#
# Copyright 2014-2015 eNovance
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
from collections import defaultdict
import hashlib
import itertools
import operator
import pkg_resources
import threading
import uuid
from gnocchiclient import exceptions as gnocchi_exc
from keystoneauth1 import exceptions as ka_exceptions
from oslo_log import log
from oslo_serialization import jsonutils
from oslo_utils import fnmatch
from oslo_utils import timeutils
import six
from stevedore import extension
import tenacity
from ceilometer import declarative
from ceilometer import dispatcher
from ceilometer.i18n import _LE, _LW
from ceilometer import gnocchi_client
from ceilometer import keystone_client
NAME_ENCODED = __name__.encode('utf-8')
CACHE_NAMESPACE = uuid.UUID(bytes=hashlib.md5(NAME_ENCODED).digest())
LOG = log.getLogger(__name__)
def cache_key_mangler(key):
"""Construct an opaque cache key."""
if six.PY2:
key = key.encode('utf-8')
return uuid.uuid5(CACHE_NAMESPACE, key).hex
EVENT_CREATE, EVENT_UPDATE, EVENT_DELETE = ("create", "update", "delete")
[docs]class ResourcesDefinition(object):
MANDATORY_FIELDS = {'resource_type': six.string_types,
'metrics': list}
MANDATORY_EVENT_FIELDS = {'id': six.string_types}
def __init__(self, definition_cfg, default_archive_policy, plugin_manager):
self._default_archive_policy = default_archive_policy
self.cfg = definition_cfg
self._check_required_and_types(self.MANDATORY_FIELDS, self.cfg)
if self.support_events():
self._check_required_and_types(self.MANDATORY_EVENT_FIELDS,
self.cfg['event_attributes'])
self._attributes = {}
for name, attr_cfg in self.cfg.get('attributes', {}).items():
self._attributes[name] = declarative.Definition(name, attr_cfg,
plugin_manager)
self._event_attributes = {}
for name, attr_cfg in self.cfg.get('event_attributes', {}).items():
self._event_attributes[name] = declarative.Definition(
name, attr_cfg, plugin_manager)
self.metrics = {}
for t in self.cfg['metrics']:
archive_policy = self.cfg.get('archive_policy',
self._default_archive_policy)
if archive_policy is None:
self.metrics[t] = {}
else:
self.metrics[t] = dict(archive_policy_name=archive_policy)
@staticmethod
def _check_required_and_types(expected, definition):
for field, field_type in expected.items():
if field not in definition:
raise declarative.ResourceDefinitionException(
_LE("Required field %s not specified") % field, definition)
if not isinstance(definition[field], field_type):
raise declarative.ResourceDefinitionException(
_LE("Required field %(field)s should be a %(type)s") %
{'field': field, 'type': field_type}, definition)
@staticmethod
def _ensure_list(value):
if isinstance(value, list):
return value
return [value]
[docs] def metric_match(self, metric_name):
for t in self.cfg['metrics']:
if fnmatch.fnmatch(metric_name, t):
return True
return False
[docs] def support_events(self):
for e in ["event_create", "event_delete", "event_update"]:
if e in self.cfg:
return True
return False
[docs] def event_match(self, event_type):
for e in self._ensure_list(self.cfg.get('event_create', [])):
if fnmatch.fnmatch(event_type, e):
return EVENT_CREATE
for e in self._ensure_list(self.cfg.get('event_delete', [])):
if fnmatch.fnmatch(event_type, e):
return EVENT_DELETE
for e in self._ensure_list(self.cfg.get('event_update', [])):
if fnmatch.fnmatch(event_type, e):
return EVENT_UPDATE
[docs] def sample_attributes(self, sample):
attrs = {}
for name, definition in self._attributes.items():
value = definition.parse(sample)
if value is not None:
attrs[name] = value
return attrs
[docs] def event_attributes(self, event):
attrs = {'type': self.cfg['resource_type']}
traits = dict([(trait[0], trait[2]) for trait in event['traits']])
for attr, field in self.cfg.get('event_attributes', {}).items():
value = traits.get(field)
if value is not None:
attrs[attr] = value
return attrs
[docs]class LockedDefaultDict(defaultdict):
"""defaultdict with lock to handle threading
Dictionary only deletes if nothing is accessing dict and nothing is holding
lock to be deleted. If both cases are not true, it will skip delete.
"""
def __init__(self, *args, **kwargs):
self.lock = threading.Lock()
super(LockedDefaultDict, self).__init__(*args, **kwargs)
def __getitem__(self, key):
with self.lock:
return super(LockedDefaultDict, self).__getitem__(key)
[docs] def pop(self, key, *args):
with self.lock:
key_lock = super(LockedDefaultDict, self).__getitem__(key)
if key_lock.acquire(False):
try:
super(LockedDefaultDict, self).pop(key, *args)
finally:
key_lock.release()
[docs]class GnocchiDispatcher(dispatcher.MeterDispatcherBase,
dispatcher.EventDispatcherBase):
"""Dispatcher class for recording metering data into database.
The dispatcher class records each meter into the gnocchi service
configured in ceilometer configuration file. An example configuration may
look like the following:
[dispatcher_gnocchi]
archive_policy = low
To enable this dispatcher, the following section needs to be present in
ceilometer.conf file
[DEFAULT]
meter_dispatchers = gnocchi
event_dispatchers = gnocchi
"""
def __init__(self, conf):
super(GnocchiDispatcher, self).__init__(conf)
self.conf = conf
self.filter_service_activity = (
conf.dispatcher_gnocchi.filter_service_activity)
self._ks_client = keystone_client.get_client(conf)
self.resources_definition = self._load_resources_definitions(conf)
self.cache = None
try:
import oslo_cache
oslo_cache.configure(self.conf)
# NOTE(cdent): The default cache backend is a real but
# noop backend. We don't want to use that here because
# we want to avoid the cache pathways entirely if the
# cache has not been configured explicitly.
if self.conf.cache.enabled:
cache_region = oslo_cache.create_region()
self.cache = oslo_cache.configure_cache_region(
self.conf, cache_region)
self.cache.key_mangler = cache_key_mangler
except ImportError:
pass
except oslo_cache.exception.ConfigurationError as exc:
LOG.warning(_LW('unable to configure oslo_cache: %s') % exc)
self._gnocchi_project_id = None
self._gnocchi_project_id_lock = threading.Lock()
self._gnocchi_resource_lock = LockedDefaultDict(threading.Lock)
self._gnocchi = gnocchi_client.get_gnocchiclient(conf)
retries = conf.storage.max_retries
@tenacity.retry(
wait=tenacity.wait_fixed(conf.storage.retry_interval),
stop=(tenacity.stop_after_attempt(retries) if retries >= 0
else tenacity.stop_never),
reraise=True)
def _get_connection():
self._gnocchi.capabilities.list()
try:
_get_connection()
except Exception:
LOG.error(_LE('Failed to connect to Gnocchi.'))
raise
@classmethod
def _load_resources_definitions(cls, conf):
plugin_manager = extension.ExtensionManager(
namespace='ceilometer.event.trait_plugin')
data = declarative.load_definitions(
conf, {}, conf.dispatcher_gnocchi.resources_definition_file,
pkg_resources.resource_filename(__name__,
"data/gnocchi_resources.yaml"))
resource_defs = []
for resource in data.get('resources', []):
try:
resource_defs.append(ResourcesDefinition(
resource,
conf.dispatcher_gnocchi.archive_policy, plugin_manager))
except Exception as exc:
LOG.error(_LE("Failed to load resource due to error %s") %
exc)
return resource_defs
@property
def gnocchi_project_id(self):
if self._gnocchi_project_id is not None:
return self._gnocchi_project_id
with self._gnocchi_project_id_lock:
if self._gnocchi_project_id is None:
try:
project = self._ks_client.projects.find(
name=self.conf.dispatcher_gnocchi.filter_project)
except ka_exceptions.NotFound:
LOG.warning(_LW('gnocchi project not found in keystone,'
' ignoring the filter_service_activity '
'option'))
self.filter_service_activity = False
return None
except Exception:
LOG.exception('fail to retrieve user of Gnocchi service')
raise
self._gnocchi_project_id = project.id
LOG.debug("gnocchi project found: %s", self.gnocchi_project_id)
return self._gnocchi_project_id
def _is_swift_account_sample(self, sample):
return bool([rd for rd in self.resources_definition
if rd.cfg['resource_type'] == 'swift_account'
and rd.metric_match(sample['counter_name'])])
def _is_gnocchi_activity(self, sample):
return (self.filter_service_activity and self.gnocchi_project_id and (
# avoid anything from the user used by gnocchi
sample['project_id'] == self.gnocchi_project_id or
# avoid anything in the swift account used by gnocchi
(sample['resource_id'] == self.gnocchi_project_id and
self._is_swift_account_sample(sample))
))
def _get_resource_definition_from_metric(self, metric_name):
for rd in self.resources_definition:
if rd.metric_match(metric_name):
return rd
def _get_resource_definition_from_event(self, event_type):
for rd in self.resources_definition:
operation = rd.event_match(event_type)
if operation:
return rd, operation
[docs] def record_metering_data(self, data):
# We may have receive only one counter on the wire
if not isinstance(data, list):
data = [data]
# NOTE(sileht): skip sample generated by gnocchi itself
data = [s for s in data if not self._is_gnocchi_activity(s)]
# FIXME(sileht): This method bulk the processing of samples
# grouped by resource_id and metric_name but this is not
# efficient yet because the data received here doesn't often
# contains a lot of different kind of samples
# So perhaps the next step will be to pool the received data from
# message bus.
data.sort(key=lambda s: (s['resource_id'], s['counter_name']))
resource_grouped_samples = itertools.groupby(
data, key=operator.itemgetter('resource_id'))
gnocchi_data = {}
measures = {}
stats = dict(measures=0, resources=0, metrics=0)
for resource_id, samples_of_resource in resource_grouped_samples:
# NOTE(sileht): / is forbidden by Gnocchi
resource_id = resource_id.replace('/', '_')
stats['resources'] += 1
metric_grouped_samples = itertools.groupby(
list(samples_of_resource),
key=operator.itemgetter('counter_name'))
res_info = {}
for metric_name, samples in metric_grouped_samples:
stats['metrics'] += 1
samples = list(samples)
rd = self._get_resource_definition_from_metric(metric_name)
if rd is None:
LOG.warning(_LW("metric %s is not handled by Gnocchi") %
metric_name)
continue
if rd.cfg.get("ignore"):
continue
res_info['resource_type'] = rd.cfg['resource_type']
res_info.setdefault("resource", {}).update({
"id": resource_id,
"user_id": samples[0]['user_id'],
"project_id": samples[0]['project_id'],
"metrics": rd.metrics,
})
for sample in samples:
res_info.setdefault("resource_extra", {}).update(
rd.sample_attributes(sample))
m = measures.setdefault(resource_id, {}).setdefault(
metric_name, [])
m.append({'timestamp': sample['timestamp'],
'value': sample['counter_volume']})
unit = sample['counter_unit']
metric = sample['counter_name']
res_info['resource']['metrics'][metric]['unit'] = unit
stats['measures'] += len(measures[resource_id][metric_name])
res_info["resource"].update(res_info["resource_extra"])
if res_info:
gnocchi_data[resource_id] = res_info
try:
self.batch_measures(measures, gnocchi_data, stats)
except (gnocchi_exc.ClientException,
ka_exceptions.ConnectFailure) as e:
LOG.error(six.text_type(e))
except Exception as e:
LOG.error(six.text_type(e), exc_info=True)
for info in gnocchi_data.values():
resource = info["resource"]
resource_type = info["resource_type"]
resource_extra = info["resource_extra"]
if not resource_extra:
continue
try:
self._if_not_cached("update", resource_type, resource,
self._update_resource, resource_extra)
except gnocchi_exc.ClientException as e:
LOG.error(six.text_type(e))
except Exception as e:
LOG.error(six.text_type(e), exc_info=True)
@staticmethod
def _extract_resources_from_error(e, resource_infos):
resource_ids = set([r['original_resource_id']
for r in e.message['detail']])
return [(resource_infos[rid]['resource_type'],
resource_infos[rid]['resource'])
for rid in resource_ids]
[docs] def batch_measures(self, measures, resource_infos, stats):
# NOTE(sileht): We don't care about error here, we want
# resources metadata always been updated
try:
self._gnocchi.metric.batch_resources_metrics_measures(
measures, create_metrics=True)
except gnocchi_exc.BadRequest as e:
if not isinstance(e.message, dict):
raise
if e.message.get('cause') != 'Unknown resources':
raise
resources = self._extract_resources_from_error(e, resource_infos)
for resource_type, resource in resources:
try:
self._if_not_cached("create", resource_type, resource,
self._create_resource)
except gnocchi_exc.ResourceAlreadyExists:
# NOTE(sileht): resource created in the meantime
pass
except gnocchi_exc.ClientException as e:
LOG.error(_LE('Error creating resource %(id)s: %(err)s'),
{'id': resource['id'], 'err': six.text_type(e)})
# We cannot post measures for this resource
# and we can't patch it later
del measures[resource['id']]
del resource_infos[resource['id']]
# NOTE(sileht): we have created missing resources/metrics,
# now retry to post measures
self._gnocchi.metric.batch_resources_metrics_measures(
measures, create_metrics=True)
# FIXME(sileht): take care of measures removed in stats
LOG.debug("%(measures)d measures posted against %(metrics)d "
"metrics through %(resources)d resources", stats)
def _create_resource(self, resource_type, resource):
self._gnocchi.resource.create(resource_type, resource)
LOG.debug('Resource %s created', resource["id"])
def _update_resource(self, resource_type, resource, resource_extra):
self._gnocchi.resource.update(resource_type,
resource["id"],
resource_extra)
LOG.debug('Resource %s updated', resource["id"])
def _if_not_cached(self, operation, resource_type, resource, method,
*args, **kwargs):
if self.cache:
cache_key = resource['id']
attribute_hash = self._check_resource_cache(cache_key, resource)
hit = False
if attribute_hash:
with self._gnocchi_resource_lock[cache_key]:
# NOTE(luogangyi): there is a possibility that the
# resource was already built in cache by another
# ceilometer-collector when we get the lock here.
attribute_hash = self._check_resource_cache(cache_key,
resource)
if attribute_hash:
method(resource_type, resource, *args, **kwargs)
self.cache.set(cache_key, attribute_hash)
else:
hit = True
LOG.debug('resource cache recheck hit for '
'%s %s', operation, cache_key)
self._gnocchi_resource_lock.pop(cache_key, None)
else:
hit = True
LOG.debug('Resource cache hit for %s %s', operation, cache_key)
if hit and operation == "create":
raise gnocchi_exc.ResourceAlreadyExists()
else:
method(resource_type, resource, *args, **kwargs)
def _check_resource_cache(self, key, resource_data):
cached_hash = self.cache.get(key)
attribute_hash = hash(frozenset(filter(lambda x: x[0] != "metrics",
resource_data.items())))
if not cached_hash or cached_hash != attribute_hash:
return attribute_hash
else:
return None
[docs] def record_events(self, events):
for event in events:
rd = self._get_resource_definition_from_event(event['event_type'])
if not rd:
LOG.debug("No gnocchi definition for event type: %s",
event['event_type'])
continue
rd, operation = rd
if operation == EVENT_DELETE:
self._delete_event(rd, event)
def _delete_event(self, rd, event):
ended_at = timeutils.utcnow().isoformat()
resource = rd.event_attributes(event)
associated_resources = rd.cfg.get('event_associated_resources', {})
if associated_resources:
to_end = itertools.chain([resource], *[
self._search_resource(resource_type, query % resource['id'])
for resource_type, query in associated_resources.items()
])
else:
to_end = [resource]
for resource in to_end:
self._set_ended_at(resource, ended_at)
def _search_resource(self, resource_type, query):
try:
return self._gnocchi.resource.search(
resource_type, jsonutils.loads(query))
except Exception:
LOG.error(_LE("Fail to search resource type %{resource_type}s "
"with '%{query}s'"),
{'resource_type': resource_type, 'query': query},
exc_info=True)
return []
def _set_ended_at(self, resource, ended_at):
try:
self._gnocchi.resource.update(resource['type'], resource['id'],
{'ended_at': ended_at})
except gnocchi_exc.ResourceNotFound:
LOG.debug("Delete event received on unexisting resource (%s), "
"ignore it.", resource['id'])
except Exception:
LOG.error(_LE("Fail to update the resource %s"), resource,
exc_info=True)
LOG.debug('Resource %s ended at %s' % (resource["id"], ended_at))