def check_parallel_run(self):
# TODO: other daemon run on nt
if os.name == 'nt':
logger.warning("The parallel daemon check is not available on nt")
self.open_pidfile(write=True)
return
# First open the pid file in open mode
self.open_pidfile()
try:
buf = self.fpid.readline().strip(' \r\n')
if not buf: # pid file was void, cool
return
pid = int(buf)
except Exception as err:
logger.info("Stale pidfile exists at %s (%s). Reusing it." % (err, self.lock_path))
return
try:
os.kill(pid, 0)
except Exception as err: # consider any exception as a stale pidfile.
# this includes :
# * PermissionError when a process with same pid exists but is executed by another user.
# * ProcessLookupError: [Errno 3] No such process.
logger.info("Stale pidfile exists (%s), Reusing it." % err)
return
logger.error("Valid previous daemon exists (pid=%s) Exiting." % pid)
raise SystemExit(2)
def _parse_cgroup_file(self, stat_file):
try:
logger.debug("Opening cgroup file: %s" % stat_file)
with open(stat_file) as fp:
return dict(map(lambda x: x.split(), fp.read().splitlines()))
except IOError:
# It is possible that the container got stopped between the API call and now
logger.info("Can't open %s. Theses metrics for this container are skipped." % stat_file)
return None
def graphite_reaper(self):
while True:
graphite_queue = self.graphite_queue
self.graphite_queue = []
if len(graphite_queue) > 0:
logger.info("Graphite queue", len(graphite_queue))
for data in graphite_queue:
T0 = time.time()
self.grok_graphite_data(data)
STATS.timer('ts.graphite.grok-graphite-data', (time.time() - T0)*1000)
time.sleep(0.1)
def w():
tid = 0
if libc:
tid = libc.syscall(186) # get the threadid when you are in it :)
logger.info('THREAD launch (%s) with thread id (%d)' % (name, tid))
try:
f(*args)
except Exception, exp:
output = cStringIO.StringIO()
traceback.print_exc(file=output)
logger.error("Thread %s is exiting on error. Back trace of this error: %s" % (name, output.getvalue()))
output.close()
def w(is_essential):
tid = 0
if libc:
tid = libc.syscall(186) # get the threadid when you are in it :)
logger.info("THREAD launch (%s) with thread id (%d)" % (name, tid))
try:
f(*args)
except Exception, exp:
output = cStringIO.StringIO()
traceback.print_exc(file=output)
logger.error("Thread %s is exiting on error. Back trace of this error: %s" % (name, output.getvalue()))
output.close()
if is_essential:
# Maybe the thread WAS an essential one (like http thread or something like this), if so
# catch it and close the whole daemon
logger.error(
"The thread %s was an essential one, we are stopping the daemon do not be in an invalid state"
% name
)
pubsub.pub("interrupt")
def do_reaper_thread(self):
while True:
now = int(time.time())
m = divmod(now, 60)[0]*60 # current minute
all_names = []
with self.data_lock:
all_names = self.data.keys()
logger.info("DOING reaper thread on %d elements" % len(all_names))
for name in all_names:
# Grok all minute entries
if name.startswith('min::'):
e = self.data.get(name, None)
# maybe some one delete the entry? should not be possible
if e is None:
continue
ctime = e['ctime']
logger.debug("REAPER old data for ", name)
# if the creation time of this structure is too old and
# really for data, force to save the entry in KV entry
if ctime < now - self.max_data_age and e['nb'] > 0:
STATS.incr('reaper-old-data', 1)
logger.debug("REAPER TOO OLD DATA FOR", name)
# get the raw metric name
_id = name[5:]
self.archive_minute(e, _id)
# the element was too old, so we can assume it won't be upadte again. Delete it's entry
try:
del self.data[name]
except:
pass
'''
# and set a new minute, the next one
n_minute = e['cur_min'] + 60
e = {'cur_min':n_minute, 'sum':0, 'min':None, 'max':None,
'values':[None for _ in xrange(60)], 'nb':0, 'ctime':now}
self.data[name] = e
'''
time.sleep(10)
def launch(self):
logger.debug('getMongoDBStatus: start')
if 'MongoDBServer' not in self.config or self.config['MongoDBServer'] == '':
logger.debug('getMongoDBStatus: config not set')
#return False
logger.debug('getMongoDBStatus: config set')
try:
import pymongo
from pymongo import Connection
except ImportError:
logger.error('Unable to import pymongo library')
return False
try:
parsed = urlparse.urlparse(self.config.get('mongodb_server', 'mongodb://localhost'))
mongoURI = ''
# Can't use attributes on Python 2.4
if parsed[0] != 'mongodb':
mongoURI = 'mongodb://'
if parsed[2]:
if parsed[0]:
mongoURI = mongoURI + parsed[0] + ':' + parsed[2]
else:
mongoURI = mongoURI + parsed[2]
else:
mongoURI = self.config.get('mongodb_server', 'mongodb://localhost')
logger.debug('-- mongoURI: %s', mongoURI)
conn = Connection(mongoURI, slave_okay=True)
logger.debug('Connected to MongoDB')
except pymongo.errors.ConnectionFailure, exp:
logger.info('Unable to connect to MongoDB server %s - Exception = %s' % (mongoURI, exp))
return False
self.containers[_id] = c
def load_containers(self):
if not self.con:
return
try:
conts = get_json('/containers/json', local_socket='/var/run/docker.sock')
except request_errors, exp:
self.connect()
return
for c in conts:
_id = c.get('Id')
self.load_container(_id)
logger.info("Loading docker container %s" % _id)
logger.debug("Container data", self.containers[_id])
def load_images(self):
if not self.con:
return
try:
self.images = get_json('/images/json', local_socket='/var/run/docker.sock')
except request_errors, exp:
self.connect()
return
def compute_stats(self):
def unlink(self):
logger.info("Unlinking lock file %s" % self.lock_path)
try:
os.unlink(self.lock_path)
except Exception, e:
logger.error("Got an error unlinking our pidfile: %s" % (e))
# halfway to daemonize..
os.setsid()
try:
pid = os.fork()
except OSError, e:
raise Exception, "%s [%d]" % (e.strerror, e.errno)
if pid != 0:
# we are the last step and the real daemon is actually correctly created at least.
# we have still the last responsibility to write the pid of the daemon itself.
self.write_pid(pid)
os._exit(0) # <-- this was the son, the real daemon is the son-son
self.fpid.close()
del self.fpid
self.pid = os.getpid()
logger.info("Daemonization done: pid=%d" % self.pid)
def do_daemon_init_and_start(self, is_daemon=False):
self.change_to_workdir()
self.check_parallel_run()
# Force the debug level if the daemon is said to start with such level
if self.debug_path:
logger.setLevel('DEBUG')
# If daemon fork() until we reach the final step
if is_daemon:
self.daemonize()
else:
self.write_pid()
except OSError, e:
logger.debug('getMemoryUsage: sysinfo not available')
except Exception, e:
logger.error('getMemoryUsage: exception = %s', traceback.format_exc())
finally:
if int(pythonVersion[1]) >= 6:
try:
proc.kill()
except Exception, e:
logger.debug('Process already terminated')
if physFree == None:
logger.info('getMemoryUsage: sysinfo not installed so falling back on sysctl. sysinfo provides more accurate memory info so is recommended. http://www.freshports.org/sysutils/sysinfo')
try:
try:
logger.debug('getMemoryUsage: attempting Popen (sysctl)')
proc = subprocess.Popen(['sysctl', '-n', 'hw.physmem'], stdout = subprocess.PIPE, close_fds = True)
physTotal = proc.communicate()[0]
if int(pythonVersion[1]) >= 6:
try:
proc.kill()
except Exception, e:
logger.debug('Process already terminated')
logger.debug('getMemoryUsage: attempting Popen (vmstat)')