class RiemannHandler(Handler):
def __init__(self, config=None):
# Initialize Handler
Handler.__init__(self, config)
if riemann_client is None:
logging.error("Failed to load riemann_client module")
return
# Initialize options
self.host = self.config['host']
self.port = int(self.config['port'])
self.transport = self.config['transport']
# Initialize client
if self.transport == 'tcp':
self.transport = TCPTransport(self.host, self.port)
else:
self.transport = UDPTransport(self.host, self.port)
self.client = Client(self.transport)
self._connect()
def get_default_config_help(self):
"""
Returns the help text for the configuration options for this handler
"""
config = super(RiemannHandler, self).get_default_config_help()
config.update({
'host': '',
'port': '',
'transport': 'tcp or udp',
})
return config
def get_default_config(self):
"""
Return the default config for the handler
"""
config = super(RiemannHandler, self).get_default_config()
config.update({
'host': '',
'port': 123,
'transport': 'tcp',
})
return config
def process(self, metric):
"""
Send a metric to Riemann.
"""
event = self._metric_to_riemann_event(metric)
try:
self.client.send_event(event)
except Exception, e:
self.log.error(
"RiemannHandler: Error sending event to Riemann: %s", e)
class RiemannHandler(Handler):
def __init__(self, config=None):
# Initialize Handler
Handler.__init__(self, config)
if riemann_client is None:
logging.error("Failed to load riemann_client module")
return
# Initialize options
self.host = self.config['host']
self.port = int(self.config['port'])
self.transport = self.config['transport']
# Initialize client
if self.transport == 'tcp':
self.transport = TCPTransport(self.host, self.port)
else:
self.transport = UDPTransport(self.host, self.port)
self.client = Client(self.transport)
self._connect()
def get_default_config_help(self):
"""
Returns the help text for the configuration options for this handler
"""
config = super(RiemannHandler, self).get_default_config_help()
config.update({
'host': '',
'port': '',
'transport': 'tcp or udp',
})
return config
def get_default_config(self):
"""
Return the default config for the handler
"""
config = super(RiemannHandler, self).get_default_config()
config.update({
'host': '',
'port': 123,
'transport': 'tcp',
})
return config
def process(self, metric):
"""
Send a metric to Riemann.
"""
event = self._metric_to_riemann_event(metric)
try:
self.client.send_event(event)
except Exception as e:
self.log.error(
"RiemannHandler: Error sending event to Riemann: %s", e)
def _metric_to_riemann_event(self, metric):
"""
Convert a metric to a dictionary representing a Riemann event.
"""
# Riemann has a separate "host" field, so remove from the path.
path = '%s.%s.%s' % (
metric.getPathPrefix(),
metric.getCollectorPath(),
metric.getMetricPath()
)
return self.client.create_event({
'host': metric.host,
'service': path,
'time': metric.timestamp,
'metric_f': float(metric.value),
'ttl': metric.ttl,
})
def _connect(self):
self.transport.connect()
def _close(self):
"""
Disconnect from Riemann.
"""
if hasattr(self, 'transport'):
self.transport.disconnect()
def __del__(self):
self._close()
class RiemannClient(MetricsStoreClientBase):
"""Riemann client"""
def __init__(self, store_endpoint, datacenter, default_metric_host,
default_metric_ttl, default_metric_state):
"""
:param store_endpoint: Riemann server endpoint
:type store_endpoint: str
:param datacenter: The current datacenter name
:type datacenter: str
:param default_metric_host: The default host value used when no other
host has been provided from the metric resource_metadata.
:type default_metric_host: str
:param default_metric_ttl: The amount of time in seconds that events
sent to Riemann are retained in its index.
:type default_metric_ttl: int
:param default_metric_state: The default state value used when no other
state has been provided from the metric resource_metadata.
:type default_metric_state: str
"""
super(RiemannClient, self).__init__()
self._store_endpoint = store_endpoint
self.store_uri = urlparse(self.store_endpoint)
self.default_metric_host = default_metric_host
self.default_metric_ttl = default_metric_ttl
self.default_metric_state = default_metric_state
self.datacenter = datacenter
self._transport = self._create_transport(self.store_uri.scheme)
self.client = Client(transport=self._transport, )
def _create_transport(self, scheme):
transport_types = dict(
tcp=transport.TCPTransport,
udp=transport.UDPTransport,
)
_transport_cls = transport_types[scheme]
return _transport_cls(host=self.store_uri.hostname,
port=self.store_uri.port)
@classmethod
def get_name(cls):
return "riemann"
@classmethod
def get_config_opts(cls):
return super(RiemannClient, cls).get_config_opts() + [
cfg.StrOpt(
'store_endpoint',
help="Complete Riemann endpoint, e.g. tcp://localhost:5555.",
default=None,
required=True),
cfg.StrOpt(
'default_metric_host',
default='openstack',
help='The default host value used when no other host value is '
'determined, such as from the metric resource_metadata.'),
cfg.IntOpt(
'default_metric_ttl',
default=86400,
help='The default amount of time in seconds that events sent '
'to Riemann are retained in its index.'),
cfg.StrOpt(
'default_metric_state',
default='ok',
help='The default state value used when no other state value '
'is determined, such as from the metric '
'resource_metadata.'),
cfg.StrOpt(
'datacenter',
required=True,
help='The current datacenter name',
),
]
@property
def store_endpoint(self):
return self._store_endpoint
def connect(self):
try:
self._transport.connect()
except (OSError, RuntimeError) as exc:
raise MetricsStoreError(exc)
LOG.info("[Riemann] Client connected")
def disconnect(self):
try:
self._transport.disconnect()
except (OSError, RuntimeError) as exc:
raise MetricsStoreError(exc)
LOG.info("[Riemann] Client disconnected")
def reconnect(self):
try:
LOG.debug("[Riemann] ==> Reconnecting...")
self.disconnect()
except MetricsStoreError as exc:
LOG.exception(exc)
try:
self.connect()
except MetricsStoreError as exc:
LOG.exception(exc)
LOG.debug("[Riemann] Socket connection re-established")
def _send(self, event):
try:
self.client.send_event(event)
except (transport.RiemannError, struct.error, OSError,
RuntimeError) as exc:
LOG.warn("Could not deliver the message "
"to the Riemann server! Trying again...")
LOG.exception(exc)
raise MetricsStoreError("Could not deliver the message "
"to the Riemann server.")
except TypeError as exc:
LOG.debug("Message formatting issue with %r",
Client.create_dict(event))
LOG.exception(exc)
raise MetricsStoreError("Message formatting issue => "
"%r: '%r'" %
(Client.create_dict(event), exc))
def send(self, metric):
LOG.debug('Publishing metrics to `%s:%d`', self.store_uri.hostname,
self.store_uri.port)
try:
event_msg = self.create_event(metric)
except (MetricsStoreError, KeyError) as exc:
LOG.exception(exc)
raise MetricsStoreError(exc)
for _ in range(2): # 2 attempts
try:
self._send(event_msg)
break
except Exception as exc:
LOG.warn('Unable to send metric `%r`', metric)
LOG.exception(exc)
LOG.error("Connection issue --> try again...")
self.reconnect()
def create_event(self, metric):
"""Translates a dictionary of event attributes to an Event object
:param dict metric: The attributes to be set on the event
:returns: A protocol buffer ``Event`` object
"""
description = "" # Needs to be specified if any
tags = [] # Needs to be specified if any
metadata = metric.get("resource_metadata", {})
host = metadata.get("host", self.default_metric_host)
ttl = metadata.get("ttl", self.default_metric_ttl)
state = metadata.get("state", self.default_metric_state)
try:
timestamp = metric.get("timestamp", "")
parsed = parser.parse(timestamp)
epoch_seconds = calendar.timegm(parsed.timetuple())
except ValueError:
error_message = "Could not parse the `%r` timestamp! " % timestamp
LOG.exception(error_message)
raise MetricsStoreError(error_message)
event_dict = dict(
time=epoch_seconds,
ttl=int(ttl),
host=host,
service=metric["name"],
state=state,
metric_f=metric["value"],
description=description,
tags=tags,
attributes=self._build_attributes(metric),
)
try:
return self.event_factory(event_dict)
except TypeError as exc:
LOG.exception(exc)
raise MetricsStoreError(exc)
def event_factory(self, data):
event = Event()
event.tags.extend(data.pop('tags', []))
for key, value in data.pop('attributes', {}).items():
attribute = event.attributes.add()
attribute.key, attribute.value = key, value
for name, value in data.items():
if value is not None:
setattr(event, name, value)
return event
def _build_attributes(self, metric):
ignored_metadata = ["host", "ttl"]
metadata = metric.get("resource_metadata", {})
attributes = {
key: val
for key, val in metadata.items() if key not in ignored_metadata
}
attributes["datacenter"] = self.datacenter or ""
attributes["resource_id"] = metric.get("resource_id", "")
attributes["unit"] = metric.get("unit", "")
return attributes
class RiemannHandler(Handler):
def __init__(self, config=None):
# Initialize Handler
Handler.__init__(self, config)
if riemann_client is None:
logging.error("Failed to load riemann_client module")
return
# Initialize options
self.host = self.config['host']
self.port = int(self.config['port'])
self.transport = self.config['transport']
# Initialize client
if self.transport == 'tcp':
self.transport = TCPTransport(self.host, self.port)
else:
self.transport = UDPTransport(self.host, self.port)
self.client = Client(self.transport)
self._connect()
def get_default_config_help(self):
"""
Returns the help text for the configuration options for this handler
"""
config = super(RiemannHandler, self).get_default_config_help()
config.update({
'host': '',
'port': '',
'transport': 'tcp or udp',
})
return config
def get_default_config(self):
"""
Return the default config for the handler
"""
config = super(RiemannHandler, self).get_default_config()
config.update({
'host': '',
'port': 123,
'transport': 'tcp',
})
return config
def process(self, metric):
"""
Send a metric to Riemann.
"""
event = self._metric_to_riemann_event(metric)
try:
self.client.send_event(event)
except Exception, e:
self.log.error(
"RiemannHandler: Error sending event to Riemann: %s", e)
class RiemannBroker(BaseModule):
def __init__(self, modconf):
BaseModule.__init__(self, modconf)
self.host = getattr(modconf, 'host', 'localhost')
self.port = int(getattr(modconf, 'port', '5555'))
self.use_udp = getattr(modconf, 'use_udp', '0') == '1'
self.buffer = []
self.ticks = 0
self.tick_limit = int(getattr(modconf, 'tick_limit', '300'))
# Called by Broker so we can do init stuff
# Conf from arbiter!
def init(self):
logger.info("[riemann broker] I init the %s server connection to %s:%d" %
(self.get_name(), str(self.host), self.port))
if self.use_udp:
transport = UDPTransport(self.host, self.port)
else:
transport = TCPTransport(self.host, self.port)
transport.connect()
self.client = Client(transport)
# Returns a list of perfdata events
def get_check_result_perfdata_events(self, perf_data, timestamp, host, service):
events = []
metrics = PerfDatas(perf_data).metrics
for e in metrics.values():
event_data = {
'host': host,
'service': service,
'time': timestamp,
#'state': 'WARNING',
'metric_f': e.value,
'description': e.name,
}
attributes = {}
if e.uom is not None:
attributes['unit'] = unicode(e.uom)
if e.warning is not None:
attributes['warning'] = unicode(e.warning)
if e.critical is not None:
attributes['critical'] = unicode(e.critical)
if e.min is not None:
attributes['min'] = unicode(e.min)
if e.max is not None:
attributes['max'] = unicode(e.max)
if attributes is not {}:
event_data['attributes'] = attributes
events.append(
self.client.create_event(event_data)
)
return events
# Returns state_update points for a given check_result_brok data
def get_state_update_points(self, data):
events = []
if data['state'] != data['last_state'] or \
data['state_type'] != data['last_state_type']:
events.append(
self.client.create_event(
{
'host': data['host_name'],
'service': data['service_description'],
'time': data['last_chk'],
'attributes': {
'state': data['state'],
'state_type': data['state_type'],
'output': data['output']
}
}
)
)
return events
# A service check result brok has just arrived, we UPDATE data info with this
def manage_service_check_result_brok(self, b):
data = b.data
events = []
events.extend(
self.get_check_result_perfdata_events(
data['perf_data'],
data['last_chk'],
data['host_name'],
data['service_description']
)
)
events.extend(
self.get_state_update_points(data)
)
try:
logger.debug("[riemann broker] Launching: %s" % str(events))
except UnicodeEncodeError:
pass
self.buffer.extend(events)
# A host check result brok has just arrived, we UPDATE data info with this
def manage_host_check_result_brok(self, b):
data = b.data
events = []
events.extend(
self.get_check_result_perfdata_events(
data['perf_data'],
data['last_chk'],
data['host_name'],
None,
)
)
events.extend(
self.get_state_update_points(data)
)
try:
logger.debug("[riemann broker] Launching: %s" % str(events))
except UnicodeEncodeError:
pass
self.buffer.extend(events)
def manage_unknown_host_check_result_brok(self, b):
data = b.data
events = []
events.extend(
self.get_check_result_perfdata_events(
data['perf_data'],
data['last_chk'],
data['host_name'],
None,
)
)
try:
logger.debug("[riemann broker] Launching: %s" % str(events))
except UnicodeEncodeError:
pass
self.buffer.extend(events)
def manage_unknown_service_check_result_brok(self, b):
data = b.data
events = []
events.extend(
self.get_check_result_perfdata_events(
data['perf_data'],
data['time_stamp'],
data['host_name'],
data['service_description']
)
)
try:
logger.debug("[riemann broker] Launching: %s" % str(events))
except UnicodeEncodeError:
pass
self.buffer.extend(events)
# A log brok has arrived, we UPDATE data info with this
def manage_log_brok(self, b):
log = b.data['log']
log_event = LogEvent(log)
if len(log_event) > 0:
event_data = {
'host': b.data['hostname'],
'service': b.data['service_desc'],
'description': b.data['event_type'],
'time': b.data['time']
}
attributes = {}
# Add each property of the service in the point
for prop in log_event:
attributes[prop[0]] = unicode(prop[1])
self.buffer.append(
self.client.create_event(event_data)
)
def hook_tick(self, brok):
if self.ticks >= self.tick_limit:
logger.error("[riemann broker] Buffering ticks exceeded. Freeing buffer")
self.buffer = []
self.ticks = 0
if len(self.buffer) > 0:
try:
for event in self.buffer:
self.client.send_event(event)
self.buffer.remove(event)
self.ticks = 0
except:
self.ticks += 1
logger.error("[riemann broker] Sending data Failed. Buffering state : %s / %s"
% (self.ticks, self.tick_limit))
class RiemannClient(MetricsStoreClientBase):
"""Riemann client"""
def __init__(self, store_endpoint, datacenter, default_metric_host,
default_metric_ttl, default_metric_state):
"""
:param store_endpoint: Riemann server endpoint
:type store_endpoint: str
:param datacenter: The current datacenter name
:type datacenter: str
:param default_metric_host: The default host value used when no other
host has been provided from the metric resource_metadata.
:type default_metric_host: str
:param default_metric_ttl: The amount of time in seconds that events
sent to Riemann are retained in its index.
:type default_metric_ttl: int
:param default_metric_state: The default state value used when no other
state has been provided from the metric resource_metadata.
:type default_metric_state: str
"""
super(RiemannClient, self).__init__()
self._store_endpoint = store_endpoint
self.store_uri = urlparse(self.store_endpoint)
self.default_metric_host = default_metric_host
self.default_metric_ttl = default_metric_ttl
self.default_metric_state = default_metric_state
self.datacenter = datacenter
self._transport = self._create_transport(self.store_uri.scheme)
self.client = Client(
transport=self._transport,
)
def _create_transport(self, scheme):
transport_types = dict(
tcp=transport.TCPTransport,
udp=transport.UDPTransport,
)
_transport_cls = transport_types[scheme]
return _transport_cls(
host=self.store_uri.hostname,
port=self.store_uri.port
)
@classmethod
def get_name(cls):
return "riemann"
@classmethod
def get_config_opts(cls):
return super(RiemannClient, cls).get_config_opts() + [
cfg.StrOpt(
'store_endpoint',
help="Complete Riemann endpoint, e.g. tcp://localhost:5555.",
default=None, required=True),
cfg.StrOpt(
'default_metric_host',
default='openstack',
help='The default host value used when no other host value is '
'determined, such as from the metric resource_metadata.'
),
cfg.IntOpt(
'default_metric_ttl',
default=86400,
help='The default amount of time in seconds that events sent '
'to Riemann are retained in its index.'
),
cfg.StrOpt(
'default_metric_state',
default='ok',
help='The default state value used when no other state value '
'is determined, such as from the metric '
'resource_metadata.'
),
cfg.StrOpt(
'datacenter',
required=True,
help='The current datacenter name',
),
]
@property
def store_endpoint(self):
return self._store_endpoint
def connect(self):
try:
self._transport.connect()
except (OSError, RuntimeError) as exc:
raise MetricsStoreError(exc)
LOG.info("[Riemann] Client connected")
def disconnect(self):
try:
self._transport.disconnect()
except (OSError, RuntimeError) as exc:
raise MetricsStoreError(exc)
LOG.info("[Riemann] Client disconnected")
def reconnect(self):
try:
LOG.debug("[Riemann] ==> Reconnecting...")
self.disconnect()
except MetricsStoreError as exc:
LOG.exception(exc)
try:
self.connect()
except MetricsStoreError as exc:
LOG.exception(exc)
LOG.debug("[Riemann] Socket connection re-established")
def _send(self, event):
try:
self.client.send_event(event)
except (transport.RiemannError, struct.error,
OSError, RuntimeError) as exc:
LOG.warn("Could not deliver the message "
"to the Riemann server! Trying again...")
LOG.exception(exc)
raise MetricsStoreError("Could not deliver the message "
"to the Riemann server.")
except TypeError as exc:
LOG.debug("Message formatting issue with %r",
Client.create_dict(event))
LOG.exception(exc)
raise MetricsStoreError(
"Message formatting issue => "
"%r: '%r'" % (Client.create_dict(event), exc)
)
def send(self, metric):
LOG.debug('Publishing metrics to `%s:%d`',
self.store_uri.hostname, self.store_uri.port)
try:
event_msg = self.create_event(metric)
except (MetricsStoreError, KeyError) as exc:
LOG.exception(exc)
raise MetricsStoreError(exc)
for _ in range(2): # 2 attempts
try:
self._send(event_msg)
break
except Exception as exc:
LOG.warn('Unable to send metric `%r`', metric)
LOG.exception(exc)
LOG.error("Connection issue --> try again...")
self.reconnect()
def create_event(self, metric):
"""Translates a dictionary of event attributes to an Event object
:param dict metric: The attributes to be set on the event
:returns: A protocol buffer ``Event`` object
"""
description = "" # Needs to be specified if any
tags = [] # Needs to be specified if any
metadata = metric.get("resource_metadata", {})
host = metadata.get("host", self.default_metric_host)
ttl = metadata.get("ttl", self.default_metric_ttl)
state = metadata.get("state", self.default_metric_state)
try:
timestamp = metric.get("timestamp", "")
parsed = parser.parse(timestamp)
epoch_seconds = calendar.timegm(parsed.timetuple())
except ValueError:
error_message = "Could not parse the `%r` timestamp! " % timestamp
LOG.exception(error_message)
raise MetricsStoreError(error_message)
event_dict = dict(
time=epoch_seconds,
ttl=int(ttl),
host=host,
service=metric["name"],
state=state,
metric_f=metric["value"],
description=description,
tags=tags,
attributes=self._build_attributes(metric),
)
try:
return self.event_factory(event_dict)
except TypeError as exc:
LOG.exception(exc)
raise MetricsStoreError(exc)
def event_factory(self, data):
event = Event()
event.tags.extend(data.pop('tags', []))
for key, value in data.pop('attributes', {}).items():
attribute = event.attributes.add()
attribute.key, attribute.value = key, value
for name, value in data.items():
if value is not None:
setattr(event, name, value)
return event
def _build_attributes(self, metric):
ignored_metadata = ["host", "ttl"]
metadata = metric.get("resource_metadata", {})
attributes = {
key: val for key, val in metadata.items()
if key not in ignored_metadata
}
attributes["datacenter"] = self.datacenter or ""
attributes["resource_id"] = metric.get("resource_id", "")
attributes["unit"] = metric.get("unit", "")
return attributes