deb-ceilometer/ceilometer/pipeline.py
gordon chung e38ffd09e8 distributed coordinated notifications
this patch enables a queue per pipeline per agent. each agent can
send data to pipeline queue set of any agent but only listens to it's
own pipeline queue. when receiving a notification, the agent will:
1. build datapoint
2. calculate hash key based on (hardcoded) attribute and mod hash
   by number of active agents
3. sends the data to single agents pipeline queue set.

ex. two agents, sample1 with res_id=1 and sample2 with res_id=2
1. agent1 builds both samples
2. hash(sample1) % 2 (num agents) == agent1 sends sample to agent2
   pipeline queue.
2a. agent2 process sample1
3. hash(sample2) % 2 (num agents) == agent1 sends sample to agent1
   pipeline queue.
3a. agent1 process sample2

Implements blueprint distributed-coordinated-notifications

Change-Id: Iab52cae0a5bfbc747a2918e67dfe8da5fd0fda84
2015-07-24 17:04:49 -04:00

834 lines
28 KiB
Python

#
# Copyright 2013 Intel Corp.
# Copyright 2014 Red Hat, Inc
#
# Authors: Yunhong Jiang <yunhong.jiang@intel.com>
# Eoghan Glynn <eglynn@redhat.com>
#
# 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.
import abc
import fnmatch
import hashlib
import os
from oslo_config import cfg
from oslo_log import log
from oslo_utils import timeutils
import six
from stevedore import extension
import yaml
from ceilometer.event.storage import models
from ceilometer.i18n import _
from ceilometer import publisher
from ceilometer.publisher import utils as publisher_utils
from ceilometer import sample as sample_util
OPTS = [
cfg.StrOpt('pipeline_cfg_file',
default="pipeline.yaml",
help="Configuration file for pipeline definition."
),
cfg.StrOpt('event_pipeline_cfg_file',
default="event_pipeline.yaml",
help="Configuration file for event pipeline definition."
),
cfg.BoolOpt('refresh_pipeline_cfg',
default=False,
help="Refresh Pipeline configuration on-the-fly."
),
cfg.IntOpt('pipeline_polling_interval',
default=20,
help="Polling interval for pipeline file configuration"
" in seconds."
),
]
cfg.CONF.register_opts(OPTS)
LOG = log.getLogger(__name__)
class PipelineException(Exception):
def __init__(self, message, pipeline_cfg):
self.msg = message
self.pipeline_cfg = pipeline_cfg
def __str__(self):
return 'Pipeline %s: %s' % (self.pipeline_cfg, self.msg)
@six.add_metaclass(abc.ABCMeta)
class PipelineEndpoint(object):
def __init__(self, context, pipeline):
self.publish_context = PublishContext(context, [pipeline])
@abc.abstractmethod
def sample(self, ctxt, publisher_id, event_type, payload, metadata):
pass
class SamplePipelineEndpoint(PipelineEndpoint):
def sample(self, ctxt, publisher_id, event_type, payload, metadata):
samples = [
sample_util.Sample(name=s['counter_name'],
type=s['counter_type'],
unit=s['counter_unit'],
volume=s['counter_volume'],
user_id=s['user_id'],
project_id=s['project_id'],
resource_id=s['resource_id'],
timestamp=s['timestamp'],
resource_metadata=s['resource_metadata'],
source=s.get('source'))
for s in payload if publisher_utils.verify_signature(
s, cfg.CONF.publisher.telemetry_secret)
]
with self.publish_context as p:
p(samples)
class EventPipelineEndpoint(PipelineEndpoint):
def sample(self, ctxt, publisher_id, event_type, payload, metadata):
events = [
models.Event(
message_id=ev['message_id'],
event_type=ev['event_type'],
generated=timeutils.normalize_time(
timeutils.parse_isotime(ev['generated'])),
traits=[models.Trait(name, dtype,
models.Trait.convert_value(dtype, value))
for name, dtype, value in ev['traits']],
raw=ev.get('raw', {}))
for ev in payload if publisher_utils.verify_signature(
ev, cfg.CONF.publisher.telemetry_secret)
]
with self.publish_context as p:
p(events)
class _PipelineTransportManager(object):
def __init__(self):
self.transporters = []
def add_transporter(self, transporter):
self.transporters.append(transporter)
def publisher(self, context):
serializer = self.serializer
hash_to_bucketise = self.hash_to_bucketise
transporters = self.transporters
filter_attr = self.filter_attr
event_type = self.event_type
class PipelinePublishContext(object):
def __enter__(self):
def p(data):
# TODO(gordc): cleanup so payload is always single
# datapoint. we can't correctly bucketise
# datapoints if batched.
data = [data] if not isinstance(data, list) else data
for datapoint in data:
serialized_data = serializer(datapoint)
for d_filter, notifiers in transporters:
if d_filter(serialized_data[filter_attr]):
key = (hash_to_bucketise(serialized_data) %
len(notifiers))
notifier = notifiers[key]
notifier.sample(context.to_dict(),
event_type=event_type,
payload=[serialized_data])
return p
def __exit__(self, exc_type, exc_value, traceback):
pass
return PipelinePublishContext()
class SamplePipelineTransportManager(_PipelineTransportManager):
filter_attr = 'counter_name'
event_type = 'ceilometer.pipeline'
@staticmethod
def hash_to_bucketise(datapoint):
return hash(datapoint['resource_id'])
@staticmethod
def serializer(data):
return publisher_utils.meter_message_from_counter(
data, cfg.CONF.publisher.telemetry_secret)
class EventPipelineTransportManager(_PipelineTransportManager):
filter_attr = 'event_type'
event_type = 'pipeline.event'
@staticmethod
def hash_to_bucketise(datapoint):
return hash(datapoint['event_type'])
@staticmethod
def serializer(data):
return publisher_utils.message_from_event(
data, cfg.CONF.publisher.telemetry_secret)
class PublishContext(object):
def __init__(self, context, pipelines=None):
pipelines = pipelines or []
self.pipelines = set(pipelines)
self.context = context
def add_pipelines(self, pipelines):
self.pipelines.update(pipelines)
def __enter__(self):
def p(data):
for p in self.pipelines:
p.publish_data(self.context, data)
return p
def __exit__(self, exc_type, exc_value, traceback):
for p in self.pipelines:
p.flush(self.context)
class Source(object):
"""Represents a source of samples or events."""
def __init__(self, cfg):
self.cfg = cfg
try:
self.name = cfg['name']
self.sinks = cfg.get('sinks')
except KeyError as err:
raise PipelineException(
"Required field %s not specified" % err.args[0], cfg)
def __str__(self):
return self.name
def check_sinks(self, sinks):
if not self.sinks:
raise PipelineException(
"No sink defined in source %s" % self,
self.cfg)
for sink in self.sinks:
if sink not in sinks:
raise PipelineException(
"Dangling sink %s from source %s" % (sink, self),
self.cfg)
def check_source_filtering(self, data, d_type):
"""Source data rules checking
- At least one meaningful datapoint exist
- Included type and excluded type can't co-exist on the same pipeline
- Included type meter and wildcard can't co-exist at same pipeline
"""
if not data:
raise PipelineException('No %s specified' % d_type, self.cfg)
if ([x for x in data if x[0] not in '!*'] and
[x for x in data if x[0] == '!']):
raise PipelineException(
'Both included and excluded %s specified' % d_type,
cfg)
if '*' in data and [x for x in data if x[0] not in '!*']:
raise PipelineException(
'Included %s specified with wildcard' % d_type,
self.cfg)
@staticmethod
def is_supported(dataset, data_name):
# Support wildcard like storage.* and !disk.*
# Start with negation, we consider that the order is deny, allow
if any(fnmatch.fnmatch(data_name, datapoint[1:])
for datapoint in dataset if datapoint[0] == '!'):
return False
if any(fnmatch.fnmatch(data_name, datapoint)
for datapoint in dataset if datapoint[0] != '!'):
return True
# if we only have negation, we suppose the default is allow
return all(datapoint.startswith('!') for datapoint in dataset)
class EventSource(Source):
"""Represents a source of events.
In effect it is a set of notification handlers capturing events for a set
of matching notifications.
"""
def __init__(self, cfg):
super(EventSource, self).__init__(cfg)
try:
self.events = cfg['events']
except KeyError as err:
raise PipelineException(
"Required field %s not specified" % err.args[0], cfg)
self.check_source_filtering(self.events, 'events')
def support_event(self, event_name):
return self.is_supported(self.events, event_name)
class SampleSource(Source):
"""Represents a source of samples.
In effect it is a set of pollsters and/or notification handlers emitting
samples for a set of matching meters. Each source encapsulates meter name
matching, polling interval determination, optional resource enumeration or
discovery, and mapping to one or more sinks for publication.
"""
def __init__(self, cfg):
super(SampleSource, self).__init__(cfg)
try:
# Support 'counters' for backward compatibility
self.meters = cfg.get('meters', cfg.get('counters'))
except KeyError as err:
raise PipelineException(
"Required field %s not specified" % err.args[0], cfg)
try:
self.interval = int(cfg.get('interval', 600))
except ValueError:
raise PipelineException("Invalid interval value", cfg)
if self.interval <= 0:
raise PipelineException("Interval value should > 0", cfg)
self.resources = cfg.get('resources') or []
if not isinstance(self.resources, list):
raise PipelineException("Resources should be a list", cfg)
self.discovery = cfg.get('discovery') or []
if not isinstance(self.discovery, list):
raise PipelineException("Discovery should be a list", cfg)
self.check_source_filtering(self.meters, 'meters')
def get_interval(self):
return self.interval
# (yjiang5) To support meters like instance:m1.tiny,
# which include variable part at the end starting with ':'.
# Hope we will not add such meters in future.
@staticmethod
def _variable_meter_name(name):
m = name.partition(':')
if m[1] == ':':
return m[1].join((m[0], '*'))
else:
return name
def support_meter(self, meter_name):
meter_name = self._variable_meter_name(meter_name)
return self.is_supported(self.meters, meter_name)
class Sink(object):
"""Represents a sink for the transformation and publication of data.
Each sink config is concerned *only* with the transformation rules
and publication conduits for data.
In effect, a sink describes a chain of handlers. The chain starts
with zero or more transformers and ends with one or more publishers.
The first transformer in the chain is passed data from the
corresponding source, takes some action such as deriving rate of
change, performing unit conversion, or aggregating, before passing
the modified data to next step.
The subsequent transformers, if any, handle the data similarly.
At the end of the chain, publishers publish the data. The exact
publishing method depends on publisher type, for example, pushing
into data storage via the message bus providing guaranteed delivery,
or for loss-tolerant data UDP may be used.
If no transformers are included in the chain, the publishers are
passed data directly from the sink which are published unchanged.
"""
def __init__(self, cfg, transformer_manager):
self.cfg = cfg
try:
self.name = cfg['name']
# It's legal to have no transformer specified
self.transformer_cfg = cfg.get('transformers') or []
except KeyError as err:
raise PipelineException(
"Required field %s not specified" % err.args[0], cfg)
if not cfg.get('publishers'):
raise PipelineException("No publisher specified", cfg)
self.publishers = []
for p in cfg['publishers']:
if '://' not in p:
# Support old format without URL
p = p + "://"
try:
self.publishers.append(publisher.get_publisher(p,
self.NAMESPACE))
except Exception:
LOG.exception(_("Unable to load publisher %s"), p)
self.transformers = self._setup_transformers(cfg, transformer_manager)
def __str__(self):
return self.name
def _setup_transformers(self, cfg, transformer_manager):
transformers = []
for transformer in self.transformer_cfg:
parameter = transformer['parameters'] or {}
try:
ext = transformer_manager[transformer['name']]
except KeyError:
raise PipelineException(
"No transformer named %s loaded" % transformer['name'],
cfg)
transformers.append(ext.plugin(**parameter))
LOG.info(_(
"Pipeline %(pipeline)s: Setup transformer instance %(name)s "
"with parameter %(param)s") % ({'pipeline': self,
'name': transformer['name'],
'param': parameter}))
return transformers
class EventSink(Sink):
NAMESPACE = 'ceilometer.event.publisher'
def publish_events(self, ctxt, events):
if events:
for p in self.publishers:
try:
p.publish_events(ctxt, events)
except Exception:
LOG.exception(_(
"Pipeline %(pipeline)s: Continue after error "
"from publisher %(pub)s") % ({'pipeline': self,
'pub': p}))
def flush(self, ctxt):
"""Flush data after all events have been injected to pipeline."""
pass
class SampleSink(Sink):
NAMESPACE = 'ceilometer.publisher'
def _transform_sample(self, start, ctxt, sample):
try:
for transformer in self.transformers[start:]:
sample = transformer.handle_sample(ctxt, sample)
if not sample:
LOG.debug(_(
"Pipeline %(pipeline)s: Sample dropped by "
"transformer %(trans)s") % ({'pipeline': self,
'trans': transformer}))
return
return sample
except Exception as err:
# TODO(gordc): only use one log level.
LOG.warning(_("Pipeline %(pipeline)s: "
"Exit after error from transformer "
"%(trans)s for %(smp)s") % ({'pipeline': self,
'trans': transformer,
'smp': sample}))
LOG.exception(err)
def _publish_samples(self, start, ctxt, samples):
"""Push samples into pipeline for publishing.
:param start: The first transformer that the sample will be injected.
This is mainly for flush() invocation that transformer
may emit samples.
:param ctxt: Execution context from the manager or service.
:param samples: Sample list.
"""
transformed_samples = []
if not self.transformers:
transformed_samples = samples
else:
for sample in samples:
LOG.debug(_(
"Pipeline %(pipeline)s: Transform sample "
"%(smp)s from %(trans)s transformer") % ({'pipeline': self,
'smp': sample,
'trans': start}))
sample = self._transform_sample(start, ctxt, sample)
if sample:
transformed_samples.append(sample)
if transformed_samples:
for p in self.publishers:
try:
p.publish_samples(ctxt, transformed_samples)
except Exception:
LOG.exception(_(
"Pipeline %(pipeline)s: Continue after error "
"from publisher %(pub)s") % ({'pipeline': self,
'pub': p}))
def publish_samples(self, ctxt, samples):
self._publish_samples(0, ctxt, samples)
def flush(self, ctxt):
"""Flush data after all samples have been injected to pipeline."""
for (i, transformer) in enumerate(self.transformers):
try:
self._publish_samples(i + 1, ctxt,
list(transformer.flush(ctxt)))
except Exception as err:
LOG.warning(_(
"Pipeline %(pipeline)s: Error flushing "
"transformer %(trans)s") % ({'pipeline': self,
'trans': transformer}))
LOG.exception(err)
@six.add_metaclass(abc.ABCMeta)
class Pipeline(object):
"""Represents a coupling between a sink and a corresponding source."""
def __init__(self, source, sink):
self.source = source
self.sink = sink
self.name = str(self)
def __str__(self):
return (self.source.name if self.source.name == self.sink.name
else '%s:%s' % (self.source.name, self.sink.name))
def flush(self, ctxt):
self.sink.flush(ctxt)
@property
def publishers(self):
return self.sink.publishers
@abc.abstractmethod
def publish_data(self, ctxt, data):
"""Publish data from pipeline."""
class EventPipeline(Pipeline):
"""Represents a pipeline for Events."""
def __str__(self):
# NOTE(gordc): prepend a namespace so we ensure event and sample
# pipelines do not have the same name.
return 'event:%s' % super(EventPipeline, self).__str__()
def support_event(self, event_type):
return self.source.support_event(event_type)
def publish_data(self, ctxt, events):
if not isinstance(events, list):
events = [events]
supported = [e for e in events
if self.source.support_event(e.event_type)]
self.sink.publish_events(ctxt, supported)
class SamplePipeline(Pipeline):
"""Represents a pipeline for Samples."""
def get_interval(self):
return self.source.interval
@property
def resources(self):
return self.source.resources
@property
def discovery(self):
return self.source.discovery
def support_meter(self, meter_name):
return self.source.support_meter(meter_name)
def publish_data(self, ctxt, samples):
if not isinstance(samples, list):
samples = [samples]
supported = [s for s in samples if self.source.support_meter(s.name)]
self.sink.publish_samples(ctxt, supported)
SAMPLE_TYPE = {'pipeline': SamplePipeline,
'source': SampleSource,
'sink': SampleSink}
EVENT_TYPE = {'pipeline': EventPipeline,
'source': EventSource,
'sink': EventSink}
class PipelineManager(object):
"""Pipeline Manager
Pipeline manager sets up pipelines according to config file
Usually only one pipeline manager exists in the system.
"""
def __init__(self, cfg, transformer_manager, p_type=SAMPLE_TYPE):
"""Setup the pipelines according to config.
The configuration is supported as follows:
Decoupled: the source and sink configuration are separately
specified before being linked together. This allows source-
specific configuration, such as resource discovery, to be
kept focused only on the fine-grained source while avoiding
the necessity for wide duplication of sink-related config.
The configuration is provided in the form of separate lists
of dictionaries defining sources and sinks, for example:
{"sources": [{"name": source_1,
"interval": interval_time,
"meters" : ["meter_1", "meter_2"],
"resources": ["resource_uri1", "resource_uri2"],
"sinks" : ["sink_1", "sink_2"]
},
{"name": source_2,
"interval": interval_time,
"meters" : ["meter_3"],
"sinks" : ["sink_2"]
},
],
"sinks": [{"name": sink_1,
"transformers": [
{"name": "Transformer_1",
"parameters": {"p1": "value"}},
{"name": "Transformer_2",
"parameters": {"p1": "value"}},
],
"publishers": ["publisher_1", "publisher_2"]
},
{"name": sink_2,
"publishers": ["publisher_3"]
},
]
}
The interval determines the cadence of sample injection into
the pipeline where samples are produced under the direct control
of an agent, i.e. via a polling cycle as opposed to incoming
notifications.
Valid meter format is '*', '!meter_name', or 'meter_name'.
'*' is wildcard symbol means any meters; '!meter_name' means
"meter_name" will be excluded; 'meter_name' means 'meter_name'
will be included.
The 'meter_name" is Sample name field. For meter names with
variable like "instance:m1.tiny", it's "instance:*".
Valid meters definition is all "included meter names", all
"excluded meter names", wildcard and "excluded meter names", or
only wildcard.
The resources is list of URI indicating the resources from where
the meters should be polled. It's optional and it's up to the
specific pollster to decide how to use it.
Transformer's name is plugin name in setup.cfg.
Publisher's name is plugin name in setup.cfg
"""
self.pipelines = []
if not ('sources' in cfg and 'sinks' in cfg):
raise PipelineException("Both sources & sinks are required",
cfg)
LOG.info(_('detected decoupled pipeline config format'))
unique_names = set()
sources = []
for s in cfg.get('sources', []):
name = s.get('name')
if name in unique_names:
raise PipelineException("Duplicated source names: %s" %
name, self)
else:
unique_names.add(name)
sources.append(p_type['source'](s))
unique_names.clear()
sinks = {}
for s in cfg.get('sinks', []):
name = s.get('name')
if name in unique_names:
raise PipelineException("Duplicated sink names: %s" %
name, self)
else:
unique_names.add(name)
sinks[s['name']] = p_type['sink'](s, transformer_manager)
unique_names.clear()
for source in sources:
source.check_sinks(sinks)
for target in source.sinks:
pipe = p_type['pipeline'](source, sinks[target])
if pipe.name in unique_names:
raise PipelineException(
"Duplicate pipeline name: %s. Ensure pipeline"
" names are unique. (name is the source and sink"
" names combined)" % pipe.name, cfg)
else:
unique_names.add(pipe.name)
self.pipelines.append(pipe)
unique_names.clear()
def publisher(self, context):
"""Build a new Publisher for these manager pipelines.
:param context: The context.
"""
return PublishContext(context, self.pipelines)
class PollingManager(object):
"""Polling Manager
Polling manager sets up polling according to config file.
"""
def __init__(self, cfg):
"""Setup the polling according to config.
The configuration is the sources half of the Pipeline Config.
"""
self.sources = []
if not ('sources' in cfg and 'sinks' in cfg):
raise PipelineException("Both sources & sinks are required",
cfg)
LOG.info(_('detected decoupled pipeline config format'))
unique_names = set()
for s in cfg.get('sources', []):
name = s.get('name')
if name in unique_names:
raise PipelineException("Duplicated source names: %s" %
name, self)
else:
unique_names.add(name)
self.sources.append(SampleSource(s))
unique_names.clear()
def _setup_pipeline_manager(cfg_file, transformer_manager, p_type=SAMPLE_TYPE):
if not os.path.exists(cfg_file):
cfg_file = cfg.CONF.find_file(cfg_file)
LOG.debug(_("Pipeline config file: %s"), cfg_file)
with open(cfg_file) as fap:
data = fap.read()
pipeline_cfg = yaml.safe_load(data)
LOG.info(_("Pipeline config: %s"), pipeline_cfg)
return PipelineManager(pipeline_cfg,
transformer_manager or
extension.ExtensionManager(
'ceilometer.transformer',
), p_type)
def _setup_polling_manager(cfg_file):
if not os.path.exists(cfg_file):
cfg_file = cfg.CONF.find_file(cfg_file)
LOG.debug(_("Polling config file: %s"), cfg_file)
with open(cfg_file) as fap:
data = fap.read()
pipeline_cfg = yaml.safe_load(data)
LOG.info(_("Pipeline config: %s"), pipeline_cfg)
return PollingManager(pipeline_cfg)
def setup_event_pipeline(transformer_manager=None):
"""Setup event pipeline manager according to yaml config file."""
cfg_file = cfg.CONF.event_pipeline_cfg_file
return _setup_pipeline_manager(cfg_file, transformer_manager, EVENT_TYPE)
def setup_pipeline(transformer_manager=None):
"""Setup pipeline manager according to yaml config file."""
cfg_file = cfg.CONF.pipeline_cfg_file
return _setup_pipeline_manager(cfg_file, transformer_manager)
def _get_pipeline_cfg_file(p_type=SAMPLE_TYPE):
if p_type == EVENT_TYPE:
cfg_file = cfg.CONF.event_pipeline_cfg_file
else:
cfg_file = cfg.CONF.pipeline_cfg_file
if not os.path.exists(cfg_file):
cfg_file = cfg.CONF.find_file(cfg_file)
return cfg_file
def get_pipeline_mtime(p_type=SAMPLE_TYPE):
cfg_file = _get_pipeline_cfg_file(p_type)
return os.path.getmtime(cfg_file)
def get_pipeline_hash(p_type=SAMPLE_TYPE):
cfg_file = _get_pipeline_cfg_file(p_type)
with open(cfg_file) as fap:
data = fap.read()
if six.PY3:
data = data.encode('utf-8')
file_hash = hashlib.md5(data).hexdigest()
return file_hash
def setup_polling():
"""Setup polling manager according to yaml config file."""
cfg_file = cfg.CONF.pipeline_cfg_file
return _setup_polling_manager(cfg_file)