stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
import os
|
|
|
|
import pickle
|
|
|
|
import pprint
|
|
|
|
import sys
|
|
|
|
import time
|
|
|
|
from collections import deque
|
|
|
|
|
|
|
|
from twisted.internet import reactor, defer
|
|
|
|
from twisted.application import service
|
|
|
|
from twisted.application.internet import TimerService
|
|
|
|
from zope.interface import implements
|
|
|
|
import foolscap
|
2008-04-30 01:12:53 +00:00
|
|
|
from foolscap.eventual import eventually
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
from foolscap.logging.gatherer import get_local_ip_for
|
2008-03-04 06:55:58 +00:00
|
|
|
from twisted.internet.error import ConnectionDone, ConnectionLost
|
|
|
|
from foolscap import DeadReferenceError
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
from allmydata.util import log
|
2008-02-01 04:10:15 +00:00
|
|
|
from allmydata.interfaces import RIStatsProvider, RIStatsGatherer, IStatsProducer
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
class LoadMonitor(service.MultiService):
|
2008-02-01 04:10:15 +00:00
|
|
|
implements(IStatsProducer)
|
|
|
|
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
loop_interval = 1
|
|
|
|
num_samples = 60
|
|
|
|
|
|
|
|
def __init__(self, provider, warn_if_delay_exceeds=1):
|
|
|
|
service.MultiService.__init__(self)
|
|
|
|
self.provider = provider
|
|
|
|
self.warn_if_delay_exceeds = warn_if_delay_exceeds
|
2008-02-02 01:57:31 +00:00
|
|
|
self.started = False
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
self.last = None
|
|
|
|
self.stats = deque()
|
2008-03-04 06:55:58 +00:00
|
|
|
self.timer = None
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def startService(self):
|
2008-02-02 01:57:31 +00:00
|
|
|
if not self.started:
|
|
|
|
self.started = True
|
2008-03-04 06:55:58 +00:00
|
|
|
self.timer = reactor.callLater(self.loop_interval, self.loop)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
service.MultiService.startService(self)
|
|
|
|
|
|
|
|
def stopService(self):
|
2008-02-02 01:57:31 +00:00
|
|
|
self.started = False
|
2008-03-04 06:55:58 +00:00
|
|
|
if self.timer:
|
|
|
|
self.timer.cancel()
|
|
|
|
self.timer = None
|
|
|
|
return service.MultiService.stopService(self)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def loop(self):
|
2008-03-04 06:55:58 +00:00
|
|
|
self.timer = None
|
2008-02-02 01:57:31 +00:00
|
|
|
if not self.started:
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
return
|
|
|
|
now = time.time()
|
|
|
|
if self.last is not None:
|
|
|
|
delay = now - self.last - self.loop_interval
|
|
|
|
if delay > self.warn_if_delay_exceeds:
|
|
|
|
log.msg(format='excessive reactor delay (%ss)', args=(delay,),
|
|
|
|
level=log.UNUSUAL)
|
|
|
|
self.stats.append(delay)
|
|
|
|
while len(self.stats) > self.num_samples:
|
|
|
|
self.stats.popleft()
|
|
|
|
|
|
|
|
self.last = now
|
2008-03-04 06:55:58 +00:00
|
|
|
self.timer = reactor.callLater(self.loop_interval, self.loop)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def get_stats(self):
|
|
|
|
if self.stats:
|
|
|
|
avg = sum(self.stats) / len(self.stats)
|
|
|
|
m_x = max(self.stats)
|
|
|
|
else:
|
|
|
|
avg = m_x = 0
|
|
|
|
return { 'load_monitor.avg_load': avg,
|
|
|
|
'load_monitor.max_load': m_x, }
|
|
|
|
|
2008-04-30 01:12:53 +00:00
|
|
|
class CPUUsageMonitor(service.MultiService):
|
|
|
|
implements(IStatsProducer)
|
2008-04-30 18:39:13 +00:00
|
|
|
HISTORY_LENGTH = 15
|
|
|
|
POLL_INTERVAL = 60
|
2008-04-30 01:12:53 +00:00
|
|
|
|
|
|
|
def __init__(self):
|
|
|
|
service.MultiService.__init__(self)
|
|
|
|
# we don't use time.clock() here, because the constructor is run by
|
|
|
|
# the twistd parent process (as it loads the .tac file), whereas the
|
|
|
|
# rest of the program will be run by the child process, after twistd
|
|
|
|
# forks. Instead, set self.initial_cpu as soon as the reactor starts
|
|
|
|
# up.
|
|
|
|
self.initial_cpu = 0.0 # just in case
|
|
|
|
eventually(self._set_initial_cpu)
|
|
|
|
self.samples = []
|
|
|
|
# we provide 1min, 5min, and 15min moving averages
|
2008-04-30 18:39:13 +00:00
|
|
|
TimerService(self.POLL_INTERVAL, self.check).setServiceParent(self)
|
2008-04-30 01:12:53 +00:00
|
|
|
|
|
|
|
def _set_initial_cpu(self):
|
|
|
|
self.initial_cpu = time.clock()
|
|
|
|
|
|
|
|
def check(self):
|
|
|
|
now_wall = time.time()
|
|
|
|
now_cpu = time.clock()
|
|
|
|
self.samples.append( (now_wall, now_cpu) )
|
2008-04-30 18:39:13 +00:00
|
|
|
while len(self.samples) > self.HISTORY_LENGTH+1:
|
2008-04-30 01:12:53 +00:00
|
|
|
self.samples.pop(0)
|
|
|
|
|
|
|
|
def _average_N_minutes(self, size):
|
|
|
|
if len(self.samples) < size+1:
|
|
|
|
return None
|
|
|
|
first = -size-1
|
|
|
|
elapsed_wall = self.samples[-1][0] - self.samples[first][0]
|
|
|
|
elapsed_cpu = self.samples[-1][1] - self.samples[first][1]
|
|
|
|
fraction = elapsed_cpu / elapsed_wall
|
|
|
|
return fraction
|
|
|
|
|
|
|
|
def get_stats(self):
|
|
|
|
s = {}
|
|
|
|
avg = self._average_N_minutes(1)
|
|
|
|
if avg is not None:
|
|
|
|
s["cpu_monitor.1min_avg"] = avg
|
|
|
|
avg = self._average_N_minutes(5)
|
|
|
|
if avg is not None:
|
|
|
|
s["cpu_monitor.5min_avg"] = avg
|
|
|
|
avg = self._average_N_minutes(15)
|
|
|
|
if avg is not None:
|
|
|
|
s["cpu_monitor.15min_avg"] = avg
|
|
|
|
now_cpu = time.clock()
|
|
|
|
s["cpu_monitor.total"] = now_cpu - self.initial_cpu
|
|
|
|
return s
|
|
|
|
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
class StatsProvider(foolscap.Referenceable, service.MultiService):
|
|
|
|
implements(RIStatsProvider)
|
|
|
|
|
2008-02-01 04:10:15 +00:00
|
|
|
def __init__(self, node, gatherer_furl):
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
service.MultiService.__init__(self)
|
2008-02-01 04:10:15 +00:00
|
|
|
self.node = node
|
2008-05-08 18:37:30 +00:00
|
|
|
self.gatherer_furl = gatherer_furl # might be None
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
self.counters = {}
|
|
|
|
self.stats_producers = []
|
|
|
|
|
2008-05-08 18:37:30 +00:00
|
|
|
# only run the LoadMonitor (which submits a timer every second) if
|
|
|
|
# there is a gatherer who is going to be paying attention. Our stats
|
|
|
|
# are visible through HTTP even without a gatherer, so run the rest
|
|
|
|
# of the stats (including the once-per-minute CPUUsageMonitor)
|
|
|
|
if gatherer_furl:
|
|
|
|
self.load_monitor = LoadMonitor(self)
|
|
|
|
self.load_monitor.setServiceParent(self)
|
|
|
|
self.register_producer(self.load_monitor)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
2008-04-30 01:12:53 +00:00
|
|
|
self.cpu_monitor = CPUUsageMonitor()
|
|
|
|
self.cpu_monitor.setServiceParent(self)
|
|
|
|
self.register_producer(self.cpu_monitor)
|
|
|
|
|
2008-02-01 04:10:15 +00:00
|
|
|
def startService(self):
|
2008-05-08 18:37:30 +00:00
|
|
|
if self.node and self.gatherer_furl:
|
2008-02-01 04:10:15 +00:00
|
|
|
d = self.node.when_tub_ready()
|
|
|
|
def connect(junk):
|
|
|
|
nickname = self.node.get_config('nickname')
|
2008-05-08 18:37:30 +00:00
|
|
|
self.node.tub.connectTo(self.gatherer_furl,
|
|
|
|
self._connected, nickname)
|
2008-02-01 04:10:15 +00:00
|
|
|
d.addCallback(connect)
|
2008-02-02 01:57:31 +00:00
|
|
|
service.MultiService.startService(self)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
2008-04-11 00:25:44 +00:00
|
|
|
def count(self, name, delta=1):
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
val = self.counters.setdefault(name, 0)
|
|
|
|
self.counters[name] = val + delta
|
|
|
|
|
|
|
|
def register_producer(self, stats_producer):
|
2008-02-01 04:10:15 +00:00
|
|
|
self.stats_producers.append(IStatsProducer(stats_producer))
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
2008-04-14 21:17:08 +00:00
|
|
|
def get_stats(self):
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
stats = {}
|
|
|
|
for sp in self.stats_producers:
|
|
|
|
stats.update(sp.get_stats())
|
2008-04-09 23:10:53 +00:00
|
|
|
ret = { 'counters': self.counters, 'stats': stats }
|
2008-04-14 21:17:08 +00:00
|
|
|
log.msg(format='get_stats() -> %(stats)s', stats=ret, level=log.NOISY)
|
2008-04-09 23:10:53 +00:00
|
|
|
return ret
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
2008-04-14 21:17:08 +00:00
|
|
|
def remote_get_stats(self):
|
|
|
|
return self.get_stats()
|
|
|
|
|
2008-02-01 04:10:15 +00:00
|
|
|
def _connected(self, gatherer, nickname):
|
2008-03-04 06:55:58 +00:00
|
|
|
gatherer.callRemoteOnly('provide', self, nickname or '')
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
class StatsGatherer(foolscap.Referenceable, service.MultiService):
|
|
|
|
implements(RIStatsGatherer)
|
|
|
|
|
|
|
|
poll_interval = 60
|
|
|
|
|
2008-03-04 06:55:58 +00:00
|
|
|
def __init__(self, tub, basedir):
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
service.MultiService.__init__(self)
|
|
|
|
self.tub = tub
|
2008-03-04 06:55:58 +00:00
|
|
|
self.basedir = basedir
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
self.clients = {}
|
|
|
|
self.nicknames = {}
|
|
|
|
|
|
|
|
def startService(self):
|
2008-03-04 06:55:58 +00:00
|
|
|
# the Tub must have a location set on it by now
|
|
|
|
service.MultiService.startService(self)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
self.timer = TimerService(self.poll_interval, self.poll)
|
|
|
|
self.timer.setServiceParent(self)
|
2008-03-04 06:55:58 +00:00
|
|
|
self.registerGatherer()
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def get_furl(self):
|
2008-03-04 06:55:58 +00:00
|
|
|
return self.my_furl
|
|
|
|
|
|
|
|
def registerGatherer(self):
|
|
|
|
furl_file = os.path.join(self.basedir, "stats_gatherer.furl")
|
|
|
|
self.my_furl = self.tub.registerReference(self, furlFile=furl_file)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def get_tubid(self, rref):
|
|
|
|
return foolscap.SturdyRef(rref.tracker.getURL()).getTubRef().getTubID()
|
|
|
|
|
|
|
|
def remote_provide(self, provider, nickname):
|
|
|
|
tubid = self.get_tubid(provider)
|
2008-02-01 02:11:31 +00:00
|
|
|
if tubid == '<unauth>':
|
|
|
|
print "WARNING: failed to get tubid for %s (%s)" % (provider, nickname)
|
|
|
|
# don't add to clients to poll (polluting data) don't care about disconnect
|
|
|
|
return
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
self.clients[tubid] = provider
|
|
|
|
self.nicknames[tubid] = nickname
|
|
|
|
|
|
|
|
def poll(self):
|
|
|
|
for tubid,client in self.clients.items():
|
|
|
|
nickname = self.nicknames.get(tubid)
|
|
|
|
d = client.callRemote('get_stats')
|
2008-03-04 06:55:58 +00:00
|
|
|
d.addCallbacks(self.got_stats, self.lost_client,
|
|
|
|
callbackArgs=(tubid, nickname),
|
|
|
|
errbackArgs=(tubid,))
|
|
|
|
d.addErrback(self.log_client_error, tubid)
|
|
|
|
|
|
|
|
def lost_client(self, f, tubid):
|
|
|
|
# this is called lazily, when a get_stats request fails
|
|
|
|
del self.clients[tubid]
|
|
|
|
del self.nicknames[tubid]
|
|
|
|
f.trap(DeadReferenceError, ConnectionDone, ConnectionLost)
|
|
|
|
|
|
|
|
def log_client_error(self, f, tubid):
|
|
|
|
log.msg("StatsGatherer: error in get_stats(), peerid=%s" % tubid,
|
|
|
|
level=log.UNUSUAL, failure=f)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
|
|
|
def got_stats(self, stats, tubid, nickname):
|
|
|
|
raise NotImplementedError()
|
|
|
|
|
|
|
|
class StdOutStatsGatherer(StatsGatherer):
|
2008-03-04 06:55:58 +00:00
|
|
|
verbose = True
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
def remote_provide(self, provider, nickname):
|
|
|
|
tubid = self.get_tubid(provider)
|
2008-03-04 06:55:58 +00:00
|
|
|
if self.verbose:
|
|
|
|
print 'connect "%s" [%s]' % (nickname, tubid)
|
|
|
|
provider.notifyOnDisconnect(self.announce_lost_client, tubid)
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
StatsGatherer.remote_provide(self, provider, nickname)
|
|
|
|
|
2008-03-04 06:55:58 +00:00
|
|
|
def announce_lost_client(self, tubid):
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
print 'disconnect "%s" [%s]:' % (self.nicknames[tubid], tubid)
|
|
|
|
|
|
|
|
def got_stats(self, stats, tubid, nickname):
|
|
|
|
print '"%s" [%s]:' % (nickname, tubid)
|
|
|
|
pprint.pprint(stats)
|
|
|
|
|
2008-03-04 06:55:58 +00:00
|
|
|
class PickleStatsGatherer(StdOutStatsGatherer):
|
|
|
|
# inherit from StdOutStatsGatherer for connect/disconnect notifications
|
|
|
|
|
|
|
|
def __init__(self, tub, basedir=".", verbose=True):
|
|
|
|
self.verbose = verbose
|
|
|
|
StatsGatherer.__init__(self, tub, basedir)
|
|
|
|
self.picklefile = os.path.join(basedir, "stats.pickle")
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
|
2008-03-04 06:55:58 +00:00
|
|
|
if os.path.exists(self.picklefile):
|
|
|
|
f = open(self.picklefile, 'rb')
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
self.gathered_stats = pickle.load(f)
|
|
|
|
f.close()
|
|
|
|
else:
|
|
|
|
self.gathered_stats = {}
|
|
|
|
|
|
|
|
def got_stats(self, stats, tubid, nickname):
|
|
|
|
s = self.gathered_stats.setdefault(tubid, {})
|
|
|
|
s['timestamp'] = time.time()
|
|
|
|
s['nickname'] = nickname
|
|
|
|
s['stats'] = stats
|
|
|
|
self.dump_pickle()
|
|
|
|
|
|
|
|
def dump_pickle(self):
|
|
|
|
tmp = "%s.tmp" % (self.picklefile,)
|
|
|
|
f = open(tmp, 'wb')
|
|
|
|
pickle.dump(self.gathered_stats, f)
|
|
|
|
f.close()
|
|
|
|
if os.path.exists(self.picklefile):
|
|
|
|
os.unlink(self.picklefile)
|
|
|
|
os.rename(tmp, self.picklefile)
|
|
|
|
|
|
|
|
class GathererApp(object):
|
|
|
|
def __init__(self):
|
|
|
|
d = self.setup_tub()
|
|
|
|
d.addCallback(self._tub_ready)
|
|
|
|
|
|
|
|
def setup_tub(self):
|
|
|
|
self._tub = foolscap.Tub(certFile="stats_gatherer.pem")
|
|
|
|
self._tub.setOption("logLocalFailures", True)
|
|
|
|
self._tub.setOption("logRemoteFailures", True)
|
|
|
|
self._tub.startService()
|
|
|
|
portnumfile = "portnum"
|
|
|
|
try:
|
|
|
|
portnum = int(open(portnumfile, "r").read())
|
|
|
|
except (EnvironmentError, ValueError):
|
|
|
|
portnum = 0
|
|
|
|
self._tub.listenOn("tcp:%d" % portnum)
|
|
|
|
d = defer.maybeDeferred(get_local_ip_for)
|
|
|
|
d.addCallback(self._set_location)
|
|
|
|
d.addCallback(lambda res: self._tub)
|
|
|
|
return d
|
|
|
|
|
|
|
|
def _set_location(self, local_address):
|
|
|
|
if local_address is None:
|
|
|
|
local_addresses = ["127.0.0.1"]
|
|
|
|
else:
|
|
|
|
local_addresses = [local_address, "127.0.0.1"]
|
|
|
|
l = self._tub.getListeners()[0]
|
|
|
|
portnum = l.getPortnum()
|
|
|
|
portnumfile = "portnum"
|
|
|
|
open(portnumfile, "w").write("%d\n" % portnum)
|
|
|
|
local_addresses = [ "%s:%d" % (addr, portnum,)
|
|
|
|
for addr in local_addresses ]
|
|
|
|
assert len(local_addresses) >= 1
|
|
|
|
location = ",".join(local_addresses)
|
|
|
|
self._tub.setLocation(location)
|
|
|
|
|
|
|
|
def _tub_ready(self, tub):
|
2008-03-04 06:55:58 +00:00
|
|
|
sg = PickleStatsGatherer(tub, ".")
|
stats: add a simple stats gathering system
We have a desire to collect runtime statistics from multiple nodes primarily
for server monitoring purposes. This implements a simple implementation of
such a system, as a skeleton to build more sophistication upon.
Each client now looks for a 'stats_gatherer.furl' config file. If it has
been configured to use a stats gatherer, then it instantiates internally
a StatsProvider. This is a central place for code which wishes to offer
stats up for monitoring to report them to, either by calling
stats_provider.count('stat.name', value) to increment a counter, or by
registering a class as a stats producer with sp.register_producer(obj).
The StatsProvider connects to the StatsGatherer server and provides its
provider upon startup. The StatsGatherer is then responsible for polling
the attached providers periodically to retrieve the data provided.
The provider queries each registered producer when the gatherer queries
the provider. Both the internal 'counters' and the queried 'stats' are
then reported to the gatherer.
This provides a simple gatherer app, (c.f. make stats-gatherer-run)
which prints its furl and listens for incoming connections. Once a
minute, the gatherer polls all connected providers, and writes the
retrieved data into a pickle file.
Also included is a munin plugin which knows how to read the gatherer's
stats.pickle and output data munin can interpret. this plugin,
tahoe-stats.py can be symlinked as multiple different names within
munin's 'plugins' directory, and inspects argv to determine which
data to display, doing a lookup in a table within that file.
It looks in the environment for 'statsfile' to determine the path to
the gatherer's stats.pickle. An example plugins-conf.d file is
provided.
2008-01-31 03:11:07 +00:00
|
|
|
sg.setServiceParent(tub)
|
|
|
|
sg.verbose = True
|
|
|
|
print '\nStatsGatherer: %s\n' % (sg.get_furl(),)
|
|
|
|
|
|
|
|
def main(argv):
|
|
|
|
ga = GathererApp()
|
|
|
|
reactor.run()
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
main(sys.argv)
|