreshold)
agentInfo['disk_warning'] = []
for disk in diskUseList:
if float(disk['percent'].strip('%')) >= diskUseThreshold and disk[
'mounted'] not in self.config.AnalyticsDataCollector.ignoreDisk:
agentInfo['disk_warning'].append(disk)
# Couch process warning
couchProc = numberCouchProcess()
couchProcessThreshold = float(
self.config.AnalyticsDataCollector.couchProcessThreshold)
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo(self)
if lastDataUpload['data_last_update'] != 0:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error'] != "":
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok':
if agentInfo['disk_warning'] != []:
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if ('data_error' in agentInfo and agentInfo['data_error'] != 'ok') or \
('couch_process_warning' in agentInfo and agentInfo['couch_process_warning'] != 0):
agentInfo['status'] = "error"
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
#direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC,
uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
self.jsonFile = config.AgentStatusWatcher.jsonFile
proxyArgs = {'logger': logging.getLogger()}
self.proxy = Proxy(proxyArgs)
self.proxyFile = self.proxy.getProxyFilename() # X509_USER_PROXY
localWQUrl = config.AnalyticsDataCollector.localQueueURL
self.workqueueDS = WorkQueueDS(localWQUrl)
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({
'source': wmstatsSource,
'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"
})
# TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({
'source': t0Source,
'target': t0Target,
'filter': "T0Request/repfilter"
})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams[
"ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {
'childUrl': childURL,
'parentUrl': sanitizeURL(parentQURL)['url']
}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({
'source': sanitizeURL(parentQURL)['url'],
'target': localQInboxURL,
'filter': wqfilter,
'query_params': query_params
})
self.replicatorDocs.append({
'source':
sanitizeURL(localQInboxURL)['url'],
'target':
parentQURL,
'filter':
wqfilter,
'query_params':
query_params
})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(rp['source'],
rp['target'],
filter=rp['filter'],
query_params=rp.get(
'query_params',
False),
continuous=True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi,
myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(
self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(
self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
self.checkProxyLifetime(agentInfo)
timeSpent, wmbsInfo, _ = self.collectWMBSInfo()
wmbsInfo['total_query_time'] = int(timeSpent)
agentInfo["WMBS_INFO"] = wmbsInfo
logging.info("WMBS data collected in: %d secs", timeSpent)
if not hasattr(self.config, "Tier0Feeder"):
# Tier0 Agent doesn't have LQ.
timeSpent, localWQInfo, _ = self.collectWorkQueueInfo()
localWQInfo['total_query_time'] = int(timeSpent)
agentInfo["LocalWQ_INFO"] = localWQInfo
logging.info("Local WorkQueue data collected in: %d secs",
timeSpent)
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
# save locally json file as well
with open(self.jsonFile, 'w') as outFile:
json.dump(agentInfo, outFile, indent=2)
except Exception as ex:
logging.exception("Error occurred, will retry later.\nDetails: %s",
str(ex))
@timeFunction
def collectWorkQueueInfo(self):
"""
Collect information from local workqueue database
:return:
"""
results = {}
results['workByStatus'] = self.workqueueDS.getJobsByStatus()
results[
'workByStatusAndPriority'] = self.workqueueDS.getJobsByStatusAndPriority(
)
elements = self.workqueueDS.getElementsByStatus(
['Available', 'Acquired'])
uniSites, posSites = getGlobalSiteStatusSummary(elements,
dataLocality=True)
results['uniqueJobsPerSite'] = uniSites
results['possibleJobsPerSite'] = posSites
return results
def collectCouchDBInfo(self):
couchInfo = {
'name': 'CouchServer',
'status': 'ok',
'error_message': ""
}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(
rp['source'], rp['target'], checkUpdateSeq=False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
agentInfo['disk_warning'] = listDiskUsageOverThreshold(self.config,
updateDB=True)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
agentInfo['drain_stats'] = DrainStatusPoller.getDrainInfo()
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo()
if lastDataUpload['data_last_update']:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error']:
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode']
or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get(
'couch_process_warning', 0):
agentInfo['status'] = "error"
logging.info("List of agent components down: %s",
agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
# direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC,
uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
@timeFunction
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
logging.info("Getting wmbs job info ...")
results = {}
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.debug("Running and pending site thresholds: %s",
results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
logging.debug("Total number of jobs in WMBS sorted by status: %s",
results['wmbsCountByState'])
logging.debug(
"Total number of 'created' jobs in WMBS sorted by type: %s",
results['wmbsCreatedTypeCount'])
logging.debug(
"Total number of 'executing' jobs in WMBS sorted by type: %s",
results['wmbsExecutingTypeCount'])
logging.debug(
"Total number of active jobs in BossAir sorted by status: %s",
results['activeRunJobByStatus'])
logging.debug(
"Total number of complete jobs in BossAir sorted by status: %s",
results['completeRunJobByStatus'])
logging.debug(
"Available slots thresholds to pull work from GQ to LQ: %s",
results['thresholdsGQ2LQ'])
logging.debug(
"List of jobs pending for each site, sorted by priority: %s",
results['sitePendCountByPrio'])
return results
def checkProxyLifetime(self, agInfo):
"""
Check the proxy lifetime (usually X509_USER_CERT) and raise either
a warning or an error if the proxy validity is about to expire.
:param agInfo: dictionary with plenty of agent monitoring information in place.
:return: same dictionary object plus additional keys/values if needed.
"""
secsLeft = self.proxy.getTimeLeft(proxy=self.proxyFile)
logging.debug("Proxy '%s' lifetime is %d secs", self.proxyFile,
secsLeft)
if secsLeft <= 86400 * 3: # 3 days
proxyWarning = True
agInfo['status'] = "error"
elif secsLeft <= 86400 * 5: # 5 days
proxyWarning = True
if agInfo['status'] == "ok":
agInfo['status'] = "warning"
else:
proxyWarning = False
if proxyWarning:
warnMsg = "Agent proxy '%s' must be renewed ASAP. " % self.proxyFile
warnMsg += "Its time left is: %.2f hours." % (secsLeft / 3600.)
agInfo['proxy_warning'] = warnMsg
return
class AgentStatusPoller(BaseWorkerThread):
"""
Gether the summary data for request (workflow) from local queue,
local job couchdb, wmbs/boss air and populate summary db for monitoring
"""
def __init__(self, config):
"""
initialize properties specified from config
"""
BaseWorkerThread.__init__(self)
# set the workqueue service for REST call
self.config = config
# need to get campaign, user, owner info
self.agentInfo = initAgentInfo(self.config)
self.summaryLevel = config.AnalyticsDataCollector.summaryLevel
self.jsonFile = config.AgentStatusWatcher.jsonFile
def setUpCouchDBReplication(self):
self.replicatorDocs = []
# set up common replication code
wmstatsSource = self.config.JobStateMachine.jobSummaryDBName
wmstatsTarget = self.config.AnalyticsDataCollector.centralWMStatsURL
self.replicatorDocs.append({'source': wmstatsSource, 'target': wmstatsTarget,
'filter': "WMStatsAgent/repfilter"})
#TODO: tier0 specific code - need to make it generic
if hasattr(self.config, "Tier0Feeder"):
t0Source = self.config.Tier0Feeder.requestDBName
t0Target = self.config.AnalyticsDataCollector.centralRequestDBURL
self.replicatorDocs.append({'source': t0Source, 'target': t0Target,
'filter': "T0Request/repfilter"})
else: # set up workqueue replication
wqfilter = 'WorkQueue/queueFilter'
parentQURL = self.config.WorkQueueManager.queueParams["ParentQueueCouchUrl"]
childURL = self.config.WorkQueueManager.queueParams["QueueURL"]
query_params = {'childUrl' : childURL, 'parentUrl' : sanitizeURL(parentQURL)['url']}
localQInboxURL = "%s_inbox" % self.config.AnalyticsDataCollector.localQueueURL
self.replicatorDocs.append({'source': sanitizeURL(parentQURL)['url'], 'target': localQInboxURL,
'filter': wqfilter, 'query_params': query_params})
self.replicatorDocs.append({'source': sanitizeURL(localQInboxURL)['url'], 'target': parentQURL,
'filter': wqfilter, 'query_params': query_params})
# delete old replicator docs before setting up
self.localCouchMonitor.deleteReplicatorDocs()
for rp in self.replicatorDocs:
self.localCouchMonitor.couchServer.replicate(
rp['source'], rp['target'], filter = rp['filter'],
query_params = rp.get('query_params', False),
continuous = True)
# First cicle need to be skipped since document is not updated that fast
self.skipReplicationCheck = True
def setup(self, parameters):
"""
set db connection(couchdb, wmbs) to prepare to gather information
"""
# interface to WMBS/BossAir db
myThread = threading.currentThread()
# set wmagent db data
self.wmagentDB = WMAgentDBData(self.summaryLevel, myThread.dbi, myThread.logger)
self.centralWMStatsCouchDB = WMStatsWriter(self.config.AnalyticsDataCollector.centralWMStatsURL)
self.localCouchMonitor = CouchMonitor(self.config.JobStateMachine.couchurl)
self.setUpCouchDBReplication()
def algorithm(self, parameters):
"""
get information from wmbs, workqueue and local couch
"""
try:
agentInfo = self.collectAgentInfo()
#set the uploadTime - should be the same for all docs
wmbsInfo = self.collectWMBSInfo()
logging.info("finished collecting agent/wmbs info")
agentInfo["WMBS_INFO"] = wmbsInfo
uploadTime = int(time.time())
self.uploadAgentInfoToCentralWMStats(agentInfo, uploadTime)
#save locally json file as well
with open(self.jsonFile, 'w') as outFile:
json.dump(agentInfo, outFile, indent=2)
except Exception as ex:
logging.error("Error occurred, will retry later:")
logging.error(str(ex))
logging.error("Trace back: \n%s" % traceback.format_exc())
def collectCouchDBInfo(self):
couchInfo = {'name': 'CouchServer', 'status': 'ok', 'error_message': ""}
if self.skipReplicationCheck:
# skipping the check this round set if False so it can be checked next round.
self.skipReplicationCheck = False
return couchInfo
for rp in self.replicatorDocs:
cInfo = self.localCouchMonitor.checkCouchServerStatus(rp['source'],
rp['target'], checkUpdateSeq = False)
if cInfo['status'] != 'ok':
couchInfo['status'] = 'error'
couchInfo['error_message'] = cInfo['error_message']
return couchInfo
def collectAgentInfo(self):
"""
Monitors the general health of the agent, as:
1. status of the agent processes
2. status of the agent threads based on the database info
3. couchdb active tasks and its replications
4. check the disk usage
5. check the number of couch processes
:return: a dict with all the info collected
"""
logging.info("Getting agent info ...")
agentInfo = self.wmagentDB.getComponentStatus(self.config)
agentInfo.update(self.agentInfo)
if isDrainMode(self.config):
logging.info("Agent is in DrainMode")
agentInfo['drain_mode'] = True
else:
agentInfo['drain_mode'] = False
couchInfo = self.collectCouchDBInfo()
if couchInfo['status'] != 'ok':
agentInfo['down_components'].append(couchInfo['name'])
agentInfo['status'] = couchInfo['status']
agentInfo['down_component_detail'].append(couchInfo)
# Disk space warning
diskUseList = diskUse()
diskUseThreshold = float(self.config.AnalyticsDataCollector.diskUseThreshold)
agentInfo['disk_warning'] = []
for disk in diskUseList:
if float(disk['percent'].strip('%')) >= diskUseThreshold and \
disk['mounted'] not in self.config.AnalyticsDataCollector.ignoreDisk:
agentInfo['disk_warning'].append(disk)
# Couch process warning
couchProc = numberCouchProcess()
logging.info("CouchDB is running with %d processes", couchProc)
couchProcessThreshold = self.config.AnalyticsDataCollector.couchProcessThreshold
if couchProc >= couchProcessThreshold:
agentInfo['couch_process_warning'] = couchProc
else:
agentInfo['couch_process_warning'] = 0
# This adds the last time and message when data was updated to agentInfo
lastDataUpload = DataUploadTime.getInfo()
if lastDataUpload['data_last_update']:
agentInfo['data_last_update'] = lastDataUpload['data_last_update']
if lastDataUpload['data_error']:
agentInfo['data_error'] = lastDataUpload['data_error']
# Change status if there is data_error, couch process maxed out or disk full problems.
if agentInfo['status'] == 'ok' and (agentInfo['drain_mode'] or agentInfo['disk_warning']):
agentInfo['status'] = "warning"
if agentInfo['status'] == 'ok' or agentInfo['status'] == 'warning':
if agentInfo.get('data_error', 'ok') != 'ok' or agentInfo.get('couch_process_warning', 0):
agentInfo['status'] = "error"
if agentInfo['down_components']:
logging.info("List of agent components down: %s" % agentInfo['down_components'])
return agentInfo
def uploadAgentInfoToCentralWMStats(self, agentInfo, uploadTime):
#direct data upload to the remote to prevent data conflict when agent is cleaned up and redeployed
agentDocs = convertToAgentCouchDoc(agentInfo, self.config.ACDC, uploadTime)
self.centralWMStatsCouchDB.updateAgentInfo(agentDocs)
def collectWMBSInfo(self):
"""
Fetches WMBS job information.
In addition to WMBS, also collects RunJob info from BossAir
:return: dict with the number of jobs in each status
"""
logging.info("Getting wmbs job info ...")
results = {}
start = int(time.time())
# first retrieve the site thresholds
results['thresholds'] = self.wmagentDB.getJobSlotInfo()
logging.debug("Running and pending site thresholds: %s", results['thresholds'])
# now fetch the amount of jobs in each state and the amount of created
# jobs grouped by task
results.update(self.wmagentDB.getAgentMonitoring())
end = int(time.time())
#adding total query time
results["total_query_time"] = end - start
logging.debug("Total number of jobs in WMBS sorted by status: %s", results['wmbsCountByState'])
logging.debug("Total number of 'created' jobs in WMBS sorted by type: %s", results['wmbsCreatedTypeCount'])
logging.debug("Total number of 'executing' jobs in WMBS sorted by type: %s", results['wmbsExecutingTypeCount'])
logging.debug("Total number of active jobs in BossAir sorted by status: %s", results['activeRunJobByStatus'])
logging.debug("Total number of complete jobs in BossAir sorted by status: %s", results['completeRunJobByStatus'])
logging.debug("Available slots thresholds to pull work from GQ to LQ: %s", results['thresholdsGQ2LQ'])
logging.debug("List of jobs pending for each site, sorted by priority: %s", results['sitePendCountByPrio'])
return results
