def _run(self):
log.info("Starting %s" % self.__class__.__name__)
subscription = SaltEventSource(log, salt_config)
while not self._complete.is_set():
# No salt tag filtering: https://github.com/saltstack/salt/issues/11582
ev = subscription.get_event(full=True)
if ev is not None and ev['tag'].startswith("ceph/server"):
data = ev['data']
log.debug("ServerMonitor got ceph/server message from %s" % data['id'])
try:
# NB assumption that FQDN==minion_id is true unless
# someone has modded their salt minion config.
self.on_server_heartbeat(data['id'], data['data'])
except:
log.debug("Message detail: %s" % json.dumps(ev))
log.exception("Error handling ceph/server message from %s" % data['id'])
if ev is not None and ev['tag'].startswith("salt/presen"):
# so these exist but i'm not convinced they work, I started a minion
# and then saw several messages indicating no mininions were present
log.debug("ServerMonitor: presence %s" % ev['data'])
log.info("Completed %s" % self.__class__.__name__)
def listen(self, completion,
on_heartbeat=None,
on_job=None,
on_server_heartbeat=None,
on_running_jobs=None,
fsid=None):
"""
:param on_heartbeat: Callback for heartbeats
:param on_job: Callback for job completions
:param fsid: Optionally filter heartbeats to one FSID
"""
event = SaltEventSource(log, salt_config)
if fsid:
heartbeat_tag = "ceph/cluster/{0}".format(fsid)
else:
heartbeat_tag = "ceph/cluster/"
while not completion.is_set():
# No salt tag filtering: https://github.com/saltstack/salt/issues/11582
ev = event.get_event(full=True)
if ev is not None:
data = ev['data']
tag = ev['tag']
log.debug("_run.ev: %s/tag=%s" % (data['id'] if 'id' in data else None, tag))
# I am interested in the following tags:
# - salt/job/<jid>/ret/<minion id> where jid is one that I started
# (this includes ceph.rados_command and ceph.get_cluster_object)
# - ceph/cluster/<fsid> where fsid is my fsid
try:
if tag.startswith(heartbeat_tag):
if on_heartbeat:
on_heartbeat(data['id'], data['data'])
elif tag.startswith("ceph/server"):
if on_server_heartbeat:
on_server_heartbeat(data['id'], data['data'])
elif re.match("^salt/job/\d+/ret/[^/]+$", tag):
if data['fun'] == 'saltutil.running':
if on_running_jobs and data['success']:
on_running_jobs(data['id'], data['return'])
else:
if on_job:
on_job(data['id'], data['jid'], data['success'], data['return'],
data['fun'], data['fun_args'])
else:
# This does not concern us, ignore it
pass
except:
# Because this is our main event handling loop, swallow exceptions
# instead of letting them end the world.
log.exception("Exception handling message with tag %s" % tag)
log.debug("Message content: %s" % data)
def _run(self):
log.info("%s running" % self.__class__.__name__)
event = SaltEventSource(log, salt_config)
while not self._complete.is_set():
# No salt tag filtering: https://github.com/saltstack/salt/issues/11582
ev = event.get_event(full=True)
if ev is not None and 'tag' in ev:
tag = ev['tag']
data = ev['data']
try:
if tag.startswith("ceph/cluster/"):
cluster_data = data['data']
if not cluster_data['fsid'] in self._manager.clusters:
self._manager.on_discovery(data['id'],
cluster_data)
else:
log.debug(
"%s: heartbeat from existing cluster %s" %
(self.__class__.__name__,
cluster_data['fsid']))
elif re.match("^salt/job/\d+/ret/[^/]+$", tag):
if data['fun'] == 'saltutil.running':
self._manager.requests.on_tick_response(
data['id'], data['return'])
else:
self._manager.requests.on_completion(data)
else:
# This does not concern us, ignore it
log.debug("TopLevelEvents: ignoring %s" % tag)
pass
except:
log.exception("Exception handling message tag=%s" % tag)
log.info("%s complete" % self.__class__.__name__)
def _run(self):
self._plugin_monitor.start()
self._ready.set()
log.debug("ClusterMonitor._run: ready")
event = SaltEventSource(log, salt_config)
while not self._complete.is_set():
# No salt tag filtering: https://github.com/saltstack/salt/issues/11582
ev = event.get_event(full=True)
if ev is not None:
data = ev['data']
tag = ev['tag']
log.debug("_run.ev: %s/tag=%s" % (data['id'] if 'id' in data else None, tag))
# I am interested in the following tags:
# - salt/job/<jid>/ret/<minion id> where jid is one that I started
# (this includes ceph.rados_command and ceph.get_cluster_object)
# - ceph/cluster/<fsid> where fsid is my fsid
try:
if tag.startswith("ceph/cluster/{0}".format(self.fsid)):
# A ceph.heartbeat beacon
self.on_heartbeat(data['id'], data['data'])
elif re.match("^salt/job/\d+/ret/[^/]+$", tag):
if data['fun'] == "saltutil.running":
# Update on what jobs are running
# It would be nice to filter these down to those which really are for
# this cluster, but as long as N_clusters and N_jobs are reasonably small
# it's not an efficiency problem.
self._requests.on_tick_response(data['id'], data['return'])
# It would be much nicer to put the FSID at the start of
# the tag, if salt would only let us add custom tags to our jobs.
# Instead we enforce a convention that all calamari jobs must include
# fsid in their return value.
if (not isinstance(data, dict)) or not isinstance(data['return'], dict):
# Something not formatted for ClusterMonitor
log.warning("Ignoring event %s" % tag)
continue
if 'fsid' not in data['return'] or data['return']['fsid'] != self.fsid:
# Something for a different ClusterMonitor
log.debug("Ignoring job return, not for my FSID")
continue
if data['fun'] == 'ceph.get_cluster_object':
# A ceph.get_cluster_object response
if not data['success']:
log.error("on_sync_object: failure from %s: %s" % (data['id'], data['return']))
continue
self.on_sync_object(data['id'], data['return'])
else:
log.warning("Unexpected function '%s' (%s)" % (data['fun'], tag))
else:
# This does not concern us, ignore it
pass
except:
# Because this is our main event handling loop, swallow exceptions
# instead of letting them end the world.
log.exception("Exception handling message with tag %s" % tag)
log.debug("Message content: %s" % data)
log.info("%s complete" % self.__class__.__name__)
self._plugin_monitor.stop()
self._plugin_monitor.join()
self.done.set()