def check(self, instance):
# read config
tags = instance.get('tags', [])
host = instance.get('host', '127.0.0.1')
port = instance.get('port', 23456)
prefix = instance.get('prefix', 'plog.')
suffix = instance.get('suffix', '')
timeout = instance.get('timeout', 3)
max_size = instance.get('max_size', 65536)
# create socket, ask for stats
sock = socket(AF_INET, SOCK_DGRAM)
try:
sock.sendto("\0\0statfordatadogplease", (host, port))
# wait for a reply
ready = select([sock], [], [], timeout)
if not ready[0]:
raise socketError('timeout')
data, addr = sock.recvfrom(max_size)
stats = loads(data)
def counter(name, val):
self.rate(prefix + name + suffix, val, tags=tags)
def rate(name, val):
self.gauge(prefix + name + suffix, val, tags=tags)
rate('uptime',
stats['uptime'])
counter('exceptions',
stats['exceptions'])
counter('unhandled_objects',
stats['unhandled_objects'])
counter('udp_simple',
stats['udp_simple_messages'])
counter('holes.from_dead_port',
stats['holes_from_dead_port'])
counter('holes.from_new_message',
stats['holes_from_new_message'])
counter('invalid_checksum',
sum(stats['v0_invalid_checksum']))
counter('fragments',
sum(stats['v0_fragments']))
counter('missing_fragments',
sum(sum(a) for a in stats['dropped_fragments']))
counter('invalid_fragments',
sum(sum(a) for a in stats['v0_invalid_fragments']))
cache = stats['cache']
counter('cache.evictions',
cache['evictions'])
counter('cache.hits',
cache['hits'])
counter('cache.miss',
cache['misses'])
kafka = stats['kafka']
rate('kafka.messages',
kafka['messageRate']['rate'][0])
rate('kafka.dropped',
kafka['droppedMessageRate']['rate'][0])
rate('kafka.bytes',
kafka['byteRate']['rate'][0])
rate('kafka.resends',
kafka['resendRate']['rate'][0])
rate('kafka.failed_sends',
kafka['failedSendRate']['rate'][0])
rate('kafka.serialization_errors',
kafka['serializationErrorRate']['rate'][0])
finally:
sock.close()
def check(self, instance):
# read config
tags = instance.get('tags', [])
host = instance.get('host', '127.0.0.1')
port = instance.get('port', 23456)
prefix = instance.get('prefix', 'plog.')
suffix = instance.get('suffix', '')
timeout = instance.get('timeout', 3)
max_size = instance.get('max_size', 65536)
# create socket, ask for stats
sock = socket(AF_INET, SOCK_DGRAM)
try:
sock.sendto("\0\0statfordatadogplease", (host, port))
# wait for a reply
ready = select([sock], [], [], timeout)
if not ready[0]:
raise socketError('timeout')
data, addr = sock.recvfrom(max_size)
stats = loads(data)
finally:
sock.close()
def rate(name, val):
self.rate(prefix + name + suffix, val, tags=tags)
def gauge(name, val):
self.gauge(prefix + name + suffix, val, tags=tags)
gauge('uptime', stats['uptime'])
rate('udp_simple', stats['udp_simple_messages'])
rate('udp_invalid_version', stats['udp_invalid_version'])
rate('v0_invalid_type', stats['v0_invalid_type'])
rate('v0_invalid_multipart_header',
stats['v0_invalid_multipart_header'])
rate('unknown_command', stats['unknown_command'])
rate('v0_commands', stats['v0_commands'])
rate('exceptions', stats['exceptions'])
rate('unhandled_objects', stats['unhandled_objects'])
rate('holes.from_dead_port', stats['holes_from_dead_port'])
rate('holes.from_new_message', stats['holes_from_new_message'])
rate('fragments', sum(stats['v0_fragments']))
rate('invalid_checksum', sum(stats['v0_invalid_checksum']))
rate('invalid_fragments',
sum(sum(a) for a in stats['v0_invalid_fragments']))
rate('missing_fragments',
sum(sum(a) for a in stats['dropped_fragments']))
defragmenter = stats['defragmenter']
rate('defragmenter.evictions', defragmenter['evictions'])
rate('defragmenter.hits', defragmenter['hits'])
rate('defragmenter.miss', defragmenter['misses'])
def flatsum(json):
if isinstance(json, Number):
return json
elif isinstance(json, list):
return sum(flatsum(o) for o in json)
else:
return 0
def feed_handler_metrics(path, json):
if isinstance(json, dict):
for k in json:
feed_handler_metrics(path + '.' + k, json[k])
else:
existing = handler_metrics.get(path, 0)
handler_metrics[path] = existing + flatsum(json)
handlers = stats['handlers']
handler_metrics = {}
for handler in handlers:
handler_name = handler['name']
for metric_name in handler:
if metric_name != 'name':
metric_path = handler_name + '.' + metric_name
feed_handler_metrics(metric_path, handler[metric_name])
for path in handler_metrics:
# TODO: rates instead of gauges through YAML config?
# We can use rate(my.metric{*}) on Datadog's side for graphing, but alerts
# do not support functions _yet_.
gauge(path, handler_metrics[path])
def check(self, instance):
# read config
tags = instance.get('tags', [])
host = instance.get('host', '127.0.0.1')
port = instance.get('port', 23456)
prefix = instance.get('prefix', 'plog.')
suffix = instance.get('suffix', '')
timeout = instance.get('timeout', 3)
max_size = instance.get('max_size', 65536)
# create socket, ask for stats
sock = socket(AF_INET, SOCK_DGRAM)
try:
sock.sendto("\0\0statfordatadogplease", (host, port))
# wait for a reply
ready = select([sock], [], [], timeout)
if not ready[0]:
raise socketError('timeout')
data, addr = sock.recvfrom(max_size)
stats = loads(data)
finally:
sock.close()
def rate(name, val):
self.rate(prefix + name + suffix, val, tags=tags)
def gauge(name, val):
self.gauge(prefix + name + suffix, val, tags=tags)
gauge('uptime',
stats['uptime'])
rate('udp_simple',
stats['udp_simple_messages'])
rate('udp_invalid_version',
stats['udp_invalid_version'])
rate('v0_invalid_type',
stats['v0_invalid_type'])
rate('v0_invalid_multipart_header',
stats['v0_invalid_multipart_header'])
rate('unknown_command',
stats['unknown_command'])
rate('v0_commands',
stats['v0_commands'])
rate('exceptions',
stats['exceptions'])
rate('unhandled_objects',
stats['unhandled_objects'])
rate('holes.from_dead_port',
stats['holes_from_dead_port'])
rate('holes.from_new_message',
stats['holes_from_new_message'])
rate('fragments',
sum(stats['v0_fragments']))
rate('invalid_checksum',
sum(stats['v0_invalid_checksum']))
rate('invalid_fragments',
sum(sum(a) for a in stats['v0_invalid_fragments']))
rate('missing_fragments',
sum(sum(a) for a in stats['dropped_fragments']))
defragmenter = stats['defragmenter']
rate('defragmenter.evictions',
defragmenter['evictions'])
rate('defragmenter.hits',
defragmenter['hits'])
rate('defragmenter.miss',
defragmenter['misses'])
def flatsum(json):
if isinstance(json, Number):
return json
elif isinstance(json, list):
return sum(flatsum(o) for o in json)
else:
return 0
def feed_handler_metrics(path, json):
if isinstance(json, dict):
for k in json:
feed_handler_metrics(path + '.' + k, json[k])
else:
existing = handler_metrics.get(path, 0)
handler_metrics[path] = existing + flatsum(json)
handlers = stats['handlers']
handler_metrics = {}
for handler in handlers:
handler_name = handler['name']
for metric_name in handler:
if metric_name != 'name':
metric_path = handler_name + '.' + metric_name
feed_handler_metrics(metric_path, handler[metric_name])
for path in handler_metrics:
# TODO: rates instead of gauges through YAML config?
# We can use rate(my.metric{*}) on Datadog's side for graphing, but alerts
# do not support functions _yet_.
gauge(path, handler_metrics[path])