class DaemonStatsTest(BaseTestCase):
"""Test the DaemonStats"""
def setUp(self):
self.daemon_stats = DaemonStats()
def testDaemonsTagsServiceId(self):
os.environ["CONTROLPLANE"] = "1"
os.environ["CONTROLPLANE_SERVICE_ID"] = "ID"
os.environ["CONTROLPLANE_TENANT_ID"] = "foo"
os.environ["CONTROLPLANE_INSTANCE_ID"] = "bar"
self.daemon_stats.config( "name", "monitor", None, None, None)
self.assertEqual(
{'daemon': 'name', 'instance': 'bar', 'internal': True,
'monitor': 'monitor', 'metricType': 'type', 'serviceId': 'ID',
'tenantId': 'foo'},
self.daemon_stats._tags("type")
)
def testDaemonsDoesNotTagServiceId(self):
if "CONTROLPLANE" in os.environ:
del os.environ["CONTROLPLANE"]
if "CONTROLPLANE_SERVICE_ID" in os.environ:
del os.environ["CONTROLPLANE_SERVICE_ID"]
self.daemon_stats.config( "name", "monitor", None, None, None)
self.assertEqual(
{'daemon': 'name', 'internal': True, 'monitor': 'monitor', 'metricType': 'type'},
self.daemon_stats._tags("type")
)
def get_rrd_stats(self, hub_config, send_event):
rrd_stats = DaemonStats()
thresholds = hub_config.getThresholdInstances(BuiltInDS.sourcetype)
threshold_notifier = ThresholdNotifier(send_event, thresholds)
derivative_tracker = DerivativeTracker()
rrd_stats.config(
'zenhub',
hub_config.id,
self.metric_writer,
threshold_notifier,
derivative_tracker
)
return rrd_stats
def get_rrd_stats(self, hub_config, send_event):
rrd_stats = DaemonStats()
thresholds = hub_config.getThresholdInstances(BuiltInDS.sourcetype)
threshold_notifier = ThresholdNotifier(send_event, thresholds)
derivative_tracker = DerivativeTracker()
rrd_stats.config(
'zenhub',
hub_config.id,
self.metric_writer,
threshold_notifier,
derivative_tracker
)
return rrd_stats
def getRRDStats(self):
"""
Return the most recent RRD statistic information.
"""
rrdStats = DaemonStats()
perfConf = self._getConf()
from Products.ZenModel.BuiltInDS import BuiltInDS
threshs = perfConf.getThresholdInstances(BuiltInDS.sourcetype)
threshold_notifier = ThresholdNotifier(self.zem.sendEvent, threshs)
derivative_tracker = DerivativeTracker()
rrdStats.config('zenhub', perfConf.id, self._metric_writer,
threshold_notifier, derivative_tracker)
return rrdStats
class DaemonStatsTest(BaseTestCase):
"""Test the DaemonStats"""
def setUp(self):
self.daemon_stats = DaemonStats()
def testDaemonsTagsServiceId(self):
os.environ["CONTROLPLANE"] = "1"
os.environ["CONTROLPLANE_SERVICE_ID"] = "ID"
os.environ["CONTROLPLANE_TENANT_ID"] = "foo"
os.environ["CONTROLPLANE_INSTANCE_ID"] = "bar"
self.daemon_stats.config("name", "monitor", None, None, None)
self.assertEqual(
{
'daemon': 'name',
'instance': 'bar',
'internal': True,
'monitor': 'monitor',
'metricType': 'type',
'serviceId': 'ID',
'tenantId': 'foo'
}, self.daemon_stats._tags("type"))
def testDaemonsDoesNotTagServiceId(self):
if "CONTROLPLANE" in os.environ:
del os.environ["CONTROLPLANE"]
if "CONTROLPLANE_SERVICE_ID" in os.environ:
del os.environ["CONTROLPLANE_SERVICE_ID"]
self.daemon_stats.config("name", "monitor", None, None, None)
self.assertEqual(
{
'daemon': 'name',
'internal': True,
'monitor': 'monitor',
'metricType': 'type'
}, self.daemon_stats._tags("type"))
class ZenModeler(PBDaemon):
"""
Daemon class to attach to zenhub and pass along
device configuration information.
"""
name = 'zenmodeler'
initialServices = PBDaemon.initialServices + ['ModelerService']
generateEvents = True
configCycleInterval = 360
classCollectorPlugins = ()
def __init__(self, single=False):
"""
Initalizer
@param single: collect from a single device?
@type single: boolean
"""
PBDaemon.__init__(self)
# FIXME: cleanup --force option #2660
self.options.force = True
self.start = None
self.startat = None
self.rrdStats = DaemonStats()
self.single = single
if self.options.device:
self.single = True
self.modelerCycleInterval = self.options.cycletime
# get the minutes and convert to fraction of a day
self.collage = float(self.options.collage) / 1440.0
self.pendingNewClients = False
self.clients = []
self.finished = []
self.devicegen = None
self.counters = collections.Counter()
self.configFilter = None
self.configLoaded = False
# Make sendEvent() available to plugins
zope.component.provideUtility(self, IEventService)
# Delay start for between 10 and 60 seconds when run as a daemon.
self.started = False
self.startDelay = 0
self.immediate = 1
if self.options.daemon or self.options.cycle:
if self.options.now:
self.log.debug('option "now" specified, starting immediately.')
else:
# self.startDelay = randint(10, 60) * 60
self.startDelay = randint(10, 60) * 1
self.immediate = 0
self.log.info(
'option "now" not specified, waiting %s seconds to start.'
% self.startDelay)
else:
self.log.debug("Run in foreground, starting immediately.")
# ZEN-26637
self.collectorLoopIteration = 0
self.mainLoopGotDeviceList = False
self._modeledDevicesMetric = Metrology.meter(
"zenmodeler.modeledDevices")
self._failuresMetric = Metrology.counter("zenmodeler.failures")
def reportError(self, error):
"""
Log errors that have occurred
@param error: error message
@type error: string
"""
self.log.error("Error occured: %s", error)
def connected(self):
"""
Called after connected to the zenhub service
"""
reactor.callLater(_CONFIG_PULLING_TIMEOUT, self._checkConfigLoad)
d = self.configure()
d.addCallback(self.heartbeat)
d.addErrback(self.reportError)
def _checkConfigLoad(self):
"""
Looping call to check whether zenmodeler got configuration
from ZenHub.
"""
if not self.configLoaded:
self.log.info("Modeling has not started pending configuration "
"pull from ZenHub. Is ZenHub overloaded?")
reactor.callLater(_CONFIG_PULLING_TIMEOUT, self._checkConfigLoad)
def configure(self):
"""
Get our configuration from zenhub
"""
# add in the code to fetch cycle time, etc.
self.log.info("Getting configuration from ZenHub...")
def inner(driver):
"""
Generator function to gather our configuration
@param driver: driver object
@type driver: driver object
"""
self.log.debug('fetching monitor properties')
yield self.config().callRemote('propertyItems')
items = dict(driver.next())
# If the cycletime option is not specified or zero, then use the
# modelerCycleInterval value in the database.
if not self.options.cycletime:
self.modelerCycleInterval = items.get('modelerCycleInterval',
_DEFAULT_CYCLE_INTERVAL)
self.configCycleInterval = items.get('configCycleInterval',
self.configCycleInterval)
reactor.callLater(self.configCycleInterval * 60, self.configure)
self.log.debug("Getting threshold classes...")
yield self.config().callRemote('getThresholdClasses')
self.remote_updateThresholdClasses(driver.next())
self.log.debug("Getting collector thresholds...")
yield self.config().callRemote('getCollectorThresholds')
thresholds = driver.next()
threshold_notifier = ThresholdNotifier(self.sendEvent, thresholds)
self.rrdStats.config(self.name, self.options.monitor,
self.metricWriter(), threshold_notifier,
self.derivativeTracker())
self.log.debug("Getting collector plugins for each DeviceClass")
yield self.config().callRemote('getClassCollectorPlugins')
self.classCollectorPlugins = driver.next()
self.configLoaded = True
return drive(inner)
def config(self):
"""
Get the ModelerService
"""
return self.services.get('ModelerService', FakeRemote())
def selectPlugins(self, device, transport):
"""
Build a list of active plugins for a device, based on:
* the --collect command-line option which is a regex
* the --ignore command-line option which is a regex
* transport which is a string describing the type of plugin
@param device: device to collect against
@type device: string
@param transport: python, ssh, snmp, telnet, cmd
@type transport: string
@return: results of the plugin
@type: string
@todo: determine if an event for the collector AND the device should be sent
"""
plugins = []
valid_loaders = []
for loader in device.plugins:
try:
plugin = loader.create()
self.log.debug("Loaded plugin %s" % plugin.name())
plugins.append(plugin)
valid_loaders.append(loader)
except (SystemExit, KeyboardInterrupt), ex:
self.log.info("Interrupted by external signal (%s)" % str(ex))
raise
except Plugins.PluginImportError, import_error:
import socket
component, _ = os.path.splitext(os.path.basename(sys.argv[0]))
collector_host = socket.gethostname()
# NB: an import errror affects all devices,
# so report the issue against the collector
# TOOD: determine if an event for the collector AND the device should be sent
evt = {
"eventClass": "/Status/Update",
"component": component,
"agent": collector_host,
"device": collector_host,
"severity": Error
}
info = "Problem loading plugin %s" % import_error.plugin
self.log.error(info)
evt['summary'] = info
info = import_error.traceback
self.log.error(info)
evt['message'] = info
info = ("Due to import errors, removing the %s plugin"
" from this collection cycle.") % import_error.plugin
self.log.error(info)
evt['message'] += "%s\n" % info
self.sendEvent(evt)
class PBDaemon(ZenDaemon, pb.Referenceable):
name = 'pbdaemon'
initialServices = ['EventService']
heartbeatEvent = {'eventClass':Heartbeat}
heartbeatTimeout = 60*3
_customexitcode = 0
_pushEventsDeferred = None
_eventHighWaterMark = None
_healthMonitorInterval = 30
def __init__(self, noopts=0, keeproot=False, name=None):
# if we were provided our collector name via the constructor instead of
# via code, be sure to store it correctly.
if name is not None:
self.name = name
self.mname = name
try:
ZenDaemon.__init__(self, noopts, keeproot)
except IOError:
import traceback
self.log.critical(traceback.format_exc(0))
sys.exit(1)
self._thresholds = None
self._threshold_notifier = None
self.rrdStats = DaemonStats()
self.lastStats = 0
self.perspective = None
self.services = {}
self.eventQueueManager = EventQueueManager(self.options, self.log)
self.startEvent = startEvent.copy()
self.stopEvent = stopEvent.copy()
details = dict(component=self.name, device=self.options.monitor)
for evt in self.startEvent, self.stopEvent, self.heartbeatEvent:
evt.update(details)
self.initialConnect = defer.Deferred()
self.stopped = False
self.counters = collections.Counter()
self._pingedZenhub = None
self._connectionTimeout = None
self._publisher = None
self._internal_publisher = None
self._metric_writer = None
self._derivative_tracker = None
self._metrologyReporter = None
# Add a shutdown trigger to send a stop event and flush the event queue
reactor.addSystemEventTrigger('before', 'shutdown', self._stopPbDaemon)
# Set up a looping call to support the health check.
self.healthMonitor = task.LoopingCall(self._checkZenHub)
self.healthMonitor.start(self._healthMonitorInterval)
def publisher(self):
if not self._publisher:
host, port = urlparse(self.options.redisUrl).netloc.split(':')
try:
port = int(port)
except ValueError:
self.log.exception("redis url contains non-integer port " +
"value {port}, defaulting to {default}".
format(port=port, default=publisher.defaultRedisPort))
port = publisher.defaultRedisPort
self._publisher = publisher.RedisListPublisher(
host, port, self.options.metricBufferSize,
channel=self.options.metricsChannel, maxOutstandingMetrics=self.options.maxOutstandingMetrics
)
return self._publisher
def internalPublisher(self):
if not self._internal_publisher:
url = os.environ.get( "CONTROLPLANE_CONSUMER_URL", None)
username = os.environ.get( "CONTROLPLANE_CONSUMER_USERNAME", "")
password = os.environ.get( "CONTROLPLANE_CONSUMER_PASSWORD", "")
if url:
self._internal_publisher = publisher.HttpPostPublisher( username, password, url)
return self._internal_publisher
def metricWriter(self):
if not self._metric_writer:
publisher = self.publisher()
metric_writer = MetricWriter(publisher)
if os.environ.get( "CONTROLPLANE", "0") == "1":
internal_publisher = self.internalPublisher()
if internal_publisher:
internal_metric_filter = lambda metric, value, timestamp, tags:\
tags and tags.get("internal", False)
internal_metric_writer = FilteredMetricWriter(internal_publisher, internal_metric_filter)
self._metric_writer = AggregateMetricWriter( [metric_writer, internal_metric_writer])
else:
self._metric_writer = metric_writer
return self._metric_writer
def derivativeTracker(self):
if not self._derivative_tracker:
self._derivative_tracker = DerivativeTracker()
return self._derivative_tracker
def connecting(self):
"""
Called when about to connect to zenhub
"""
self.log.info("Attempting to connect to zenhub")
def getZenhubInstanceId(self):
"""
Called after we connected to zenhub.
"""
def callback(result):
self.log.info("Connected to the zenhub/%s instance", result)
def errback(result):
self.log.info("Unexpected error appeared while getting zenhub instance number %s", result)
d = self.perspective.callRemote('getHubInstanceId')
d.addCallback(callback)
d.addErrback(errback)
return d
def gotPerspective(self, perspective):
"""
This gets called every time we reconnect.
@parameter perspective: Twisted perspective object
@type perspective: Twisted perspective object
"""
self.perspective = perspective
self.getZenhubInstanceId()
# Cancel the connection timeout timer as it's no longer needed.
if self._connectionTimeout:
try:
self._connectionTimeout.cancel()
except AlreadyCalled:
pass
self._connectionTimeout = None
d2 = self.getInitialServices()
if self.initialConnect:
self.log.debug('Chaining getInitialServices with d2')
self.initialConnect, d = None, self.initialConnect
d2.chainDeferred(d)
def connect(self):
pingInterval = self.options.zhPingInterval
factory = ReconnectingPBClientFactory(connectTimeout=60, pingPerspective=self.options.pingPerspective,
pingInterval=pingInterval, pingtimeout=pingInterval * 5)
self.log.info("Connecting to %s:%d" % (self.options.hubhost, self.options.hubport))
factory.connectTCP(self.options.hubhost, self.options.hubport)
username = self.options.hubusername
password = self.options.hubpassword
self.log.debug("Logging in as %s" % username)
c = credentials.UsernamePassword(username, password)
factory.gotPerspective = self.gotPerspective
factory.connecting = self.connecting
factory.setCredentials(c)
def timeout(d):
if not d.called:
self.connectTimeout()
self._connectionTimeout = reactor.callLater(
self.options.hubtimeout, timeout, self.initialConnect)
return self.initialConnect
def connectTimeout(self):
self.log.error('Timeout connecting to zenhub: is it running?')
pass
def eventService(self):
return self.getServiceNow('EventService')
def getServiceNow(self, svcName):
if not svcName in self.services:
self.log.warning('No service named %r: ZenHub may be disconnected' % svcName)
return self.services.get(svcName, None) or FakeRemote()
def getService(self, serviceName, serviceListeningInterface=None):
"""
Attempt to get a service from zenhub. Returns a deferred.
When service is retrieved it is stashed in self.services with
serviceName as the key. When getService is called it will first
check self.services and if serviceName is already there it will return
the entry from self.services wrapped in a defer.succeed
"""
if serviceName in self.services:
return defer.succeed(self.services[serviceName])
def removeService(ignored):
self.log.debug('Removing service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
def callback(result, serviceName):
self.log.debug('Loaded service %s from zenhub' % serviceName)
self.services[serviceName] = result
result.notifyOnDisconnect(removeService)
return result
def errback(error, serviceName):
self.log.debug('errback after getting service %s' % serviceName)
self.log.error('Could not retrieve service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
return error
d = self.perspective.callRemote('getService',
serviceName,
self.options.monitor,
serviceListeningInterface or self,
self.options.__dict__)
d.addCallback(callback, serviceName)
d.addErrback(errback, serviceName)
return d
def getInitialServices(self):
"""
After connecting to zenhub, gather our initial list of services.
"""
def errback(error):
if isinstance(error, Failure):
self.log.critical( "Invalid monitor: %s" % self.options.monitor)
reactor.stop()
return defer.fail(RemoteBadMonitor(
"Invalid monitor: %s" % self.options.monitor, ''))
return error
self.log.debug('Setting up initial services: %s' % \
', '.join(self.initialServices))
d = defer.DeferredList(
[self.getService(name) for name in self.initialServices],
fireOnOneErrback=True, consumeErrors=True)
d.addErrback(errback)
return d
def connected(self):
pass
def _getThresholdNotifier(self):
if not self._threshold_notifier:
self._threshold_notifier = ThresholdNotifier(self.sendEvent, self.getThresholds())
return self._threshold_notifier
def getThresholds(self):
if not self._thresholds:
self._thresholds = Thresholds()
return self._thresholds
def run(self):
def stopReporter():
if self._metrologyReporter:
return self._metrologyReporter.stop()
# Order of the shutdown triggers matter. Want to stop reporter first, calling self.metricWriter() below
# registers shutdown triggers for the actual metric http and redis publishers.
reactor.addSystemEventTrigger('before', 'shutdown', stopReporter)
threshold_notifier = self._getThresholdNotifier()
self.rrdStats.config(self.name,
self.options.monitor,
self.metricWriter(),
threshold_notifier,
self.derivativeTracker())
self.log.debug('Starting PBDaemon initialization')
d = self.connect()
def callback(result):
self.sendEvent(self.startEvent)
self.pushEventsLoop()
self.log.debug('Calling connected.')
self.connected()
return result
def startStatsLoop():
self.log.debug("Starting Statistic posting")
loop = task.LoopingCall(self.postStatistics)
loop.start(self.options.writeStatistics, now=False)
daemonTags = {
'zenoss_daemon': self.name,
'zenoss_monitor': self.options.monitor,
'internal': True
}
self._metrologyReporter = TwistedMetricReporter(self.options.writeStatistics, self.metricWriter(), daemonTags)
self._metrologyReporter.start()
if self.options.cycle:
reactor.callWhenRunning(startStatsLoop)
d.addCallback(callback)
d.addErrback(twisted.python.log.err)
reactor.run()
if self._customexitcode:
sys.exit(self._customexitcode)
def setExitCode(self, exitcode):
self._customexitcode = exitcode
def stop(self, ignored=''):
if reactor.running:
try:
reactor.stop()
except ReactorNotRunning:
self.log.debug("Tried to stop reactor that was stopped")
else:
self.log.debug("stop() called when not running")
def _stopPbDaemon(self):
if self.stopped:
return
self.stopped = True
if 'EventService' in self.services:
# send stop event if we don't have an implied --cycle,
# or if --cycle has been specified
if not hasattr(self.options, 'cycle') or \
getattr(self.options, 'cycle', True):
self.sendEvent(self.stopEvent)
self.log.debug("Sent a 'stop' event")
if self._pushEventsDeferred:
self.log.debug("Currently sending events. Queueing next call")
d = self._pushEventsDeferred
# Schedule another call to flush any additional queued events
d.addBoth(lambda unused: self.pushEvents())
else:
d = self.pushEvents()
return d
self.log.debug("No event sent as no EventService available.")
def sendEvents(self, events):
map(self.sendEvent, events)
def sendEvent(self, event, **kw):
""" Add event to queue of events to be sent. If we have an event
service then process the queue.
"""
generatedEvent = self.generateEvent(event, **kw)
self.eventQueueManager.addEvent(generatedEvent)
self.counters['eventCount'] += 1
if self._eventHighWaterMark:
return self._eventHighWaterMark
elif self.eventQueueManager.event_queue_length >= self.options.maxqueuelen * self.options.queueHighWaterMark:
return self.pushEvents()
else:
return defer.succeed(None)
def generateEvent(self, event, **kw):
""" Add event to queue of events to be sent. If we have an event
service then process the queue.
"""
if not reactor.running:
return
eventCopy = {}
for k, v in chain(event.items(), kw.items()):
if isinstance(v, basestring):
#default max size is 512k
size = LIMITS.get(k, DEFAULT_LIMIT)
eventCopy[k] = v[0:size] if len(v)>size else v
else:
eventCopy[k] = v
eventCopy['agent'] = self.name
eventCopy['monitor'] = self.options.monitor
eventCopy['manager'] = self.fqdn
return eventCopy
@defer.inlineCallbacks
def pushEventsLoop(self):
"""Periodially, wake up and flush events to ZenHub.
"""
reactor.callLater(self.options.eventflushseconds, self.pushEventsLoop)
yield self.pushEvents()
# Record the number of events in the queue up to every 2 seconds.
now = time.time()
if self.rrdStats.name and now >= (self.lastStats + 2):
self.lastStats = now
self.rrdStats.gauge(
'eventQueueLength', self.eventQueueManager.event_queue_length)
@defer.inlineCallbacks
def pushEvents(self):
"""Flush events to ZenHub.
"""
# are we already shutting down?
if not reactor.running:
self.log.debug("Skipping event sending - reactor not running.")
return
if self.eventQueueManager.event_queue_length >= self.options.maxqueuelen * self.options.queueHighWaterMark and not self._eventHighWaterMark:
self.log.debug("Queue length exceeded high water mark, %s ;creating high water mark deferred", self.eventQueueManager.event_queue_length)
self._eventHighWaterMark = defer.Deferred()
# are still connected to ZenHub?
evtSvc = self.services.get('EventService', None)
if not evtSvc:
self.log.error("No event service: %r", evtSvc)
yield task.deferLater(reactor, 0, lambda:None)
if self._eventHighWaterMark:
d, self._eventHighWaterMark = self._eventHighWaterMark, None
#not connected, release throttle and let things queue
d.callback("No Event Service")
defer.returnValue(None)
if self._pushEventsDeferred:
self.log.debug("Skipping event sending - previous call active.")
defer.returnValue("Push Pending")
sent = 0
try:
#only set _pushEventsDeferred after we know we have an evtSvc/connectivity
self._pushEventsDeferred = defer.Deferred()
def repush(val):
if self.eventQueueManager.event_queue_length >= self.options.eventflushchunksize:
self.pushEvents()
return val
# conditionally push more events after this pushEvents call finishes
self._pushEventsDeferred.addCallback(repush)
discarded_events = self.eventQueueManager.discarded_events
if discarded_events:
self.log.error(
'Discarded oldest %d events because maxqueuelen was '
'exceeded: %d/%d',
discarded_events,
discarded_events + self.options.maxqueuelen,
self.options.maxqueuelen)
self.counters['discardedEvents'] += discarded_events
self.eventQueueManager.discarded_events = 0
send_events_fn = partial(evtSvc.callRemote, 'sendEvents')
try:
sent = yield self.eventQueueManager.sendEvents(send_events_fn)
except ConnectionLost as ex:
self.log.error('Error sending event: %s', ex)
#let the reactor have time to clean up any connection errors and make callbacks
yield task.deferLater(reactor, 0, lambda:None)
except Exception as ex:
self.log.exception(ex)
#let the reactor have time to clean up any connection errors and make callbacks
yield task.deferLater(reactor, 0, lambda:None)
finally:
if self._pushEventsDeferred:
d, self._pushEventsDeferred = self._pushEventsDeferred, None
d.callback('sent %s' % sent)
if self._eventHighWaterMark and self.eventQueueManager.event_queue_length < self.options.maxqueuelen * self.options.queueHighWaterMark:
self.log.debug("Queue restored to below high water mark: %s", self.eventQueueManager.event_queue_length)
d, self._eventHighWaterMark = self._eventHighWaterMark, None
d.callback("Queue length below high water mark")
def heartbeat(self):
"""if cycling, send a heartbeat, else, shutdown"""
if not self.options.cycle:
self.stop()
return
heartbeatEvent = self.generateEvent(self.heartbeatEvent, timeout=self.heartbeatTimeout)
self.eventQueueManager.addHeartbeatEvent(heartbeatEvent)
# heartbeat is normally 3x cycle time
self.niceDoggie(self.heartbeatTimeout / 3)
def postStatisticsImpl(self):
pass
def postStatistics(self):
# save daemon counter stats
for name, value in self.counters.items():
self.log.info("Counter %s, value %d", name, value)
self.rrdStats.counter(name, value)
# persist counters values
self.postStatisticsImpl()
def _pickleName(self):
instance_id = os.environ.get('CONTROLPLANE_INSTANCE_ID')
return 'var/%s_%s_counters.pickle' % (self.name, instance_id)
def remote_getName(self):
return self.name
def remote_shutdown(self, unused):
self.stop()
self.sigTerm()
def remote_setPropertyItems(self, items):
pass
@translateError
def remote_updateThresholdClasses(self, classes):
from Products.ZenUtils.Utils import importClass
self.log.debug("Loading classes %s", classes)
for c in classes:
try:
importClass(c)
except ImportError:
self.log.error("Unable to import class %s", c)
def _checkZenHub(self):
"""
Check status of ZenHub (using ping method of service).
@return: if ping occurs, return deferred with result of ping attempt.
"""
self.log.debug('_checkZenHub: entry')
def callback(result):
self.log.debug('ZenHub health check: Got result %s' % result)
if result == 'pong':
self.log.debug('ZenHub health check: Success - received pong from ZenHub ping service.')
self._signalZenHubAnswering(True)
else:
self.log.error('ZenHub health check did not respond as expected.')
self._signalZenHubAnswering(False)
def errback(error):
self.log.error('Error pinging ZenHub: %s (%s).' % (error, getattr(error, 'message', '')))
self._signalZenHubAnswering(False)
try:
if self.perspective:
self.log.debug('ZenHub health check: perspective found. attempting remote ping call.')
d = self.perspective.callRemote('ping')
d.addCallback(callback)
d.addErrback(errback)
return d
else:
self.log.debug('ZenHub health check: ZenHub may be down.')
self._signalZenHubAnswering(False)
except pb.DeadReferenceError:
self.log.warning("ZenHub health check: DeadReferenceError - lost connection to ZenHub.")
self._signalZenHubAnswering(False)
except Exception as e:
self.log.error('ZenHub health check: caught %s exception: %s' % (e.__class__, e.message))
self._signalZenHubAnswering(False)
def _signalZenHubAnswering(self, answering):
"""
Write or remove file that the ZenHub_answering health check uses to report status.
@param answering: true if ZenHub is answering, False, otherwise.
"""
self.log.debug('_signalZenHubAnswering(%s)' % answering)
filename = 'zenhub_connected'
signalFilePath = zenPath('var', filename)
if answering:
self.log.debug('writing file at %s' % signalFilePath)
atomicWrite(signalFilePath, '')
else:
try:
self.log.debug('removing file at %s' % signalFilePath)
os.remove(signalFilePath)
except Exception as e:
self.log.debug('ignoring %s exception (%s) removing file %s' % (e.__class__, e.message, signalFilePath))
def buildOptions(self):
ZenDaemon.buildOptions(self)
self.parser.add_option('--hubhost',
dest='hubhost',
default=DEFAULT_HUB_HOST,
help='Host of zenhub daemon.'
' Default is %s.' % DEFAULT_HUB_HOST)
self.parser.add_option('--hubport',
dest='hubport',
type='int',
default=DEFAULT_HUB_PORT,
help='Port zenhub listens on.'
'Default is %s.' % DEFAULT_HUB_PORT)
self.parser.add_option('--hubusername',
dest='hubusername',
default=DEFAULT_HUB_USERNAME,
help='Username for zenhub login.'
' Default is %s.' % DEFAULT_HUB_USERNAME)
self.parser.add_option('--hubpassword',
dest='hubpassword',
default=DEFAULT_HUB_PASSWORD,
help='Password for zenhub login.'
' Default is %s.' % DEFAULT_HUB_PASSWORD)
self.parser.add_option('--monitor',
dest='monitor',
default=DEFAULT_HUB_MONITOR,
help='Name of monitor instance to use for'
' configuration. Default is %s.'
% DEFAULT_HUB_MONITOR)
self.parser.add_option('--initialHubTimeout',
dest='hubtimeout',
type='int',
default=30,
help='Initial time to wait for a ZenHub '
'connection')
self.parser.add_option('--allowduplicateclears',
dest='allowduplicateclears',
default=False,
action='store_true',
help='Send clear events even when the most '
'recent event was also a clear event.')
self.parser.add_option('--duplicateclearinterval',
dest='duplicateclearinterval',
default=0,
type='int',
help=('Send a clear event every [DUPLICATECLEARINTEVAL] '
'events.')
)
self.parser.add_option('--eventflushseconds',
dest='eventflushseconds',
default=5.,
type='float',
help='Seconds between attempts to flush '
'events to ZenHub.')
self.parser.add_option('--eventflushchunksize',
dest='eventflushchunksize',
default=50,
type='int',
help='Number of events to send to ZenHub'
'at one time')
self.parser.add_option('--maxqueuelen',
dest='maxqueuelen',
default=5000,
type='int',
help='Maximum number of events to queue')
self.parser.add_option('--queuehighwatermark',
dest='queueHighWaterMark',
default=0.75,
type='float',
help='The size, in percent, of the event queue when event pushback starts')
self.parser.add_option('--zenhubpinginterval',
dest='zhPingInterval',
default=120,
type='int',
help='How often to ping zenhub')
self.parser.add_option('--disable-event-deduplication',
dest='deduplicate_events',
default=True,
action='store_false',
help='Disable event de-duplication')
self.parser.add_option('--redis-url',
dest='redisUrl',
type='string',
default='redis://localhost:{default}/0'.format(default=publisher.defaultRedisPort),
help='redis connection string: redis://[hostname]:[port]/[db], default: %default')
self.parser.add_option('--metricBufferSize',
dest='metricBufferSize',
type='int',
default=publisher.defaultMetricBufferSize,
help='Number of metrics to buffer if redis goes down')
self.parser.add_option('--metricsChannel',
dest='metricsChannel',
type='string',
default=publisher.defaultMetricsChannel,
help='redis channel to which metrics are published')
self.parser.add_option('--maxOutstandingMetrics',
dest='maxOutstandingMetrics',
type='int',
default=publisher.defaultMaxOutstandingMetrics,
help='Max Number of metrics to allow in redis')
self.parser.add_option('--disable-ping-perspective',
dest='pingPerspective',
help="Enable or disable ping perspective",
default=True,
action='store_false')
self.parser.add_option('--writeStatistics',
dest='writeStatistics',
type='int',
default=30,
help='How often to write internal statistics value in seconds')
class PBDaemon(ZenDaemon, pb.Referenceable):
name = 'pbdaemon'
initialServices = ['EventService']
heartbeatEvent = {'eventClass': Heartbeat}
heartbeatTimeout = 60 * 3
_customexitcode = 0
_pushEventsDeferred = None
_eventHighWaterMark = None
_healthMonitorInterval = 30
def __init__(self, noopts=0, keeproot=False, name=None):
# if we were provided our collector name via the constructor instead of
# via code, be sure to store it correctly.
if name is not None:
self.name = name
self.mname = name
try:
ZenDaemon.__init__(self, noopts, keeproot)
except IOError:
import traceback
self.log.critical(traceback.format_exc(0))
sys.exit(1)
self._thresholds = None
self._threshold_notifier = None
self.rrdStats = DaemonStats()
self.lastStats = 0
self.perspective = None
self.services = {}
self.eventQueueManager = EventQueueManager(self.options, self.log)
self.startEvent = startEvent.copy()
self.stopEvent = stopEvent.copy()
details = dict(component=self.name, device=self.options.monitor)
for evt in self.startEvent, self.stopEvent, self.heartbeatEvent:
evt.update(details)
self.initialConnect = defer.Deferred()
self.stopped = False
self.counters = collections.Counter()
self._pingedZenhub = None
self._connectionTimeout = None
self._publisher = None
self._internal_publisher = None
self._metric_writer = None
self._derivative_tracker = None
self._metrologyReporter = None
# Add a shutdown trigger to send a stop event and flush the event queue
reactor.addSystemEventTrigger('before', 'shutdown', self._stopPbDaemon)
# Set up a looping call to support the health check.
self.healthMonitor = task.LoopingCall(self._checkZenHub)
self.healthMonitor.start(self._healthMonitorInterval)
def publisher(self):
if not self._publisher:
host, port = urlparse(self.options.redisUrl).netloc.split(':')
try:
port = int(port)
except ValueError:
self.log.exception(
"redis url contains non-integer port " +
"value {port}, defaulting to {default}".format(
port=port, default=publisher.defaultRedisPort))
port = publisher.defaultRedisPort
self._publisher = publisher.RedisListPublisher(
host,
port,
self.options.metricBufferSize,
channel=self.options.metricsChannel,
maxOutstandingMetrics=self.options.maxOutstandingMetrics)
return self._publisher
def internalPublisher(self):
if not self._internal_publisher:
url = os.environ.get("CONTROLPLANE_CONSUMER_URL", None)
username = os.environ.get("CONTROLPLANE_CONSUMER_USERNAME", "")
password = os.environ.get("CONTROLPLANE_CONSUMER_PASSWORD", "")
if url:
self._internal_publisher = publisher.HttpPostPublisher(
username, password, url)
return self._internal_publisher
def metricWriter(self):
if not self._metric_writer:
publisher = self.publisher()
metric_writer = MetricWriter(publisher)
if os.environ.get("CONTROLPLANE", "0") == "1":
internal_publisher = self.internalPublisher()
if internal_publisher:
internal_metric_filter = lambda metric, value, timestamp, tags:\
tags and tags.get("internal", False)
internal_metric_writer = FilteredMetricWriter(
internal_publisher, internal_metric_filter)
self._metric_writer = AggregateMetricWriter(
[metric_writer, internal_metric_writer])
else:
self._metric_writer = metric_writer
return self._metric_writer
def derivativeTracker(self):
if not self._derivative_tracker:
self._derivative_tracker = DerivativeTracker()
return self._derivative_tracker
def connecting(self):
"""
Called when about to connect to zenhub
"""
self.log.info("Attempting to connect to zenhub")
def getZenhubInstanceId(self):
"""
Called after we connected to zenhub.
"""
def callback(result):
self.log.info("Connected to the zenhub/%s instance", result)
def errback(result):
self.log.info(
"Unexpected error appeared while getting zenhub instance number %s",
result)
d = self.perspective.callRemote('getHubInstanceId')
d.addCallback(callback)
d.addErrback(errback)
return d
def gotPerspective(self, perspective):
"""
This gets called every time we reconnect.
@parameter perspective: Twisted perspective object
@type perspective: Twisted perspective object
"""
self.perspective = perspective
self.getZenhubInstanceId()
# Cancel the connection timeout timer as it's no longer needed.
if self._connectionTimeout:
try:
self._connectionTimeout.cancel()
except AlreadyCalled:
pass
self._connectionTimeout = None
d2 = self.getInitialServices()
if self.initialConnect:
self.log.debug('Chaining getInitialServices with d2')
self.initialConnect, d = None, self.initialConnect
d2.chainDeferred(d)
def connect(self):
pingInterval = self.options.zhPingInterval
factory = ReconnectingPBClientFactory(
connectTimeout=60,
pingPerspective=self.options.pingPerspective,
pingInterval=pingInterval,
pingtimeout=pingInterval * 5)
self.log.info("Connecting to %s:%d" %
(self.options.hubhost, self.options.hubport))
factory.connectTCP(self.options.hubhost, self.options.hubport)
username = self.options.hubusername
password = self.options.hubpassword
self.log.debug("Logging in as %s" % username)
c = credentials.UsernamePassword(username, password)
factory.gotPerspective = self.gotPerspective
factory.connecting = self.connecting
factory.setCredentials(c)
def timeout(d):
if not d.called:
self.connectTimeout()
self._connectionTimeout = reactor.callLater(self.options.hubtimeout,
timeout,
self.initialConnect)
return self.initialConnect
def connectTimeout(self):
self.log.error('Timeout connecting to zenhub: is it running?')
pass
def eventService(self):
return self.getServiceNow('EventService')
def getServiceNow(self, svcName):
if not svcName in self.services:
self.log.warning(
'No service named %r: ZenHub may be disconnected' % svcName)
return self.services.get(svcName, None) or FakeRemote()
def getService(self, serviceName, serviceListeningInterface=None):
"""
Attempt to get a service from zenhub. Returns a deferred.
When service is retrieved it is stashed in self.services with
serviceName as the key. When getService is called it will first
check self.services and if serviceName is already there it will return
the entry from self.services wrapped in a defer.succeed
"""
if serviceName in self.services:
return defer.succeed(self.services[serviceName])
def removeService(ignored):
self.log.debug('Removing service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
def callback(result, serviceName):
self.log.debug('Loaded service %s from zenhub' % serviceName)
self.services[serviceName] = result
result.notifyOnDisconnect(removeService)
return result
def errback(error, serviceName):
self.log.debug('errback after getting service %s' % serviceName)
self.log.error('Could not retrieve service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
return error
d = self.perspective.callRemote('getService', serviceName,
self.options.monitor,
serviceListeningInterface or self,
self.options.__dict__)
d.addCallback(callback, serviceName)
d.addErrback(errback, serviceName)
return d
def getInitialServices(self):
"""
After connecting to zenhub, gather our initial list of services.
"""
def errback(error):
if isinstance(error, Failure):
self.log.critical("Invalid monitor: %s" % self.options.monitor)
reactor.stop()
return defer.fail(
RemoteBadMonitor(
"Invalid monitor: %s" % self.options.monitor, ''))
return error
self.log.debug('Setting up initial services: %s' % \
', '.join(self.initialServices))
d = defer.DeferredList(
[self.getService(name) for name in self.initialServices],
fireOnOneErrback=True,
consumeErrors=True)
d.addErrback(errback)
return d
def connected(self):
pass
def _getThresholdNotifier(self):
if not self._threshold_notifier:
self._threshold_notifier = ThresholdNotifier(
self.sendEvent, self.getThresholds())
return self._threshold_notifier
def getThresholds(self):
if not self._thresholds:
self._thresholds = Thresholds()
return self._thresholds
def run(self):
def stopReporter():
if self._metrologyReporter:
return self._metrologyReporter.stop()
# Order of the shutdown triggers matter. Want to stop reporter first, calling self.metricWriter() below
# registers shutdown triggers for the actual metric http and redis publishers.
reactor.addSystemEventTrigger('before', 'shutdown', stopReporter)
threshold_notifier = self._getThresholdNotifier()
self.rrdStats.config(self.name, self.options.monitor,
self.metricWriter(), threshold_notifier,
self.derivativeTracker())
self.log.debug('Starting PBDaemon initialization')
d = self.connect()
def callback(result):
self.sendEvent(self.startEvent)
self.pushEventsLoop()
self.log.debug('Calling connected.')
self.connected()
return result
def startStatsLoop():
self.log.debug("Starting Statistic posting")
loop = task.LoopingCall(self.postStatistics)
loop.start(self.options.writeStatistics, now=False)
daemonTags = {
'zenoss_daemon': self.name,
'zenoss_monitor': self.options.monitor,
'internal': True
}
self._metrologyReporter = TwistedMetricReporter(
self.options.writeStatistics, self.metricWriter(), daemonTags)
self._metrologyReporter.start()
reactor.callWhenRunning(startStatsLoop)
d.addCallback(callback)
d.addErrback(twisted.python.log.err)
reactor.run()
if self._customexitcode:
sys.exit(self._customexitcode)
def setExitCode(self, exitcode):
self._customexitcode = exitcode
def stop(self, ignored=''):
if reactor.running:
try:
reactor.stop()
except ReactorNotRunning:
self.log.debug("Tried to stop reactor that was stopped")
else:
self.log.debug("stop() called when not running")
def _stopPbDaemon(self):
if self.stopped:
return
self.stopped = True
if 'EventService' in self.services:
# send stop event if we don't have an implied --cycle,
# or if --cycle has been specified
if not hasattr(self.options, 'cycle') or \
getattr(self.options, 'cycle', True):
self.sendEvent(self.stopEvent)
self.log.debug("Sent a 'stop' event")
if self._pushEventsDeferred:
self.log.debug("Currently sending events. Queueing next call")
d = self._pushEventsDeferred
# Schedule another call to flush any additional queued events
d.addBoth(lambda unused: self.pushEvents())
else:
d = self.pushEvents()
return d
self.log.debug("No event sent as no EventService available.")
def sendEvents(self, events):
map(self.sendEvent, events)
def sendEvent(self, event, **kw):
""" Add event to queue of events to be sent. If we have an event
service then process the queue.
"""
generatedEvent = self.generateEvent(event, **kw)
self.eventQueueManager.addEvent(generatedEvent)
self.counters['eventCount'] += 1
if self._eventHighWaterMark:
return self._eventHighWaterMark
elif self.eventQueueManager.event_queue_length >= self.options.maxqueuelen * self.options.queueHighWaterMark:
return self.pushEvents()
else:
return defer.succeed(None)
def generateEvent(self, event, **kw):
""" Add event to queue of events to be sent. If we have an event
service then process the queue.
"""
if not reactor.running:
return
eventCopy = {}
for k, v in chain(event.items(), kw.items()):
if isinstance(v, basestring):
#default max size is 512k
size = LIMITS.get(k, DEFAULT_LIMIT)
eventCopy[k] = v[0:size] if len(v) > size else v
else:
eventCopy[k] = v
eventCopy['agent'] = self.name
eventCopy['monitor'] = self.options.monitor
eventCopy['manager'] = self.fqdn
return eventCopy
@defer.inlineCallbacks
def pushEventsLoop(self):
"""Periodially, wake up and flush events to ZenHub.
"""
reactor.callLater(self.options.eventflushseconds, self.pushEventsLoop)
yield self.pushEvents()
# Record the number of events in the queue up to every 2 seconds.
now = time.time()
if self.rrdStats.name and now >= (self.lastStats + 2):
self.lastStats = now
self.rrdStats.gauge('eventQueueLength',
self.eventQueueManager.event_queue_length)
@defer.inlineCallbacks
def pushEvents(self):
"""Flush events to ZenHub.
"""
# are we already shutting down?
if not reactor.running:
self.log.debug("Skipping event sending - reactor not running.")
return
if self.eventQueueManager.event_queue_length >= self.options.maxqueuelen * self.options.queueHighWaterMark and not self._eventHighWaterMark:
self.log.debug(
"Queue length exceeded high water mark, %s ;creating high water mark deferred",
self.eventQueueManager.event_queue_length)
self._eventHighWaterMark = defer.Deferred()
# are still connected to ZenHub?
evtSvc = self.services.get('EventService', None)
if not evtSvc:
self.log.error("No event service: %r", evtSvc)
yield task.deferLater(reactor, 0, lambda: None)
if self._eventHighWaterMark:
d, self._eventHighWaterMark = self._eventHighWaterMark, None
#not connected, release throttle and let things queue
d.callback("No Event Service")
defer.returnValue(None)
if self._pushEventsDeferred:
self.log.debug("Skipping event sending - previous call active.")
defer.returnValue("Push Pending")
sent = 0
try:
#only set _pushEventsDeferred after we know we have an evtSvc/connectivity
self._pushEventsDeferred = defer.Deferred()
def repush(val):
if self.eventQueueManager.event_queue_length >= self.options.eventflushchunksize:
self.pushEvents()
return val
# conditionally push more events after this pushEvents call finishes
self._pushEventsDeferred.addCallback(repush)
discarded_events = self.eventQueueManager.discarded_events
if discarded_events:
self.log.error(
'Discarded oldest %d events because maxqueuelen was '
'exceeded: %d/%d', discarded_events,
discarded_events + self.options.maxqueuelen,
self.options.maxqueuelen)
self.counters['discardedEvents'] += discarded_events
self.eventQueueManager.discarded_events = 0
send_events_fn = partial(evtSvc.callRemote, 'sendEvents')
try:
sent = yield self.eventQueueManager.sendEvents(send_events_fn)
except ConnectionLost as ex:
self.log.error('Error sending event: %s', ex)
#let the reactor have time to clean up any connection errors and make callbacks
yield task.deferLater(reactor, 0, lambda: None)
except Exception as ex:
self.log.exception(ex)
#let the reactor have time to clean up any connection errors and make callbacks
yield task.deferLater(reactor, 0, lambda: None)
finally:
if self._pushEventsDeferred:
d, self._pushEventsDeferred = self._pushEventsDeferred, None
d.callback('sent %s' % sent)
if self._eventHighWaterMark and self.eventQueueManager.event_queue_length < self.options.maxqueuelen * self.options.queueHighWaterMark:
self.log.debug("Queue restored to below high water mark: %s",
self.eventQueueManager.event_queue_length)
d, self._eventHighWaterMark = self._eventHighWaterMark, None
d.callback("Queue length below high water mark")
def heartbeat(self):
"""if cycling, send a heartbeat, else, shutdown"""
if not self.options.cycle:
self.stop()
return
heartbeatEvent = self.generateEvent(self.heartbeatEvent,
timeout=self.heartbeatTimeout)
self.eventQueueManager.addHeartbeatEvent(heartbeatEvent)
# heartbeat is normally 3x cycle time
self.niceDoggie(self.heartbeatTimeout / 3)
def postStatisticsImpl(self):
pass
def postStatistics(self):
# save daemon counter stats
for name, value in self.counters.items():
self.log.info("Counter %s, value %d", name, value)
self.rrdStats.counter(name, value)
# persist counters values
self.postStatisticsImpl()
def _pickleName(self):
instance_id = os.environ.get('CONTROLPLANE_INSTANCE_ID')
return 'var/%s_%s_counters.pickle' % (self.name, instance_id)
def remote_getName(self):
return self.name
def remote_shutdown(self, unused):
self.stop()
self.sigTerm()
def remote_setPropertyItems(self, items):
pass
@translateError
def remote_updateThresholdClasses(self, classes):
from Products.ZenUtils.Utils import importClass
self.log.debug("Loading classes %s", classes)
for c in classes:
try:
importClass(c)
except ImportError:
self.log.error("Unable to import class %s", c)
def _checkZenHub(self):
"""
Check status of ZenHub (using ping method of service).
@return: if ping occurs, return deferred with result of ping attempt.
"""
self.log.debug('_checkZenHub: entry')
def callback(result):
self.log.debug('ZenHub health check: Got result %s' % result)
if result == 'pong':
self.log.debug(
'ZenHub health check: Success - received pong from ZenHub ping service.'
)
self._signalZenHubAnswering(True)
else:
self.log.error(
'ZenHub health check did not respond as expected.')
self._signalZenHubAnswering(False)
def errback(error):
self.log.error('Error pinging ZenHub: %s (%s).' %
(error, getattr(error, 'message', '')))
self._signalZenHubAnswering(False)
try:
if self.perspective:
self.log.debug(
'ZenHub health check: perspective found. attempting remote ping call.'
)
d = self.perspective.callRemote('ping')
d.addCallback(callback)
d.addErrback(errback)
return d
else:
self.log.debug('ZenHub health check: ZenHub may be down.')
self._signalZenHubAnswering(False)
except pb.DeadReferenceError:
self.log.warning(
"ZenHub health check: DeadReferenceError - lost connection to ZenHub."
)
self._signalZenHubAnswering(False)
except Exception as e:
self.log.error('ZenHub health check: caught %s exception: %s' %
(e.__class__, e.message))
self._signalZenHubAnswering(False)
def _signalZenHubAnswering(self, answering):
"""
Write or remove file that the ZenHub_answering health check uses to report status.
@param answering: true if ZenHub is answering, False, otherwise.
"""
self.log.debug('_signalZenHubAnswering(%s)' % answering)
filename = 'zenhub_connected'
signalFilePath = zenPath('var', filename)
if answering:
self.log.debug('writing file at %s' % signalFilePath)
atomicWrite(signalFilePath, '')
else:
try:
self.log.debug('removing file at %s' % signalFilePath)
os.remove(signalFilePath)
except Exception as e:
self.log.debug('ignoring %s exception (%s) removing file %s' %
(e.__class__, e.message, signalFilePath))
def buildOptions(self):
self.parser.add_option('--hubhost',
dest='hubhost',
default=DEFAULT_HUB_HOST,
help='Host of zenhub daemon.'
' Default is %s.' % DEFAULT_HUB_HOST)
self.parser.add_option('--hubport',
dest='hubport',
type='int',
default=DEFAULT_HUB_PORT,
help='Port zenhub listens on.'
'Default is %s.' % DEFAULT_HUB_PORT)
self.parser.add_option('--hubusername',
dest='hubusername',
default=DEFAULT_HUB_USERNAME,
help='Username for zenhub login.'
' Default is %s.' % DEFAULT_HUB_USERNAME)
self.parser.add_option('--hubpassword',
dest='hubpassword',
default=DEFAULT_HUB_PASSWORD,
help='Password for zenhub login.'
' Default is %s.' % DEFAULT_HUB_PASSWORD)
self.parser.add_option('--monitor',
dest='monitor',
default=DEFAULT_HUB_MONITOR,
help='Name of monitor instance to use for'
' configuration. Default is %s.' %
DEFAULT_HUB_MONITOR)
self.parser.add_option('--initialHubTimeout',
dest='hubtimeout',
type='int',
default=30,
help='Initial time to wait for a ZenHub '
'connection')
self.parser.add_option('--allowduplicateclears',
dest='allowduplicateclears',
default=False,
action='store_true',
help='Send clear events even when the most '
'recent event was also a clear event.')
self.parser.add_option(
'--duplicateclearinterval',
dest='duplicateclearinterval',
default=0,
type='int',
help=('Send a clear event every [DUPLICATECLEARINTEVAL] '
'events.'))
self.parser.add_option('--eventflushseconds',
dest='eventflushseconds',
default=5.,
type='float',
help='Seconds between attempts to flush '
'events to ZenHub.')
self.parser.add_option('--eventflushchunksize',
dest='eventflushchunksize',
default=50,
type='int',
help='Number of events to send to ZenHub'
'at one time')
self.parser.add_option('--maxqueuelen',
dest='maxqueuelen',
default=5000,
type='int',
help='Maximum number of events to queue')
self.parser.add_option(
'--queuehighwatermark',
dest='queueHighWaterMark',
default=0.75,
type='float',
help=
'The size, in percent, of the event queue when event pushback starts'
)
self.parser.add_option('--zenhubpinginterval',
dest='zhPingInterval',
default=120,
type='int',
help='How often to ping zenhub')
self.parser.add_option('--disable-event-deduplication',
dest='deduplicate_events',
default=True,
action='store_false',
help='Disable event de-duplication')
self.parser.add_option(
'--redis-url',
dest='redisUrl',
type='string',
default='redis://localhost:{default}/0'.format(
default=publisher.defaultRedisPort),
help=
'redis connection string: redis://[hostname]:[port]/[db], default: %default'
)
self.parser.add_option(
'--metricBufferSize',
dest='metricBufferSize',
type='int',
default=publisher.defaultMetricBufferSize,
help='Number of metrics to buffer if redis goes down')
self.parser.add_option(
'--metricsChannel',
dest='metricsChannel',
type='string',
default=publisher.defaultMetricsChannel,
help='redis channel to which metrics are published')
self.parser.add_option('--maxOutstandingMetrics',
dest='maxOutstandingMetrics',
type='int',
default=publisher.defaultMaxOutstandingMetrics,
help='Max Number of metrics to allow in redis')
self.parser.add_option('--disable-ping-perspective',
dest='pingPerspective',
help="Enable or disable ping perspective",
default=True,
action='store_false')
self.parser.add_option(
'--writeStatistics',
dest='writeStatistics',
type='int',
default=30,
help='How often to write internal statistics value in seconds')
ZenDaemon.buildOptions(self)
class ZenModeler(PBDaemon):
"""
Daemon class to attach to zenhub and pass along
device configuration information.
"""
name = 'zenmodeler'
initialServices = PBDaemon.initialServices + ['ModelerService']
generateEvents = True
configCycleInterval = 360
classCollectorPlugins = ()
def __init__(self, single=False ):
"""
Initalizer
@param single: collect from a single device?
@type single: boolean
"""
PBDaemon.__init__(self)
# FIXME: cleanup --force option #2660
self.options.force = True
self.start = None
self.startat = None
self.rrdStats = DaemonStats()
self.single = single
if self.options.device:
self.single = True
self.modelerCycleInterval = self.options.cycletime
self.collage = float( self.options.collage ) / 1440.0
self.pendingNewClients = False
self.clients = []
self.finished = []
self.devicegen = None
self.counters = collections.Counter()
# Make sendEvent() available to plugins
zope.component.provideUtility(self, IEventService)
# Delay start for between 10 and 60 seconds when run as a daemon.
self.started = False
self.startDelay = 0
self.immediate = 1
if self.options.daemon:
if self.options.now:
self.log.debug("Run as a daemon, starting immediately.")
else:
# self.startDelay = randint(10, 60) * 60
self.startDelay = randint(10, 60) * 1
self.immediate = 0
self.log.info("Run as a daemon, waiting %s seconds to start." %
self.startDelay)
else:
self.log.debug("Run in foreground, starting immediately.")
# load performance counters
self.loadCounters()
def reportError(self, error):
"""
Log errors that have occurred
@param error: error message
@type error: string
"""
self.log.error("Error occured: %s", error)
def connected(self):
"""
Called after connected to the zenhub service
"""
d = self.configure()
d.addCallback(self.heartbeat)
d.addErrback(self.reportError)
def configure(self):
"""
Get our configuration from zenhub
"""
# add in the code to fetch cycle time, etc.
def inner(driver):
"""
Generator function to gather our configuration
@param driver: driver object
@type driver: driver object
"""
self.log.debug('fetching monitor properties')
yield self.config().callRemote('propertyItems')
items = dict(driver.next())
# If the cycletime option is not specified or zero, then use the
# modelerCycleInterval value in the database.
if not self.options.cycletime:
self.modelerCycleInterval = items.get('modelerCycleInterval',
_DEFAULT_CYCLE_INTERVAL)
self.configCycleInterval = items.get('configCycleInterval',
self.configCycleInterval)
reactor.callLater(self.configCycleInterval * 60, self.configure)
self.log.debug("Getting threshold classes...")
yield self.config().callRemote('getThresholdClasses')
self.remote_updateThresholdClasses(driver.next())
self.log.debug("Getting collector thresholds...")
yield self.config().callRemote('getCollectorThresholds')
thresholds = driver.next()
threshold_notifier = ThresholdNotifier(self.sendEvent, thresholds)
self.rrdStats.config(self.name,
self.options.monitor,
self.metricWriter(),
threshold_notifier,
self.derivativeTracker())
self.log.debug("Getting collector plugins for each DeviceClass")
yield self.config().callRemote('getClassCollectorPlugins')
self.classCollectorPlugins = driver.next()
return drive(inner)
def config(self):
"""
Get the ModelerService
"""
return self.services.get('ModelerService', FakeRemote())
def selectPlugins(self, device, transport):
"""
Build a list of active plugins for a device, based on:
* the --collect command-line option which is a regex
* the --ignore command-line option which is a regex
* transport which is a string describing the type of plugin
@param device: device to collect against
@type device: string
@param transport: python, ssh, snmp, telnet, cmd
@type transport: string
@return: results of the plugin
@type: string
@todo: determine if an event for the collector AND the device should be sent
"""
plugins = []
valid_loaders = []
for loader in device.plugins:
try:
plugin= loader.create()
self.log.debug( "Loaded plugin %s" % plugin.name() )
plugins.append( plugin )
valid_loaders.append( loader )
except (SystemExit, KeyboardInterrupt), ex:
self.log.info( "Interrupted by external signal (%s)" % str(ex) )
raise
except Plugins.PluginImportError, import_error:
import socket
component, _ = os.path.splitext( os.path.basename( sys.argv[0] ) )
collector_host= socket.gethostname()
# NB: an import errror affects all devices,
# so report the issue against the collector
# TOOD: determine if an event for the collector AND the device should be sent
evt= { "eventClass":"/Status/Update", "component":component,
"agent":collector_host, "device":collector_host,
"severity":Error }
info= "Problem loading plugin %s" % import_error.plugin
self.log.error( info )
evt[ 'summary' ]= info
info= import_error.traceback
self.log.error( info )
evt[ 'message' ]= info
info= ("Due to import errors, removing the %s plugin"
" from this collection cycle.") % import_error.plugin
self.log.error( info )
evt[ 'message' ] += "%s\n" % info
self.sendEvent( evt )
class PBDaemon(ZenDaemon, pb.Referenceable):
name = 'pbdaemon'
initialServices = ['EventService']
heartbeatEvent = {'eventClass': Heartbeat}
heartbeatTimeout = 60 * 3
_customexitcode = 0
_pushEventsDeferred = None
def __init__(self, noopts=0, keeproot=False, name=None):
# if we were provided our collector name via the constructor instead of
# via code, be sure to store it correctly.
if name is not None:
self.name = name
self.mname = name
try:
ZenDaemon.__init__(self, noopts, keeproot)
except IOError:
import traceback
self.log.critical(traceback.format_exc(0))
sys.exit(1)
self.rrdStats = DaemonStats()
self.lastStats = 0
self.perspective = None
self.services = {}
self.eventQueueManager = EventQueueManager(self.options, self.log)
self.startEvent = startEvent.copy()
self.stopEvent = stopEvent.copy()
details = dict(component=self.name, device=self.options.monitor)
for evt in self.startEvent, self.stopEvent, self.heartbeatEvent:
evt.update(details)
self.initialConnect = defer.Deferred()
self.stopped = False
self.counters = collections.Counter()
self.loadCounters()
self._pingedZenhub = None
self._connectionTimeout = None
# Add a shutdown trigger to send a stop event and flush the event queue
reactor.addSystemEventTrigger('before', 'shutdown', self._stopPbDaemon)
def connecting(self):
"""
Called when about to connect to zenhub
"""
self.log.info("Attempting to connect to zenhub")
def gotPerspective(self, perspective):
"""
This gets called every time we reconnect.
@parameter perspective: Twisted perspective object
@type perspective: Twisted perspective object
"""
self.log.info("Connected to ZenHub")
self.perspective = perspective
# Cancel the connection timeout timer as it's no longer needed.
if self._connectionTimeout:
try:
self._connectionTimeout.cancel()
except AlreadyCalled:
pass
self._connectionTimeout = None
d2 = self.getInitialServices()
if self.initialConnect:
self.log.debug('Chaining getInitialServices with d2')
self.initialConnect, d = None, self.initialConnect
d2.chainDeferred(d)
def connect(self):
factory = ReconnectingPBClientFactory(connectTimeout=60)
self.log.info("Connecting to %s:%d" %
(self.options.hubhost, self.options.hubport))
factory.connectTCP(self.options.hubhost, self.options.hubport)
username = self.options.hubusername
password = self.options.hubpassword
self.log.debug("Logging in as %s" % username)
c = credentials.UsernamePassword(username, password)
factory.gotPerspective = self.gotPerspective
factory.connecting = self.connecting
factory.startLogin(c)
def timeout(d):
if not d.called:
self.connectTimeout()
self._connectionTimeout = reactor.callLater(self.options.hubtimeout,
timeout,
self.initialConnect)
return self.initialConnect
def connectTimeout(self):
self.log.error('Timeout connecting to zenhub: is it running?')
pass
def eventService(self):
return self.getServiceNow('EventService')
def getServiceNow(self, svcName):
if not svcName in self.services:
self.log.warning(
'No service named %r: ZenHub may be disconnected' % svcName)
return self.services.get(svcName, None) or FakeRemote()
def getService(self, serviceName, serviceListeningInterface=None):
"""
Attempt to get a service from zenhub. Returns a deferred.
When service is retrieved it is stashed in self.services with
serviceName as the key. When getService is called it will first
check self.services and if serviceName is already there it will return
the entry from self.services wrapped in a defer.succeed
"""
if serviceName in self.services:
return defer.succeed(self.services[serviceName])
def removeService(ignored):
self.log.debug('Removing service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
def callback(result, serviceName):
self.log.debug('Loaded service %s from zenhub' % serviceName)
self.services[serviceName] = result
result.notifyOnDisconnect(removeService)
return result
def errback(error, serviceName):
self.log.debug('errback after getting service %s' % serviceName)
self.log.error('Could not retrieve service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
return error
d = self.perspective.callRemote('getService', serviceName,
self.options.monitor,
serviceListeningInterface or self)
d.addCallback(callback, serviceName)
d.addErrback(errback, serviceName)
return d
def getInitialServices(self):
"""
After connecting to zenhub, gather our initial list of services.
"""
def errback(error):
if isinstance(error, Failure):
self.log.critical("Invalid monitor: %s" % self.options.monitor)
reactor.stop()
return defer.fail(
RemoteBadMonitor(
"Invalid monitor: %s" % self.options.monitor, ''))
return error
self.log.debug('Setting up initial services: %s' % \
', '.join(self.initialServices))
d = defer.DeferredList(
[self.getService(name) for name in self.initialServices],
fireOnOneErrback=True,
consumeErrors=True)
d.addErrback(errback)
return d
def connected(self):
pass
def run(self):
self.rrdStats.config(self.options.monitor, self.name, [])
self.log.debug('Starting PBDaemon initialization')
d = self.connect()
def callback(result):
self.sendEvent(self.startEvent)
self.pushEventsLoop()
self.log.debug('Calling connected.')
self.connected()
return result
d.addCallback(callback)
d.addErrback(twisted.python.log.err)
reactor.run()
if self._customexitcode:
sys.exit(self._customexitcode)
def setExitCode(self, exitcode):
self._customexitcode = exitcode
def stop(self, ignored=''):
if reactor.running:
try:
reactor.stop()
except ReactorNotRunning:
self.log.debug("Tried to stop reactor that was stopped")
else:
self.log.debug("stop() called when not running")
def _stopPbDaemon(self):
if self.stopped:
return
self.stopped = True
if 'EventService' in self.services:
# send stop event if we don't have an implied --cycle,
# or if --cycle has been specified
if not hasattr(self.options, 'cycle') or\
getattr(self.options, 'cycle', True):
self.sendEvent(self.stopEvent)
self.log.debug("Sent a 'stop' event")
if self._pushEventsDeferred:
self.log.debug("Currently sending events. Queueing next call")
d = self._pushEventsDeferred
# Schedule another call to flush any additional queued events
d.addBoth(lambda unused: self.pushEvents())
else:
d = self.pushEvents()
d.addBoth(lambda unused: self.saveCounters())
return d
self.log.debug("No event sent as no EventService available.")
self.saveCounters()
def sendEvents(self, events):
map(self.sendEvent, events)
def sendEvent(self, event, **kw):
''' Add event to queue of events to be sent. If we have an event
service then process the queue.
'''
generatedEvent = self.generateEvent(event, **kw)
self.eventQueueManager.addEvent(generatedEvent)
self.counters['eventCount'] += 1
def generateEvent(self, event, **kw):
''' Add event to queue of events to be sent. If we have an event
service then process the queue.
'''
if not reactor.running: return
event = event.copy()
event['agent'] = self.name
event['monitor'] = self.options.monitor
event['manager'] = self.fqdn
event.update(kw)
return event
@defer.inlineCallbacks
def pushEventsLoop(self):
"""Periodially, wake up and flush events to ZenHub.
"""
reactor.callLater(self.options.eventflushseconds, self.pushEventsLoop)
yield self.pushEvents()
# Record the number of events in the queue every 5 minutes.
now = time.time()
if self.rrdStats.name and now >= (self.lastStats + 300):
self.lastStats = now
events = self.rrdStats.gauge(
'eventQueueLength', 300,
self.eventQueueManager.event_queue_length)
for event in events:
self.eventQueueManager.addPerformanceEvent(event)
@defer.inlineCallbacks
def pushEvents(self):
"""Flush events to ZenHub.
"""
# are we already shutting down?
if not reactor.running:
self.log.debug("Skipping event sending - reactor not running.")
return
if self._pushEventsDeferred:
self.log.debug("Skipping event sending - previous call active.")
return
try:
self._pushEventsDeferred = defer.Deferred()
# are still connected to ZenHub?
evtSvc = self.services.get('EventService', None)
if not evtSvc:
self.log.error("No event service: %r", evtSvc)
return
discarded_events = self.eventQueueManager.discarded_events
if discarded_events:
self.log.error(
'Discarded oldest %d events because maxqueuelen was '
'exceeded: %d/%d', discarded_events,
discarded_events + self.options.maxqueuelen,
self.options.maxqueuelen)
self.counters['discardedEvents'] += discarded_events
self.eventQueueManager.discarded_events = 0
send_events_fn = partial(evtSvc.callRemote, 'sendEvents')
try:
yield self.eventQueueManager.sendEvents(send_events_fn)
except ConnectionLost as ex:
self.log.error('Error sending event: %s', ex)
except ConnectionDone:
pass
except Exception as ex:
self.log.exception(ex)
finally:
d, self._pushEventsDeferred = self._pushEventsDeferred, None
d.callback('sent')
def heartbeat(self):
'if cycling, send a heartbeat, else, shutdown'
if not self.options.cycle:
self.stop()
return
heartbeatEvent = self.generateEvent(self.heartbeatEvent,
timeout=self.heartbeatTimeout)
self.eventQueueManager.addHeartbeatEvent(heartbeatEvent)
# heartbeat is normally 3x cycle time
self.niceDoggie(self.heartbeatTimeout / 3)
events = []
# save daemon counter stats
for name, value in self.counters.items():
self.log.info("Counter %s, value %d", name, value)
events += self.rrdStats.counter(name, 300, value)
self.sendEvents(events)
# persist counters values
self.saveCounters()
def saveCounters(self):
atomicWrite(
zenPath('var/%s_counters.pickle' % self.name),
pickle.dumps(self.counters),
raiseException=False,
)
def loadCounters(self):
try:
self.counters = pickle.load(
open(zenPath('var/%s_counters.pickle' % self.name)))
except Exception:
pass
def remote_getName(self):
return self.name
def remote_shutdown(self, unused):
self.stop()
self.sigTerm()
def remote_setPropertyItems(self, items):
pass
@translateError
def remote_updateThresholdClasses(self, classes):
from Products.ZenUtils.Utils import importClass
self.log.debug("Loading classes %s", classes)
for c in classes:
try:
importClass(c)
except ImportError:
self.log.error("Unable to import class %s", c)
def buildOptions(self):
self.parser.add_option('--hubhost',
dest='hubhost',
default=DEFAULT_HUB_HOST,
help='Host of zenhub daemon.'
' Default is %s.' % DEFAULT_HUB_HOST)
self.parser.add_option('--hubport',
dest='hubport',
type='int',
default=DEFAULT_HUB_PORT,
help='Port zenhub listens on.'
'Default is %s.' % DEFAULT_HUB_PORT)
self.parser.add_option('--hubusername',
dest='hubusername',
default=DEFAULT_HUB_USERNAME,
help='Username for zenhub login.'
' Default is %s.' % DEFAULT_HUB_USERNAME)
self.parser.add_option('--hubpassword',
dest='hubpassword',
default=DEFAULT_HUB_PASSWORD,
help='Password for zenhub login.'
' Default is %s.' % DEFAULT_HUB_PASSWORD)
self.parser.add_option('--monitor',
dest='monitor',
default=DEFAULT_HUB_MONITOR,
help='Name of monitor instance to use for'
' configuration. Default is %s.' %
DEFAULT_HUB_MONITOR)
self.parser.add_option('--initialHubTimeout',
dest='hubtimeout',
type='int',
default=30,
help='Initial time to wait for a ZenHub '
'connection')
self.parser.add_option('--allowduplicateclears',
dest='allowduplicateclears',
default=False,
action='store_true',
help='Send clear events even when the most '
'recent event was also a clear event.')
self.parser.add_option(
'--duplicateclearinterval',
dest='duplicateclearinterval',
default=0,
type='int',
help=('Send a clear event every [DUPLICATECLEARINTEVAL] '
'events.'))
self.parser.add_option('--eventflushseconds',
dest='eventflushseconds',
default=5.,
type='float',
help='Seconds between attempts to flush '
'events to ZenHub.')
self.parser.add_option('--eventflushchunksize',
dest='eventflushchunksize',
default=50,
type='int',
help='Number of events to send to ZenHub'
'at one time')
self.parser.add_option('--maxqueuelen',
dest='maxqueuelen',
default=5000,
type='int',
help='Maximum number of events to queue')
self.parser.add_option('--zenhubpinginterval',
dest='zhPingInterval',
default=30,
type='int',
help='How often to ping zenhub')
self.parser.add_option('--disable-event-deduplication',
dest='deduplicate_events',
default=True,
action='store_false',
help='Disable event de-duplication')
ZenDaemon.buildOptions(self)
class EventServer(PBDaemon):
'Base class for a daemon whose primary job is to post events'
name = 'EventServer'
def __init__(self):
PBDaemon.__init__(self, keeproot=True)
self.stats = Stats()
self.rrdStats = DaemonStats()
def connected(self):
self.sendEvent(
dict(device=self.options.monitor,
eventClass=App_Start,
summary="%s started" % self.name,
severity=0,
component=self.name))
self.log.info("started")
self.configure()
def model(self):
return self.services.get('EventService', FakeRemote())
def configure(self):
def inner(driver):
self.log.info("fetching default RRDCreateCommand")
yield self.model().callRemote('getDefaultRRDCreateCommand')
createCommand = driver.next()
self.log.info("getting threshold classes")
yield self.model().callRemote('getThresholdClasses')
self.remote_updateThresholdClasses(driver.next())
self.log.info("getting collector thresholds")
yield self.model().callRemote('getCollectorThresholds')
self.rrdStats.config(self.options.monitor, self.name,
driver.next(), createCommand)
self.heartbeat()
self.reportCycle()
d = drive(inner)
def error(result):
self.log.error("Unexpected error in configure: %s" % result)
d.addErrback(error)
return d
def sendEvent(self, event, **kw):
# FIXME: get real event processing stats
if 'firstTime' in event:
self.stats.add(min(time.time() - event['firstTime'], 0))
PBDaemon.sendEvent(self, event, **kw)
def useUdpFileDescriptor(self, fd):
from twisted.internet import udp
s = socket.fromfd(fd, socket.AF_INET, socket.SOCK_DGRAM)
import os
os.close(fd)
port = s.getsockname()[1]
transport = udp.Port(port, self)
s.setblocking(0)
transport.socket = s
transport.fileno = s.fileno
transport.connected = 1
transport._realPortNumber = port
self.transport = transport
# hack around startListening not being called
self.numPorts = 1
transport.startReading()
def useTcpFileDescriptor(self, fd, factory):
import os, socket
for i in range(19800, 19999):
try:
p = reactor.listenTCP(i, factory)
os.dup2(fd, p.socket.fileno())
p.socket.listen(p.backlog)
p.socket.setblocking(False)
p.socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
os.close(fd)
return p
except socket.error:
pass
raise socket.error("Unable to find an open socket to listen on")
def reportCycle(self):
if self.options.statcycle:
self.report()
reactor.callLater(self.options.statcycle, self.reportCycle)
def heartbeat(self):
"""Since we don't do anything on a regular basis, just
push heartbeats regularly"""
seconds = self.heartbeatTimeout / 3
reactor.callLater(self.heartbeatTimeout / 3, self.heartbeat)
PBDaemon.heartbeat(self)
totalTime, totalEvents, maxTime = self.stats.report()
for ev in (self.rrdStats.counter('events', seconds, totalEvents) +
self.rrdStats.counter('totalTime', seconds,
int(totalTime * 1000))):
self.sendEvent(ev)
def report(self):
'report some simple diagnostics at shutdown'
totalTime, totalEvents, maxTime = self.stats.report()
self.log.info("%d events processed in %.2f seconds", totalEvents,
totalTime)
if totalEvents > 0:
self.log.info("%.5f average seconds per event",
(totalTime / totalEvents))
self.log.info("Maximum processing time for one event was %.5f",
maxTime)
def buildOptions(self):
PBDaemon.buildOptions(self)
self.parser.add_option(
'--statcycle',
dest='statcycle',
type='int',
help='Number of seconds between the writing of statistics',
default=0)
class ZenModeler(PBDaemon):
"""
Daemon class to attach to zenhub and pass along
device configuration information.
"""
name = 'zenmodeler'
initialServices = PBDaemon.initialServices + ['ModelerService']
generateEvents = True
configCycleInterval = 360
classCollectorPlugins = ()
def __init__(self, single=False):
"""
Initalizer
@param single: collect from a single device?
@type single: boolean
"""
PBDaemon.__init__(self)
# FIXME: cleanup --force option #2660
self.options.force = True
self.start = None
self.startat = None
self.rrdStats = DaemonStats()
self.single = single
if self.options.device:
self.single = True
self.modelerCycleInterval = self.options.cycletime
# get the minutes and convert to fraction of a day
self.collage = float(self.options.collage) / 1440.0
self.pendingNewClients = False
self.clients = []
self.finished = []
self.devicegen = None
self.counters = collections.Counter()
self.configFilter = None
self.configLoaded = False
# Make sendEvent() available to plugins
zope.component.provideUtility(self, IEventService)
# Delay start for between 10 and 60 seconds when run as a daemon.
self.started = False
self.startDelay = 0
self.immediate = 1
if self.options.daemon or self.options.cycle:
if self.options.now:
self.log.debug('option "now" specified, starting immediately.')
else:
# self.startDelay = randint(10, 60) * 60
self.startDelay = randint(10, 60) * 1
self.immediate = 0
self.log.info(
'option "now" not specified, waiting %s seconds to start.'
% self.startDelay)
else:
self.log.debug("Run in foreground, starting immediately.")
# ZEN-26637
self.collectorLoopIteration = 0
self.mainLoopGotDeviceList = False
self.isMainScheduled = False
self._modeledDevicesMetric = Metrology.meter(
"zenmodeler.modeledDevices")
self._failuresMetric = Metrology.counter("zenmodeler.failures")
def reportError(self, error):
"""
Log errors that have occurred
@param error: error message
@type error: string
"""
self.log.error("Error occured: %s", error)
def connected(self):
"""
Called after connected to the zenhub service
"""
reactor.callLater(_CONFIG_PULLING_TIMEOUT, self._checkConfigLoad)
d = self.configure()
d.addCallback(self.heartbeat)
d.addErrback(self.reportError)
def _checkConfigLoad(self):
"""
Looping call to check whether zenmodeler got configuration
from ZenHub.
"""
if not self.configLoaded:
self.log.info("Modeling has not started pending configuration "
"pull from ZenHub. Is ZenHub overloaded?")
reactor.callLater(_CONFIG_PULLING_TIMEOUT, self._checkConfigLoad)
def configure(self):
"""
Get our configuration from zenhub
"""
# add in the code to fetch cycle time, etc.
self.log.info("Getting configuration from ZenHub...")
def inner(driver):
"""
Generator function to gather our configuration
@param driver: driver object
@type driver: driver object
"""
self.log.debug('fetching monitor properties')
yield self.config().callRemote('propertyItems')
items = dict(driver.next())
# If the cycletime option is not specified or zero, then use the
# modelerCycleInterval value in the database.
if not self.options.cycletime:
self.modelerCycleInterval = items.get('modelerCycleInterval',
_DEFAULT_CYCLE_INTERVAL)
self.configCycleInterval = items.get('configCycleInterval',
self.configCycleInterval)
reactor.callLater(self.configCycleInterval * 60, self.configure)
self.log.debug("Getting threshold classes...")
yield self.config().callRemote('getThresholdClasses')
self.remote_updateThresholdClasses(driver.next())
self.log.debug("Getting collector thresholds...")
yield self.config().callRemote('getCollectorThresholds')
thresholds = driver.next()
threshold_notifier = ThresholdNotifier(self.sendEvent, thresholds)
self.rrdStats.config(self.name, self.options.monitor,
self.metricWriter(), threshold_notifier,
self.derivativeTracker())
self.log.debug("Getting collector plugins for each DeviceClass")
yield self.config().callRemote('getClassCollectorPlugins')
self.classCollectorPlugins = driver.next()
self.configLoaded = True
return drive(inner)
def config(self):
"""
Get the ModelerService
"""
return self.services.get('ModelerService', FakeRemote())
def selectPlugins(self, device, transport):
"""
Build a list of active plugins for a device, based on:
* the --collect command-line option which is a regex
* the --ignore command-line option which is a regex
* transport which is a string describing the type of plugin
@param device: device to collect against
@type device: string
@param transport: python, ssh, snmp, telnet, cmd
@type transport: string
@return: results of the plugin
@type: string
@todo: determine if an event for the collector AND the device should be sent
"""
plugins = []
valid_loaders = []
for loader in device.plugins:
try:
plugin = loader.create()
self.log.debug("Loaded plugin %s" % plugin.name())
plugins.append(plugin)
valid_loaders.append(loader)
except Plugins.PluginImportError as import_error:
import socket
component, _ = os.path.splitext(os.path.basename(sys.argv[0]))
collector_host = socket.gethostname()
# NB: an import errror affects all devices,
# so report the issue against the collector
# TOOD: determine if an event for the collector AND the device should be sent
evt = {
"eventClass": "/Status/Update",
"component": component,
"agent": collector_host,
"device": collector_host,
"severity": Error
}
info = "Problem loading plugin %s" % import_error.plugin
self.log.error(info)
evt['summary'] = info
info = import_error.traceback
self.log.error(info)
evt['message'] = info
info = ("Due to import errors, removing the %s plugin"
" from this collection cycle.") % import_error.plugin
self.log.error(info)
evt['message'] += "%s\n" % info
self.sendEvent(evt)
# Make sure that we don't generate messages for bad loaders again
# NB: doesn't update the device's zProperties
if len(device.plugins) != len(valid_loaders):
device.plugins = valid_loaders
# Create functions to search for what plugins we will and
# won't supply to the device
collectTest = lambda x: False
ignoreTest = lambda x: False
if self.options.collectPlugins:
collectTest = re.compile(self.options.collectPlugins).search
elif self.options.ignorePlugins:
ignoreTest = re.compile(self.options.ignorePlugins).search
result = []
for plugin in plugins:
if plugin.transport != transport:
continue
name = plugin.name()
if ignoreTest(name):
self.log.debug("Ignoring %s on %s because of --ignore flag",
name, device.id)
elif collectTest(name):
self.log.debug("Using %s on %s because of --collect flag",
name, device.id)
result.append(plugin)
elif not self.options.collectPlugins:
self.log.debug("Using %s on %s", name, device.id)
result.append(plugin)
return result
def collectDevice(self, device):
"""
Collect data from a single device.
@param device: device to collect against
@type device: string
"""
clientTimeout = getattr(device, 'zCollectorClientTimeout', 180)
ip = device.manageIp
timeout = clientTimeout + time.time()
if USE_WMI:
self.wmiCollect(device, ip, timeout)
else:
self.log.info(
"skipping WMI-based collection, PySamba zenpack not installed")
self.log.info("Collect on device {} for collector loop #{:03d}".format(
device.id, self.collectorLoopIteration))
self.pythonCollect(device, ip, timeout)
self.cmdCollect(device, ip, timeout)
self.snmpCollect(device, ip, timeout)
self.portscanCollect(device, ip, timeout)
def wmiCollect(self, device, ip, timeout):
"""
Start the Windows Management Instrumentation (WMI) collector
@param device: device to collect against
@type device: string
@param ip: IP address of device to collect against
@type ip: string
@param timeout: timeout before failing the connection
@type timeout: integer
"""
if self.options.nowmi:
return
client = None
try:
plugins = self.selectPlugins(device, 'wmi')
if not plugins:
self.log.info("No WMI plugins found for %s" % device.id)
return
if self.checkCollection(device):
self.log.info('WMI collector method for device %s' % device.id)
self.log.info("plugins: %s",
", ".join(map(lambda p: p.name(), plugins)))
client = WMIClient(device, self, plugins)
if not client or not plugins:
self.log.warn("WMI collector creation failed")
return
except Exception:
self.log.exception("Error opening WMI collector")
self.addClient(client, timeout, 'WMI', device.id)
def pythonCollect(self, device, ip, timeout):
"""
Start local Python collection client.
@param device: device to collect against
@type device: string
@param ip: IP address of device to collect against
@type ip: string
@param timeout: timeout before failing the connection
@type timeout: integer
"""
client = None
try:
plugins = self.selectPlugins(device, "python")
if not plugins:
self.log.info("No Python plugins found for %s" % device.id)
return
if self.checkCollection(device):
self.log.info('Python collection device %s' % device.id)
self.log.info("plugins: %s",
", ".join(map(lambda p: p.name(), plugins)))
client = PythonClient(device, self, plugins)
if not client or not plugins:
self.log.warn("Python client creation failed")
return
except Exception:
self.log.exception("Error opening pythonclient")
self.addClient(client, timeout, 'python', device.id)
def cmdCollect(self, device, ip, timeout):
"""
Start shell command collection client.
@param device: device to collect against
@type device: string
@param ip: IP address of device to collect against
@type ip: string
@param timeout: timeout before failing the connection
@type timeout: integer
"""
client = None
clientType = 'snmp' # default to SNMP if we can't figure out a protocol
hostname = device.id
try:
plugins = self.selectPlugins(device, "command")
if not plugins:
self.log.info("No command plugins found for %s" % hostname)
return
protocol = getattr(device, 'zCommandProtocol', defaultProtocol)
commandPort = getattr(device, 'zCommandPort', defaultPort)
# don't even create a client if we shouldn't collect/model yet
if not self.checkCollection(device):
return
if protocol == "ssh":
client = SshClient(hostname,
ip,
commandPort,
options=self.options,
plugins=plugins,
device=device,
datacollector=self,
isLoseConnection=True)
clientType = 'ssh'
self.log.info('Using SSH collection method for device %s' %
hostname)
elif protocol == 'telnet':
if commandPort == 22: commandPort = 23 #set default telnet
client = TelnetClient(hostname,
ip,
commandPort,
options=self.options,
plugins=plugins,
device=device,
datacollector=self)
clientType = 'telnet'
self.log.info('Using telnet collection method for device %s' %
hostname)
else:
info = ("Unknown protocol %s for device %s -- "
"defaulting to %s collection method" %
(protocol, hostname, clientType))
self.log.warn(info)
import socket
component, _ = os.path.splitext(os.path.basename(sys.argv[0]))
collector_host = socket.gethostname()
evt = {
"eventClass": "/Status/Update",
"agent": collector_host,
"device": hostname,
"severity": Error
}
evt['summary'] = info
self.sendEvent(evt)
return
if not client:
self.log.warn("Shell command collector creation failed")
else:
self.log.info("plugins: %s",
", ".join(map(lambda p: p.name(), plugins)))
except Exception:
self.log.exception("Error opening command collector")
self.addClient(client, timeout, clientType, device.id)
def snmpCollect(self, device, ip, timeout):
"""
Start SNMP collection client.
@param device: device to collect against
@type device: string
@param ip: IP address of device to collect against
@type ip: string
@param timeout: timeout before failing the connection
@type timeout: integer
"""
client = None
try:
hostname = device.id
if getattr(device, "zSnmpMonitorIgnore", True):
self.log.info("SNMP monitoring off for %s" % hostname)
return
if not ip:
self.log.info("No manage IP for %s" % hostname)
return
plugins = []
plugins = self.selectPlugins(device, "snmp")
if not plugins:
self.log.info("No SNMP plugins found for %s" % hostname)
return
if self.checkCollection(device):
self.log.info('SNMP collection device %s' % hostname)
self.log.info("plugins: %s",
", ".join(map(lambda p: p.name(), plugins)))
client = SnmpClient(device.id, ip, self.options, device, self,
plugins)
if not client or not plugins:
self.log.warn("SNMP collector creation failed")
return
except Exception:
self.log.exception("Error opening the SNMP collector")
self.addClient(client, timeout, 'SNMP', device.id)
######## need to make async test for snmp work at some point -EAD #########
# def checkSnmpConnection(self, device):
# """
# Check to see if our current community string is still valid
#
# @param device: the device against which we will check
# @type device: a Device instance
# @return: result is None or a tuple containing
# (community, port, version, snmp name)
# @rtype: deferred: Twisted deferred
# """
# from pynetsnmp.twistedsnmp import AgentProxy
#
# def inner(driver):
# self.log.debug("Checking SNMP community %s on %s",
# device.zSnmpCommunity, device.id)
#
# oid = ".1.3.6.1.2.1.1.5.0"
# proxy = AgentProxy(device.id,
# device.zSnmpPort,
# timeout=device.zSnmpTimeout,
# community=device.zSnmpCommunity,
# snmpVersion=device.zSnmpVer,
# tries=2)
# proxy.open()
# yield proxy.get([oid])
# devname = driver.next().values()[0]
# if devname:
# yield succeed(True)
# yield succeed(False)
#
# return drive(inner)
def addClient(self, device, timeout, clientType, name):
"""
If device is not None, schedule the device to be collected.
Otherwise log an error.
@param device: device to collect against
@type device: string
@param timeout: timeout before failing the connection
@type timeout: integer
@param clientType: description of the plugin type
@type clientType: string
@param name: plugin name
@type name: string
"""
if device:
device.timeout = timeout
device.timedOut = False
self.clients.append(device)
device.run()
else:
self.log.warn('Unable to create a %s collector for %s', clientType,
name)
# XXX double-check this, once the implementation is in place
def portscanCollect(self, device, ip, timeout):
"""
Start portscan collection client.
@param device: device to collect against
@type device: string
@param ip: IP address of device to collect against
@type ip: string
@param timeout: timeout before failing the connection
@type timeout: integer
"""
client = None
try:
hostname = device.id
plugins = self.selectPlugins(device, "portscan")
if not plugins:
self.log.info("No portscan plugins found for %s" % hostname)
return
if self.checkCollection(device):
self.log.info('Portscan collector method for device %s' %
hostname)
self.log.info("plugins: %s",
", ".join(map(lambda p: p.name(), plugins)))
client = PortscanClient(device.id, ip, self.options, device,
self, plugins)
if not client or not plugins:
self.log.warn("Portscan collector creation failed")
return
except Exception:
self.log.exception("Error opening portscan collector")
self.addClient(client, timeout, 'portscan', device.id)
def checkCollection(self, device):
"""
See how old the data is that we've collected
@param device: device to collect against
@type device: string
@return: is the last collection time + collage older than now or is the SNMP status number > 0 ?
@type: boolean
"""
delay = device.getSnmpLastCollection() + self.collage
if delay >= float(
DateTime.DateTime()) and device.getSnmpStatusNumber() == 0:
self.log.info("Skipped collection of %s" % device.id)
return False
return True
def clientFinished(self, collectorClient):
"""
Callback that processes the return values from a device.
Python iterable.
@param collectorClient: collector instance
@type collectorClient: collector class
@return: Twisted deferred object
@type: Twisted deferred object
"""
device = collectorClient.device
self.log.debug("Client for %s finished collecting", device.id)
def processClient(driver):
try:
if (isinstance(collectorClient, SnmpClient)
and collectorClient.connInfo.changed == True):
self.log.info(
"SNMP connection info for %s changed. Updating...",
device.id)
yield self.config().callRemote(
'setSnmpConnectionInfo', device.id,
collectorClient.connInfo.zSnmpVer,
collectorClient.connInfo.zSnmpPort,
collectorClient.connInfo.zSnmpCommunity)
driver.next()
pluginStats = {}
self.log.debug("Processing data for device %s", device.id)
devchanged = False
maps = []
for plugin, results in collectorClient.getResults():
if plugin is None: continue
self.log.debug("Processing plugin %s on device %s ...",
plugin.name(), device.id)
if not results:
self.log.warn("The plugin %s returned no results.",
plugin.name())
continue
if self.options.save_raw_results:
self.savePluginData(device.id, plugin.name(), 'raw',
results)
self.log.debug("Plugin %s results = %s", plugin.name(),
results)
datamaps = []
try:
results = plugin.preprocess(results, self.log)
if results:
datamaps = plugin.process(device, results,
self.log)
if datamaps:
pluginStats.setdefault(plugin.name(),
plugin.weight)
except (SystemExit, KeyboardInterrupt) as ex:
self.log.info("Plugin %s terminated due to external"
" signal (%s)" %
(plugin.name(), str(ex)))
continue
except Exception as ex:
# NB: don't discard the plugin, as it might be a
# temporary issue
# Also, report it against the device, rather than at
# a collector as it might be just for this device.
import socket
collector_host = socket.gethostname()
evt = {
"eventClass": "/Status/Update",
"agent": collector_host,
"device": device.id,
"severity": Error
}
info = "Problem while executing plugin %s" % plugin.name(
)
self.log.error(info)
evt['summary'] = info
info = traceback.format_exc()
self.log.error(info)
evt['message'] = info
self.sendEvent(evt)
continue
# allow multiple maps to be returned from one plugin
if not isinstance(datamaps, (list, tuple)):
datamaps = [
datamaps,
]
if datamaps:
newmaps = [m for m in datamaps if m]
for m in newmaps:
setattr(m, PLUGIN_NAME_ATTR, plugin.name())
if self.options.save_processed_results:
self.savePluginData(device.id, plugin.name(),
'processed', newmaps)
maps += newmaps
if maps:
deviceClass = Classifier.classifyDevice(
pluginStats, self.classCollectorPlugins)
# If self.single is True, then call singleApplyDataMaps
# instead of applyDataMaps.
if not self.single:
method = "applyDataMaps"
else:
method = "singleApplyDataMaps"
yield self.config().callRemote(method, device.id, maps,
deviceClass, True)
if driver.next():
devchanged = True
if devchanged:
self.log.info("Changes in configuration applied")
else:
self.log.info("No change in configuration detected")
except Exception as ex:
self.log.exception(ex)
raise
def processClientFinished(result):
"""
Called after the client collection finishes
@param result: object (unused)
@type result: object
"""
self.counters['modeledDevicesCount'] += 1
self._modeledDevicesMetric.mark()
# result is now the result of remote_applyDataMaps (from processClient)
if result and isinstance(result, (basestring, Failure)):
self.log.error("Client %s finished with message: %s" %
(device.id, result))
self._failuresMetric.increment()
else:
self.log.debug("Client %s finished" % device.id)
try:
self.clients.remove(collectorClient)
self.finished.append(collectorClient)
except ValueError:
self.log.debug(
"Client %s not found in in the list"
" of active clients", device.id)
self.log.info(
"Finished processing client within collector loop #{0:03d}".
format(self.collectorLoopIteration))
d = drive(self.fillCollectionSlots)
d.addErrback(self.fillError)
d = drive(processClient)
d.addBoth(processClientFinished)
def savePluginData(self, deviceName, pluginName, dataType, data):
filename = "/tmp/%s.%s.%s.pickle.gz" % (deviceName, pluginName,
dataType)
try:
with gzip.open(filename, 'wb') as fd:
pickle.dump(data, fd)
except Exception as ex:
self.log.warn("Unable to save data into file '%s': %s", filename,
ex)
def fillError(self, reason):
"""
Twisted errback routine to log an error when
unable to collect some data
@param reason: error message
@type reason: string
"""
self.log.error("Unable to fill collection slots: %s" % reason)
def cycleTime(self):
"""
Return our cycle time (in minutes)
@return: cycle time
@rtype: integer
"""
return self.modelerCycleInterval * 60
def heartbeat(self, ignored=None):
"""
Twisted keep-alive mechanism to ensure that
we're still connected to zenhub
@param ignored: object (unused)
@type ignored: object
"""
ARBITRARY_BEAT = 30
reactor.callLater(ARBITRARY_BEAT, self.heartbeat)
if self.options.cycle:
evt = dict(eventClass=Heartbeat,
component='zenmodeler',
device=self.options.monitor,
timeout=self.options.heartbeatTimeout)
self.sendEvent(evt)
self.niceDoggie(self.cycleTime())
# We start modeling from here to accomodate the startup delay.
if not self.started:
if self.immediate == 0 and self.startat:
# This stuff relies on ARBITRARY_BEAT being < 60s
if self.timeMatches():
self.started = True
self.log.info("Starting modeling...")
reactor.callLater(1, self.main)
elif not self.isMainScheduled:
self.isMainScheduled = True
reactor.callLater(self.cycleTime(), self.main)
else:
self.started = True
self.log.info("Starting modeling in %s seconds.",
self.startDelay)
reactor.callLater(self.startDelay, self.main)
def postStatisticsImpl(self):
# save modeled device rate
self.rrdStats.derive('modeledDevices',
self.counters['modeledDevicesCount'])
# save running count
self.rrdStats.gauge('modeledDevicesCount',
self.counters['modeledDevicesCount'])
def _getCountersFile(self):
return zenPath('var/%s_%s.pickle' % (
self.name,
self.options.monitor,
))
@property
def _devicegen_has_items(self):
"""check it self.devicegen (an iterator) is not empty and has at least
one value. doing this check changes the iterator, so this method
restores it to its original state before returning"""
result = False
if self.devicegen is not None:
try:
first = self.devicegen.next()
except StopIteration:
pass
else:
result = True
self.devicegen = chain([first], self.devicegen)
return result
def checkStop(self, unused=None):
"""
Check to see if there's anything to do.
If there isn't, report our statistics and exit.
@param unused: unused (unused)
@type unused: string
"""
if self.pendingNewClients or self.clients: return
if self._devicegen_has_items: return
if not self.mainLoopGotDeviceList:
return # ZEN-26637 to prevent race between checkStop and mainLoop
if self.start:
runTime = time.time() - self.start
self.start = None
if not self.didCollect:
self.log.info("Did not collect during collector loop")
self.log.info("Scan time: %0.2f seconds for collector loop #%03d",
runTime, self.collectorLoopIteration)
self.log.info(
"Scanned %d of %d devices during collector loop #%03d",
self.processedDevicesCount, self.iterationDeviceCount,
self.collectorLoopIteration)
devices = len(self.finished)
timedOut = len([c for c in self.finished if c.timedOut])
self.rrdStats.gauge('cycleTime', runTime)
self.rrdStats.gauge('devices', devices)
self.rrdStats.gauge('timedOut', timedOut)
if not self.options.cycle:
self.stop()
self.finished = []
def fillCollectionSlots(self, driver):
"""
An iterator which either returns a device to collect or
calls checkStop()
@param driver: driver object
@type driver: driver object
"""
count = len(self.clients)
while count < self.options.parallel and self._devicegen_has_items \
and not self.pendingNewClients:
self.pendingNewClients = True
try:
device = self.devicegen.next()
yield self.config().callRemote('getDeviceConfig', [device],
self.options.checkStatus)
# just collect one device, and let the timer add more
devices = driver.next()
if devices:
self.processedDevicesCount = self.processedDevicesCount + 1
self.log.info(
"Filled collection slots for %d of %d devices during collector loop #%03d",
self.processedDevicesCount, self.iterationDeviceCount,
self.collectorLoopIteration
) #TODO should this message be logged at debug level?
self.didCollect = True
d = devices[0]
if d.skipModelMsg:
self.log.info(d.skipModelMsg)
else:
self.collectDevice(d)
else:
self.log.info("Device %s not returned is it down?", device)
except StopIteration:
self.devicegen = None
finally:
self.pendingNewClients = False
break
update = len(self.clients)
if update != count and update != 1:
self.log.info('Running %d clients', update)
else:
self.log.debug('Running %d clients', update)
self.checkStop()
def timeMatches(self):
"""
Check whether the current time matches a cron-like
specification, return a straight true or false
"""
if self.startat is None:
return True
def match_entity(entity, value):
if entity == '*':
return True
value = int(value)
if entity.isdigit() and int(entity) == value:
return True
if entity.startswith('*/') and entity[2:].isdigit():
if value % int(entity[2:]) == 0:
return True
if ',' in entity and any(
segment.isdigit() and int(segment) == value
for segment in entity.split(',')):
return True
return False
curtime = time.localtime()
# match minutes, hours, date, and month fields
if all(
match_entity(self.startat[a], curtime[b])
for a, b in ((0, 4), (1, 3), (2, 2), (3, 1))):
dayofweek = curtime[6] + 1
if (match_entity(self.startat[4], dayofweek)
or dayofweek == 7 and match_entity(self.startat[4], 0)):
return True
return False
def buildOptions(self):
"""
Build our list of command-line options
"""
PBDaemon.buildOptions(self)
self.parser.add_option('--debug',
dest='debug',
action="store_true",
default=False,
help="Don't fork threads for processing")
self.parser.add_option('--nowmi',
dest='nowmi',
action="store_true",
default=not USE_WMI,
help="Do not execute WMI plugins")
self.parser.add_option(
'--parallel',
dest='parallel',
type='int',
default=defaultParallel,
help="Number of devices to collect from in parallel")
self.parser.add_option('--cycletime',
dest='cycletime',
type='int',
help="Run collection every x minutes")
self.parser.add_option(
'--ignore',
dest='ignorePlugins',
default="",
help="Modeler plugins to ignore. Takes a regular expression")
self.parser.add_option(
'--collect',
dest='collectPlugins',
default="",
help="Modeler plugins to use. Takes a regular expression")
self.parser.add_option(
'-p',
'--path',
dest='path',
help="Start class path for collection ie /NetworkDevices")
self.parser.add_option(
'-d',
'--device',
dest='device',
help="Fully qualified device name ie www.confmon.com")
self.parser.add_option('--startat',
dest='startat',
help="Start string in cron(8) format")
self.parser.add_option(
'-a',
'--collage',
dest='collage',
default=0,
type='float',
help="Do not collect from devices whose collect date " +
"is within this many minutes")
self.parser.add_option('--writetries',
dest='writetries',
default=2,
type='int',
help="Number of times to try to write if a "
"read conflict is found")
# FIXME: cleanup --force option #2660
self.parser.add_option(
"-F",
"--force",
dest="force",
action='store_true',
default=True,
help="Force collection of config data (deprecated)")
self.parser.add_option(
'--portscantimeout',
dest='portscantimeout',
type='int',
default=defaultPortScanTimeout,
help="Time to wait for connection failures when port scanning")
self.parser.add_option(
'--now',
dest='now',
action="store_true",
default=False,
help="Start daemon now, do not sleep before starting")
self.parser.add_option(
'--communities',
dest='discoverCommunity',
action="store_true",
default=False,
help=
"If an snmp connection fails try and rediscover it's connection info"
)
self.parser.add_option(
'--checkstatus',
dest='checkStatus',
action="store_true",
default=False,
help="Don't model if the device is ping or snmp down")
self.parser.add_option(
'--save_raw_results',
dest='save_raw_results',
action="store_true",
default=False,
help="Save raw results for replay purposes in /tmp")
self.parser.add_option(
'--save_processed_results',
dest='save_processed_results',
action="store_true",
default=False,
help="Save modeler plugin outputs for replay purposes in /tmp")
addWorkerOptions(self.parser)
TCbuildOptions(self.parser, self.usage)
if USE_WMI:
addNTLMv2Option(self.parser)
def processOptions(self):
"""
Check what the user gave us vs what we'll accept
for command-line options
"""
if not self.options.path and not self.options.device:
self.options.path = "/Devices"
if self.options.ignorePlugins and self.options.collectPlugins:
raise SystemExit("Only one of --ignore or --collect"
" can be used at a time")
if self.options.startat:
cronmatch = re.match(
'^\s*([\*/,\d]+)\s+([\*/,\d]+)\s+([\*/,\d]+)\s+([\*/,\d]+)\s+([\*/,\d]+)\s*$',
self.options.startat)
if cronmatch:
self.startat = cronmatch.groups()
else:
self.log.error(
'startat option "%s" was invalid, carrying on anyway',
self.options.startat)
if USE_WMI:
setNTLMv2Auth(self.options)
configFilter = parseWorkerOptions(self.options.__dict__)
if configFilter:
self.configFilter = configFilter
def _timeoutClients(self):
"""
The guts of the timeoutClients method (minus the twisted reactor
stuff). Breaking this part out as a separate method facilitates unit
testing.
"""
active = []
for client in self.clients:
if client.timeout < time.time():
self.log.warn("Client %s timeout", client.hostname)
self.finished.append(client)
client.timedOut = True
try:
client.stop()
except AssertionError:
pass # session closed twice http://dev.zenoss.org/trac/ticket/6354
else:
active.append(client)
self.clients = active
def timeoutClients(self, unused=None):
"""
Check to see which clients have timed out and which ones haven't.
Stop processing anything that's timed out.
@param unused: unused (unused)
@type unused: string
"""
reactor.callLater(1, self.timeoutClients)
self._timeoutClients()
d = drive(self.fillCollectionSlots)
d.addCallback(self.checkStop)
d.addErrback(self.fillError)
def reactorLoop(self):
"""
Twisted main loop
"""
reactor.startRunning()
while reactor.running:
try:
while reactor.running:
reactor.runUntilCurrent()
timeout = reactor.timeout()
reactor.doIteration(timeout)
except Exception:
if reactor.running:
self.log.exception("Unexpected error in main loop.")
def getDeviceList(self):
"""
Get the list of devices for which we are collecting:
* if -d devicename was used, use the devicename
* if a class path flag was supplied, gather the devices
along that organizer
* otherwise get all of the devices associated with our collector
@return: list of devices
@rtype: list
"""
if self.options.device:
self.log.info("Collecting for device %s", self.options.device)
return succeed([self.options.device])
self.log.info("Collecting for path %s", self.options.path)
d = self.config().callRemote('getDeviceListByOrganizer',
self.options.path, self.options.monitor,
self.options.__dict__)
def handle(results):
if hasattr(results, "type") and results.type is HubDown:
self.log.warning("Collection aborted: %s",
results.getErrorMessage())
return
self.log.critical("%s is not a valid organizer.",
self.options.path)
reactor.running = False
sys.exit(1)
d.addErrback(handle)
return d
def mainLoop(self, driver):
"""
Main collection loop, a Python iterable
@param driver: driver object
@type driver: driver object
@return: Twisted deferred object
@rtype: Twisted deferred object
"""
if self.options.cycle:
self.isMainScheduled = True
driveLater(self.cycleTime(), self.mainLoop)
if self.clients:
self.log.error("Modeling cycle taking too long")
return
# ZEN-26637 - did we collect during collector loop?
self.didCollect = False
self.mainLoopGotDeviceList = False
self.start = time.time()
self.collectorLoopIteration = self.collectorLoopIteration + 1
self.log.info("Starting collector loop #{:03d}...".format(
self.collectorLoopIteration))
yield self.getDeviceList()
deviceList = driver.next()
self.log.debug("getDeviceList returned %s devices", len(deviceList))
self.log.debug("getDeviceList returned %s devices", deviceList)
self.devicegen = iter(deviceList)
self.iterationDeviceCount = len(deviceList)
self.processedDevicesCount = 0
self.log.info(
"Got %d devices to be scanned during collector loop #%03d",
self.iterationDeviceCount, self.collectorLoopIteration)
d = drive(self.fillCollectionSlots)
d.addErrback(self.fillError)
self.mainLoopGotDeviceList = True
yield d
driver.next()
self.log.debug("Collection slots filled")
def main(self, unused=None):
"""
Wrapper around the mainLoop
@param unused: unused (unused)
@type unused: string
@return: Twisted deferred object
@rtype: Twisted deferred object
"""
self.finished = []
d = drive(self.mainLoop)
d.addCallback(self.timeoutClients)
return d
def remote_deleteDevice(self, device):
"""
Stub function
@param device: device name (unused)
@type device: string
@todo: implement
"""
# we fetch the device list before every scan
self.log.debug("Asynch deleteDevice %s" % device)
def remote_deleteDevices(self, devices):
"""
Stub function
@param devices: device ids (unused)
@type device: set
@todo: implement
"""
# we fetch the device list before every scan
self.log.debug("Asynch deleteDevices {0}".format(len(devices)))
class PBDaemon(ZenDaemon, pb.Referenceable):
name = 'pbdaemon'
initialServices = ['EventService']
heartbeatEvent = {'eventClass':Heartbeat}
heartbeatTimeout = 60*3
_customexitcode = 0
_pushEventsDeferred = None
def __init__(self, noopts=0, keeproot=False, name=None):
# if we were provided our collector name via the constructor instead of
# via code, be sure to store it correctly.
if name is not None:
self.name = name
self.mname = name
try:
ZenDaemon.__init__(self, noopts, keeproot)
except IOError:
import traceback
self.log.critical( traceback.format_exc( 0 ) )
sys.exit(1)
self.rrdStats = DaemonStats()
self.lastStats = 0
self.perspective = None
self.services = {}
self.eventQueueManager = EventQueueManager(self.options, self.log)
self.startEvent = startEvent.copy()
self.stopEvent = stopEvent.copy()
details = dict(component=self.name, device=self.options.monitor)
for evt in self.startEvent, self.stopEvent, self.heartbeatEvent:
evt.update(details)
self.initialConnect = defer.Deferred()
self.stopped = False
self.counters = collections.Counter()
self.loadCounters()
self._pingedZenhub = None
self._connectionTimeout = None
# Add a shutdown trigger to send a stop event and flush the event queue
reactor.addSystemEventTrigger('before', 'shutdown', self._stopPbDaemon)
def connecting(self):
"""
Called when about to connect to zenhub
"""
self.log.info("Attempting to connect to zenhub")
def gotPerspective(self, perspective):
"""
This gets called every time we reconnect.
@parameter perspective: Twisted perspective object
@type perspective: Twisted perspective object
"""
self.log.info("Connected to ZenHub")
self.perspective = perspective
# Cancel the connection timeout timer as it's no longer needed.
if self._connectionTimeout:
try:
self._connectionTimeout.cancel()
except AlreadyCalled:
pass
self._connectionTimeout = None
d2 = self.getInitialServices()
if self.initialConnect:
self.log.debug('Chaining getInitialServices with d2')
self.initialConnect, d = None, self.initialConnect
d2.chainDeferred(d)
def connect(self):
factory = ReconnectingPBClientFactory(connectTimeout=60)
self.log.info("Connecting to %s:%d" % (self.options.hubhost, self.options.hubport))
factory.connectTCP(self.options.hubhost, self.options.hubport)
username = self.options.hubusername
password = self.options.hubpassword
self.log.debug("Logging in as %s" % username)
c = credentials.UsernamePassword(username, password)
factory.gotPerspective = self.gotPerspective
factory.connecting = self.connecting
factory.startLogin(c)
def timeout(d):
if not d.called:
self.connectTimeout()
self._connectionTimeout = reactor.callLater(
self.options.hubtimeout, timeout, self.initialConnect)
return self.initialConnect
def connectTimeout(self):
self.log.error('Timeout connecting to zenhub: is it running?')
pass
def eventService(self):
return self.getServiceNow('EventService')
def getServiceNow(self, svcName):
if not svcName in self.services:
self.log.warning('No service named %r: ZenHub may be disconnected' % svcName)
return self.services.get(svcName, None) or FakeRemote()
def getService(self, serviceName, serviceListeningInterface=None):
"""
Attempt to get a service from zenhub. Returns a deferred.
When service is retrieved it is stashed in self.services with
serviceName as the key. When getService is called it will first
check self.services and if serviceName is already there it will return
the entry from self.services wrapped in a defer.succeed
"""
if serviceName in self.services:
return defer.succeed(self.services[serviceName])
def removeService(ignored):
self.log.debug('Removing service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
def callback(result, serviceName):
self.log.debug('Loaded service %s from zenhub' % serviceName)
self.services[serviceName] = result
result.notifyOnDisconnect(removeService)
return result
def errback(error, serviceName):
self.log.debug('errback after getting service %s' % serviceName)
self.log.error('Could not retrieve service %s' % serviceName)
if serviceName in self.services:
del self.services[serviceName]
return error
d = self.perspective.callRemote('getService',
serviceName,
self.options.monitor,
serviceListeningInterface or self)
d.addCallback(callback, serviceName)
d.addErrback(errback, serviceName)
return d
def getInitialServices(self):
"""
After connecting to zenhub, gather our initial list of services.
"""
def errback(error):
if isinstance(error, Failure):
self.log.critical( "Invalid monitor: %s" % self.options.monitor)
reactor.stop()
return defer.fail(RemoteBadMonitor(
"Invalid monitor: %s" % self.options.monitor, ''))
return error
self.log.debug('Setting up initial services: %s' % \
', '.join(self.initialServices))
d = defer.DeferredList(
[self.getService(name) for name in self.initialServices],
fireOnOneErrback=True, consumeErrors=True)
d.addErrback(errback)
return d
def connected(self):
pass
def run(self):
self.rrdStats.config(self.options.monitor, self.name, [])
self.log.debug('Starting PBDaemon initialization')
d = self.connect()
def callback(result):
self.sendEvent(self.startEvent)
self.pushEventsLoop()
self.log.debug('Calling connected.')
self.connected()
return result
d.addCallback(callback)
d.addErrback(twisted.python.log.err)
reactor.run()
if self._customexitcode:
sys.exit(self._customexitcode)
def setExitCode(self, exitcode):
self._customexitcode = exitcode
def stop(self, ignored=''):
if reactor.running:
try:
reactor.stop()
except ReactorNotRunning:
self.log.debug("Tried to stop reactor that was stopped")
else:
self.log.debug("stop() called when not running")
def _stopPbDaemon(self):
if self.stopped:
return
self.stopped = True
if 'EventService' in self.services:
# send stop event if we don't have an implied --cycle,
# or if --cycle has been specified
if not hasattr(self.options, 'cycle') or\
getattr(self.options, 'cycle', True):
self.sendEvent(self.stopEvent)
self.log.debug("Sent a 'stop' event")
if self._pushEventsDeferred:
self.log.debug("Currently sending events. Queueing next call")
d = self._pushEventsDeferred
# Schedule another call to flush any additional queued events
d.addBoth(lambda unused: self.pushEvents())
else:
d = self.pushEvents()
d.addBoth(lambda unused: self.saveCounters())
return d
self.log.debug("No event sent as no EventService available.")
self.saveCounters()
def sendEvents(self, events):
map(self.sendEvent, events)
def sendEvent(self, event, **kw):
''' Add event to queue of events to be sent. If we have an event
service then process the queue.
'''
generatedEvent = self.generateEvent(event, **kw)
self.eventQueueManager.addEvent(generatedEvent)
self.counters['eventCount'] += 1
def generateEvent(self, event, **kw):
''' Add event to queue of events to be sent. If we have an event
service then process the queue.
'''
if not reactor.running: return
event = event.copy()
event['agent'] = self.name
event['monitor'] = self.options.monitor
event['manager'] = self.fqdn
event.update(kw)
return event
@defer.inlineCallbacks
def pushEventsLoop(self):
"""Periodially, wake up and flush events to ZenHub.
"""
reactor.callLater(self.options.eventflushseconds, self.pushEventsLoop)
yield self.pushEvents()
# Record the number of events in the queue every 5 minutes.
now = time.time()
if self.rrdStats.name and now >= (self.lastStats + 300):
self.lastStats = now
events = self.rrdStats.gauge('eventQueueLength',
300, self.eventQueueManager.event_queue_length)
for event in events:
self.eventQueueManager.addPerformanceEvent(event)
@defer.inlineCallbacks
def pushEvents(self):
"""Flush events to ZenHub.
"""
# are we already shutting down?
if not reactor.running:
self.log.debug("Skipping event sending - reactor not running.")
return
if self._pushEventsDeferred:
self.log.debug("Skipping event sending - previous call active.")
return
try:
self._pushEventsDeferred = defer.Deferred()
# are still connected to ZenHub?
evtSvc = self.services.get('EventService', None)
if not evtSvc:
self.log.error("No event service: %r", evtSvc)
return
discarded_events = self.eventQueueManager.discarded_events
if discarded_events:
self.log.error(
'Discarded oldest %d events because maxqueuelen was '
'exceeded: %d/%d',
discarded_events,
discarded_events + self.options.maxqueuelen,
self.options.maxqueuelen)
self.counters['discardedEvents'] += discarded_events
self.eventQueueManager.discarded_events = 0
send_events_fn = partial(evtSvc.callRemote, 'sendEvents')
try:
yield self.eventQueueManager.sendEvents(send_events_fn)
except ConnectionLost as ex:
self.log.error('Error sending event: %s', ex)
except ConnectionDone:
pass
except Exception as ex:
self.log.exception(ex)
finally:
d, self._pushEventsDeferred = self._pushEventsDeferred, None
d.callback('sent')
def heartbeat(self):
'if cycling, send a heartbeat, else, shutdown'
if not self.options.cycle:
self.stop()
return
heartbeatEvent = self.generateEvent(self.heartbeatEvent, timeout=self.heartbeatTimeout)
self.eventQueueManager.addHeartbeatEvent(heartbeatEvent)
# heartbeat is normally 3x cycle time
self.niceDoggie(self.heartbeatTimeout / 3)
events = []
# save daemon counter stats
for name, value in self.counters.items():
self.log.info("Counter %s, value %d", name, value)
events += self.rrdStats.counter(name, 300, value)
self.sendEvents(events)
# persist counters values
self.saveCounters()
def saveCounters(self):
atomicWrite(
zenPath('var/%s_counters.pickle' % self.name),
pickle.dumps(self.counters),
raiseException=False,
)
def loadCounters(self):
try:
self.counters = pickle.load(open(zenPath('var/%s_counters.pickle'% self.name)))
except Exception:
pass
def remote_getName(self):
return self.name
def remote_shutdown(self, unused):
self.stop()
self.sigTerm()
def remote_setPropertyItems(self, items):
pass
@translateError
def remote_updateThresholdClasses(self, classes):
from Products.ZenUtils.Utils import importClass
self.log.debug("Loading classes %s", classes)
for c in classes:
try:
importClass(c)
except ImportError:
self.log.error("Unable to import class %s", c)
def buildOptions(self):
self.parser.add_option('--hubhost',
dest='hubhost',
default=DEFAULT_HUB_HOST,
help='Host of zenhub daemon.'
' Default is %s.' % DEFAULT_HUB_HOST)
self.parser.add_option('--hubport',
dest='hubport',
type='int',
default=DEFAULT_HUB_PORT,
help='Port zenhub listens on.'
'Default is %s.' % DEFAULT_HUB_PORT)
self.parser.add_option('--hubusername',
dest='hubusername',
default=DEFAULT_HUB_USERNAME,
help='Username for zenhub login.'
' Default is %s.' % DEFAULT_HUB_USERNAME)
self.parser.add_option('--hubpassword',
dest='hubpassword',
default=DEFAULT_HUB_PASSWORD,
help='Password for zenhub login.'
' Default is %s.' % DEFAULT_HUB_PASSWORD)
self.parser.add_option('--monitor',
dest='monitor',
default=DEFAULT_HUB_MONITOR,
help='Name of monitor instance to use for'
' configuration. Default is %s.'
% DEFAULT_HUB_MONITOR)
self.parser.add_option('--initialHubTimeout',
dest='hubtimeout',
type='int',
default=30,
help='Initial time to wait for a ZenHub '
'connection')
self.parser.add_option('--allowduplicateclears',
dest='allowduplicateclears',
default=False,
action='store_true',
help='Send clear events even when the most '
'recent event was also a clear event.')
self.parser.add_option('--duplicateclearinterval',
dest='duplicateclearinterval',
default=0,
type='int',
help=('Send a clear event every [DUPLICATECLEARINTEVAL] '
'events.')
)
self.parser.add_option('--eventflushseconds',
dest='eventflushseconds',
default=5.,
type='float',
help='Seconds between attempts to flush '
'events to ZenHub.')
self.parser.add_option('--eventflushchunksize',
dest='eventflushchunksize',
default=50,
type='int',
help='Number of events to send to ZenHub'
'at one time')
self.parser.add_option('--maxqueuelen',
dest='maxqueuelen',
default=5000,
type='int',
help='Maximum number of events to queue')
self.parser.add_option('--zenhubpinginterval',
dest='zhPingInterval',
default=30,
type='int',
help='How often to ping zenhub')
self.parser.add_option('--disable-event-deduplication',
dest='deduplicate_events',
default=True,
action='store_false',
help='Disable event de-duplication')
ZenDaemon.buildOptions(self)
class EventServer(PBDaemon):
'Base class for a daemon whose primary job is to post events'
name = 'EventServer'
def __init__(self):
PBDaemon.__init__(self, keeproot=True)
self.stats = Stats()
self.rrdStats = DaemonStats()
def connected(self):
self.sendEvent(dict(device=self.options.monitor,
eventClass=App_Start,
summary="%s started" % self.name,
severity=0,
component=self.name))
self.log.info("started")
self.configure()
def model(self):
return self.services.get('EventService', FakeRemote())
def configure(self):
def inner(driver):
self.log.info("fetching default RRDCreateCommand")
yield self.model().callRemote('getDefaultRRDCreateCommand')
createCommand = driver.next()
self.log.info("getting threshold classes")
yield self.model().callRemote('getThresholdClasses')
self.remote_updateThresholdClasses(driver.next())
self.log.info("getting collector thresholds")
yield self.model().callRemote('getCollectorThresholds')
self.rrdStats.config(self.options.monitor, self.name,
driver.next(), createCommand)
self.heartbeat()
self.reportCycle()
d = drive(inner)
def error(result):
self.log.error("Unexpected error in configure: %s" % result)
d.addErrback(error)
return d
def sendEvent(self, event, **kw):
# FIXME: get real event processing stats
if 'firstTime' in event:
self.stats.add(min(time.time() - event['firstTime'], 0))
PBDaemon.sendEvent(self, event, **kw)
def useUdpFileDescriptor(self, fd):
from twisted.internet import udp
s = socket.fromfd(fd, socket.AF_INET, socket.SOCK_DGRAM)
import os
os.close(fd)
port = s.getsockname()[1]
transport = udp.Port(port, self)
s.setblocking(0)
transport.socket = s
transport.fileno = s.fileno
transport.connected = 1
transport._realPortNumber = port
self.transport = transport
# hack around startListening not being called
self.numPorts = 1
transport.startReading()
def useTcpFileDescriptor(self, fd, factory):
import os, socket
for i in range(19800, 19999):
try:
p = reactor.listenTCP(i, factory)
os.dup2(fd, p.socket.fileno())
p.socket.listen(p.backlog)
p.socket.setblocking(False)
p.socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
os.close(fd)
return p
except socket.error:
pass
raise socket.error("Unable to find an open socket to listen on")
def reportCycle(self):
if self.options.statcycle:
self.report()
reactor.callLater(self.options.statcycle, self.reportCycle)
def heartbeat(self):
"""Since we don't do anything on a regular basis, just
push heartbeats regularly"""
seconds = self.heartbeatTimeout / 3
reactor.callLater(self.heartbeatTimeout / 3, self.heartbeat)
PBDaemon.heartbeat(self)
totalTime, totalEvents, maxTime = self.stats.report()
for ev in (self.rrdStats.counter('events',
seconds,
totalEvents) +
self.rrdStats.counter('totalTime',
seconds,
int(totalTime * 1000))):
self.sendEvent(ev)
def report(self):
'report some simple diagnostics at shutdown'
totalTime, totalEvents, maxTime = self.stats.report()
self.log.info("%d events processed in %.2f seconds",
totalEvents,
totalTime)
if totalEvents > 0:
self.log.info("%.5f average seconds per event",
(totalTime / totalEvents))
self.log.info("Maximum processing time for one event was %.5f",
maxTime)
def buildOptions(self):
PBDaemon.buildOptions(self)
self.parser.add_option('--statcycle',
dest='statcycle',
type='int',
help='Number of seconds between the writing of statistics',
default=0)