def __init__(self, persister, eventer, requests):
super(ServerMonitor, self).__init__()
# FQDN to ServerState
self.servers = {}
# Hostname to ServerState
self.hostname_to_server = {}
# FSID to ServiceState list
self.fsid_services = defaultdict(list)
# Service (fsid, type, id) to ServiceState
self.services = {}
self._complete = event.Event()
self._eventer = eventer
self._persister = persister
self._requests = requests
# Cache things we look up from pillar to avoid hitting disk repeatedly
self._contact_period_cache = {}
self.remote = get_remote()
def __init__(self, persister, eventer, requests):
super(ServerMonitor, self).__init__()
# FQDN to ServerState
self.servers = {}
# Hostname to ServerState
self.hostname_to_server = {}
# FSID to ServiceState list
self.fsid_services = defaultdict(list)
# Service (fsid, type, id) to ServiceState
self.services = {}
self._complete = event.Event()
self._eventer = eventer
self._persister = persister
self._requests = requests
# Cache things we look up from pillar to avoid hitting disk repeatedly
self._contact_period_cache = {}
self.remote = get_remote()
def _run(self):
log.info("%s running" % self.__class__.__name__)
remote = get_remote()
remote.listen(self._complete,
on_heartbeat=self.on_heartbeat,
on_job=self.on_job,
on_running_jobs=self._manager.requests.on_tick_response)
log.info("%s complete" % self.__class__.__name__)
def __init__(self, manager):
super(RequestCollection, self).__init__()
self._by_request_id = {}
self._by_jid = {}
self._lock = RLock()
self._remote = get_remote()
self._manager = manager
def __init__(self, manager):
super(RequestCollection, self).__init__()
self._by_request_id = {}
self._by_jid = {}
self._lock = RLock()
self._remote = get_remote()
self._manager = manager
def _run(self):
log.info("%s running" % self.__class__.__name__)
remote = get_remote()
remote.listen(self._complete,
on_heartbeat=self.on_heartbeat,
on_job=self.on_job,
on_running_jobs=self._manager.requests.on_tick_response)
log.info("%s complete" % self.__class__.__name__)
from calamari_common.remote import get_remote, Unavailable
from cthulhu.manager import config
from cthulhu.gevent_util import nosleep, nosleep_mgr
from cthulhu.log import log
from cthulhu.manager.crush_node_request_factory import CrushNodeRequestFactory
from cthulhu.manager.crush_rule_request_factory import CrushRuleRequestFactory
from cthulhu.manager.crush_request_factory import CrushRequestFactory
from cthulhu.manager.osd_request_factory import OsdRequestFactory
from cthulhu.manager.pool_request_factory import PoolRequestFactory
from cthulhu.manager.plugin_monitor import PluginMonitor
from calamari_common.types import CRUSH_NODE, CRUSH_RULE, CRUSH_MAP, SYNC_OBJECT_STR_TYPE, SYNC_OBJECT_TYPES, OSD, POOL, OsdMap, MdsMap, MonMap, MonStatus
from cthulhu.util import now
remote = get_remote()
FAVORITE_TIMEOUT_FACTOR = int(config.get('cthulhu', 'favorite_timeout_factor'))
class ClusterUnavailable(Exception):
pass
class SyncObjects(object):
"""
A collection of versioned objects, keyed by their class (which
must be a SyncObject subclass).
The objects are immutable, so it is safe to hand out references: new
versions are new objects.
def __init__(self, manager):
self._manager = manager
self._remote = get_remote()
import logging
import uuid
from calamari_common.remote import get_remote
from cthulhu.log import log
from calamari_common.types import OsdMap, PgSummary, USER_REQUEST_COMPLETE, USER_REQUEST_SUBMITTED
from cthulhu.util import now
remote = get_remote()
class UserRequestBase(object):
"""
A request acts on one or more Ceph-managed objects, i.e.
mon, mds, osd, pg.
Amidst the terminology mess of 'jobs', 'commands', 'operations', this class
is named for clarity: it's an operation at an end-user level of
granularity, something that might be a button in the UI.
UserRequests are usually remotely executed on a mon. However, there
may be a final step of updating the state of ClusterMonitor in order
that subsequent REST API consumer reads return values consistent with
the job having completed, e.g. waiting for the OSD map to be up
to date before calling a pool creation complete. For this reason,
UserRequests have a local ID and completion state that is independent
of their remote ID (salt jid). UserRequests may also execute more than
one JID in the course of their lifetime.
Requests have the following lifecycle:
NEW object is created, it has all the information needed to do its job
def __init__(self, manager):
self._manager = manager
self._remote = get_remote()