def cancel(self, request_id):
"""
Immediately mark a request as cancelled, and in the background
try and cancel any outstanding JID for it.
"""
request = self._by_request_id[request_id]
# Idempotent behaviour: no-op if already cancelled
if request.state == request.COMPLETE:
return
with self._update_index(request):
# I will take over cancelling the JID from the request
cancel_jid = request.jid
request.jid = None
# Request is now done, no further calls
request.set_error("Cancelled")
request.complete()
# In the background, try to cancel the request's JID on a best-effort basis
if cancel_jid:
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
client.run_job(request.minion_id, 'saltutil.kill_job',
[cancel_jid])
def fetch(self, minion_id, sync_type):
log.debug("SyncObjects.fetch: %s/%s" % (minion_id, sync_type))
if minion_id is None:
# We're probably being replayed to from the database
log.warn("SyncObjects.fetch called with minion_id=None")
return
self._fetching_at[sync_type] = now()
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
# TODO clean up unused 'since' argument
pub_data = client.run_job(
minion_id, 'ceph.get_cluster_object',
condition_kwarg(
[], {
'cluster_name': self._cluster_name,
'sync_type': sync_type.str,
'since': None
}))
if not pub_data:
log.error("Failed to start fetch job %s/%s" %
(minion_id, sync_type))
# Don't throw an exception because if a fetch fails we should always
else:
log.debug("SyncObjects.fetch: jid=%s minions=%s" %
(pub_data['jid'], pub_data['minions']))
def get_running(self, minions):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(minions,
'saltutil.running', [],
expr_form="list")
if not pub_data:
log.warning(
"Failed to publish saltutil.running to {0}".format(minions))
def run_job(self, fqdn, cmd, args):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(fqdn, cmd, condition_kwarg([], args))
if not pub_data:
# FIXME: LocalClient uses 'print' to record the
# details of what went wrong :-(
raise Unavailable()
else:
return pub_data['jid']
def run_job(self, fqdn, cmd, args):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(fqdn, cmd, condition_kwarg([], args))
if not pub_data:
# FIXME: LocalClient uses 'print' to record the
# details of what went wrong :-(
raise Unavailable()
else:
return pub_data['jid']
def _submit(self):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(self._minion_id, self._cmd, self._args)
if not pub_data:
# FIXME: LocalClient uses 'print' to record the
# details of what went wrong :-(
raise PublishError("Failed to publish job")
self.log.info("Request %s started job %s" % (self.id, pub_data['jid']))
self.alive_at = now()
self.jid = pub_data['jid']
return self.jid
def tick(self):
"""
For walltime-based monitoring of running requests. Long-running requests
get a periodic call to saltutil.running to verify that things really
are still happening.
"""
if not self._by_jid:
return
else:
log.debug("RequestCollection.tick: %s JIDs underway" %
len(self._by_jid))
# Identify JIDs who haven't had a saltutil.running reponse for too long.
# Kill requests in a separate phase because request:JID is not 1:1
stale_jobs = set()
_now = now()
for request in self._by_jid.values():
if _now - request.alive_at > datetime.timedelta(
seconds=TICK_PERIOD * 3):
log.error("Request %s JID %s stale: now=%s, alive_at=%s" %
(request.id, request.jid, _now, request.alive_at))
stale_jobs.add(request)
# Any identified stale jobs are errored out.
for request in stale_jobs:
with self._update_index(request):
request.set_error("Lost contact")
request.jid = None
request.complete()
# Identify minions associated with JIDs in flight
query_minions = set()
for jid, request in self._by_jid.items():
query_minions.add(request.minion_id)
# Attempt to emit a saltutil.running to ping jobs, next tick we
# will see if we got updates to the alive_at attribute to indicate non-staleness
if query_minions:
log.info("RequestCollection.tick: sending saltutil.running to {0}".
format(query_minions))
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(list(query_minions),
'saltutil.running', [],
expr_form="list")
if not pub_data:
log.warning("Failed to publish saltutil.running to {0}".format(
query_minions))
def fetch(self, minion_id, sync_type):
log.debug("SyncObjects.fetch: %s/%s" % (minion_id, sync_type))
if minion_id is None:
# We're probably being replayed to from the database
log.warn("SyncObjects.fetch called with minion_id=None")
return
self._fetching_at[sync_type] = now()
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
# TODO clean up unused 'since' argument
pub_data = client.run_job(minion_id, 'ceph.get_cluster_object',
condition_kwarg([], {'cluster_name': self._cluster_name,
'sync_type': sync_type.str,
'since': None}))
if not pub_data:
log.error("Failed to start fetch job %s/%s" % (minion_id, sync_type))
# Don't throw an exception because if a fetch fails we should always
else:
log.debug("SyncObjects.fetch: jid=%s minions=%s" % (pub_data['jid'], pub_data['minions']))
def tick(self):
"""
For walltime-based monitoring of running requests. Long-running requests
get a periodic call to saltutil.running to verify that things really
are still happening.
"""
if not self._by_jid:
return
else:
log.debug("RequestCollection.tick: %s JIDs underway" % len(self._by_jid))
# Identify JIDs who haven't had a saltutil.running reponse for too long.
# Kill requests in a separate phase because request:JID is not 1:1
stale_jobs = set()
_now = now()
for request in self._by_jid.values():
if _now - request.alive_at > datetime.timedelta(seconds=TICK_PERIOD * 3):
log.error("Request %s JID %s stale: now=%s, alive_at=%s" % (
request.id, request.jid, _now, request.alive_at
))
stale_jobs.add(request)
# Any identified stale jobs are errored out.
for request in stale_jobs:
with self._update_index(request):
request.set_error("Lost contact")
request.jid = None
request.complete()
# Identify minions associated with JIDs in flight
query_minions = set()
for jid, request in self._by_jid.items():
query_minions.add(request.minion_id)
# Attempt to emit a saltutil.running to ping jobs, next tick we
# will see if we got updates to the alive_at attribute to indicate non-staleness
if query_minions:
log.info("RequestCollection.tick: sending saltutil.running to {0}".format(query_minions))
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(list(query_minions), 'saltutil.running', [], expr_form="list")
if not pub_data:
log.warning("Failed to publish saltutil.running to {0}".format(query_minions))
def _submit(self, commands=None):
if commands is None:
commands = self._commands
self.log.debug("%s._submit: %s/%s/%s" % (self.__class__.__name__,
self._minion_id, self._cluster_name, commands))
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(self._minion_id, 'ceph.rados_commands',
[self.fsid, self._cluster_name, commands])
if not pub_data:
# FIXME: LocalClient uses 'print' to record the
# details of what went wrong :-(
raise PublishError("Failed to publish job")
self.log.info("Request %s started job %s" % (self.id, pub_data['jid']))
self.alive_at = now()
self.jid = pub_data['jid']
return self.jid
def cancel(self, fqdn, jid):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
client.run_job(fqdn, 'saltutil.kill_job', [jid])
def cancel(self, fqdn, jid):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
client.run_job(fqdn, 'saltutil.kill_job', [jid])
def get_running(self, minions):
client = LocalClient(config.get('cthulhu', 'salt_config_path'))
pub_data = client.run_job(minions, 'saltutil.running', [], expr_form="list")
if not pub_data:
log.warning("Failed to publish saltutil.running to {0}".format(minions))