def adjust_worker_number_by_load(self):
"""Try to create the minimum workers specified in the configuration
:return: None
"""
to_del = []
logger.debug(
"[%s] Trying to adjust worker number."
" Actual number : %d, min per module : %d, max per module : %d",
self.name, len(self.workers), self.min_workers, self.max_workers)
# I want at least min_workers by module then if I can, I add worker for load balancing
for mod in self.q_by_mod:
# At least min_workers
todo = max(0, self.min_workers - len(self.q_by_mod[mod]))
for _ in range(todo):
try:
self.create_and_launch_worker(module_name=mod)
# Maybe this modules is not a true worker one.
# if so, just delete if from q_by_mod
except NotWorkerMod:
to_del.append(mod)
break
for mod in to_del:
logger.debug(
"[%s] The module %s is not a worker one, "
"I remove it from the worker list", self.name, mod)
del self.q_by_mod[mod]
def adjust_worker_number_by_load(self):
"""Try to create the minimum workers specified in the configuration
:return: None
"""
to_del = []
logger.debug("[%s] Trying to adjust worker number."
" Actual number : %d, min per module : %d, max per module : %d",
self.name, len(self.workers), self.min_workers, self.max_workers)
# I want at least min_workers by module then if I can, I add worker for load balancing
for mod in self.q_by_mod:
# At least min_workers
todo = max(0, self.min_workers - len(self.q_by_mod[mod]))
for _ in range(todo):
try:
self.create_and_launch_worker(module_name=mod)
# Maybe this modules is not a true worker one.
# if so, just delete if from q_by_mod
except NotWorkerMod:
to_del.append(mod)
break
for mod in to_del:
logger.debug("[%s] The module %s is not a worker one, "
"I remove it from the worker list", self.name, mod)
del self.q_by_mod[mod]
def ping(self):
"""Send a HTTP request to the satellite (GET /ping)
Add failed attempt if an error occurs
Otherwise, set alive this satellite
:return: None
"""
logger.debug("Pinging %s", self.get_name())
try:
if self.con is None:
self.create_connection()
logger.debug(" (%s)", self.uri)
# If the connection failed to initialize, bail out
if self.con is None:
self.add_failed_check_attempt()
return
res = self.con.get('ping')
# Should return us pong string
if res == 'pong':
self.set_alive()
else:
self.add_failed_check_attempt()
except HTTPEXCEPTIONS, exp:
self.add_failed_check_attempt(reason=str(exp))
def ping(self):
"""Send a HTTP request to the satellite (GET /ping)
Add failed attempt if an error occurs
Otherwise, set alive this satellite
:return: None
"""
logger.debug("Pinging %s", self.get_name())
try:
if self.con is None:
self.create_connection()
logger.debug(" (%s)", self.uri)
# If the connection failed to initialize, bail out
if self.con is None:
self.add_failed_check_attempt()
return
res = self.con.get("ping")
# Should return us pong string
if res == "pong":
self.set_alive()
else:
self.add_failed_check_attempt()
except HTTPEXCEPTIONS, exp:
self.add_failed_check_attempt(reason=str(exp))
def main(self):
"""Main satellite function. Do init and then mainloop
:return: None
"""
try:
for line in self.get_header():
logger.info(line)
self.load_config_file()
# Setting log level
logger.setLevel(self.log_level)
# Force the debug level if the daemon is said to start with such level
if self.debug:
logger.setLevel('DEBUG')
# Look if we are enabled or not. If ok, start the daemon mode
self.look_for_early_exit()
self.do_daemon_init_and_start()
self.do_post_daemon_init()
self.load_modules_manager()
# We wait for initial conf
self.wait_for_initial_conf()
if not self.new_conf: # we must have either big problem or was requested to shutdown
return
self.setup_new_conf()
# We can load our modules now
self.do_load_modules(self.new_modules_conf)
# And even start external ones
self.modules_manager.start_external_instances()
# Allocate Mortal Threads
for _ in xrange(1, self.min_workers):
to_del = []
for mod in self.q_by_mod:
try:
self.create_and_launch_worker(module_name=mod)
# Maybe this modules is not a true worker one.
# if so, just delete if from q_by_mod
except NotWorkerMod:
to_del.append(mod)
for mod in to_del:
logger.debug(
"The module %s is not a worker one, "
"I remove it from the worker list", mod)
del self.q_by_mod[mod]
# Now main loop
self.do_mainloop()
except Exception:
self.print_unrecoverable(traceback.format_exc())
raise
def register(
self,
app,
name,
_type,
api_key="",
secret="",
http_proxy="",
statsd_host="localhost",
statsd_port=8125,
statsd_prefix="alignak",
statsd_enabled=False,
):
"""Init statsd instance with real values
:param app: application (arbiter, scheduler..)
:type app: alignak.daemon.Daemon
:param name: daemon name
:type name: str
:param _type: daemon type
:type _type:
:param api_key: api_key to post data
:type api_key: str
:param secret: secret to post data
:type secret: str
:param http_proxy: proxy http if necessary
:type http_proxy: str
:param statsd_host: host to post data
:type statsd_host: str
:param statsd_port: port to post data
:type statsd_port: int
:param statsd_prefix: prefix to add to metric
:type statsd_prefix: str
:param statsd_enabled: bool to enable statsd
:type statsd_enabled: bool
:return: None
"""
self.app = app
self.name = name
self.type = _type
# kernel.io part
self.api_key = api_key
self.secret = secret
self.http_proxy = http_proxy
# local statsd part
self.statsd_host = statsd_host
self.statsd_port = statsd_port
self.statsd_prefix = statsd_prefix
self.statsd_enabled = statsd_enabled
if self.statsd_enabled:
logger.debug(
"Loading statsd communication with %s:%s.%s", self.statsd_host, self.statsd_port, self.statsd_prefix
)
self.load_statsd()
# Also load the proxy if need
self.con.set_proxy(self.http_proxy)
def main(self):
"""Main satellite function. Do init and then mainloop
:return: None
"""
try:
for line in self.get_header():
logger.info(line)
self.load_config_file()
# Setting log level
logger.setLevel(self.log_level)
# Force the debug level if the daemon is said to start with such level
if self.debug:
logger.setLevel('DEBUG')
# Look if we are enabled or not. If ok, start the daemon mode
self.look_for_early_exit()
self.do_daemon_init_and_start()
self.do_post_daemon_init()
self.load_modules_manager()
# We wait for initial conf
self.wait_for_initial_conf()
if not self.new_conf: # we must have either big problem or was requested to shutdown
return
self.setup_new_conf()
# We can load our modules now
self.modules_manager.set_modules(self.modules_manager.modules)
self.do_load_modules()
# And even start external ones
self.modules_manager.start_external_instances()
# Allocate Mortal Threads
for _ in xrange(1, self.min_workers):
to_del = []
for mod in self.q_by_mod:
try:
self.create_and_launch_worker(module_name=mod)
# Maybe this modules is not a true worker one.
# if so, just delete if from q_by_mod
except NotWorkerMod:
to_del.append(mod)
for mod in to_del:
logger.debug("The module %s is not a worker one, "
"I remove it from the worker list", mod)
del self.q_by_mod[mod]
# Now main loop
self.do_mainloop()
except Exception:
self.print_unrecoverable(traceback.format_exc())
raise
def manage_host_check_result_brok(self, b):
"""
Manage a host check result brok (we UPDATE data info with this)
:param b:
:type b:
:return: None
"""
host_name = b.data['host_name']
logger.debug("[Graphite] host check result: %s", host_name)
# If host initial status brok has not been received, ignore ...
if host_name not in self.hosts_cache:
logger.warning("[Graphite] received service check result for an unknown host: %s",
host_name)
return
# Decode received metrics
couples = self.get_metric_and_value('host_check', b.data['perf_data'])
# If no values, we can exit now
if len(couples) == 0:
logger.debug("[Graphite] no metrics to send ...")
return
# Custom hosts variables
hname = self.illegal_char_hostname.sub('_', host_name)
if '_GRAPHITE_GROUP' in self.hosts_cache[host_name]:
hname = ".".join((self.hosts_cache[host_name]['_GRAPHITE_GROUP'], hname))
if '_GRAPHITE_PRE' in self.hosts_cache[host_name]:
hname = ".".join((self.hosts_cache[host_name]['_GRAPHITE_PRE'], hname))
if self.hostcheck:
hname = '.'.join((hname, self.hostcheck))
# Checks latency
if self.ignore_latency_limit >= b.data['latency'] > 0:
check_time = int(b.data['last_chk']) - int(b.data['latency'])
logger.info("[Graphite] Ignoring latency for service %s. Latency : %s",
b.data['service_description'], b.data['latency'])
else:
check_time = int(b.data['last_chk'])
# Graphite data source
if self.graphite_data_source:
path = '.'.join((hname, self.graphite_data_source))
else:
path = hname
lines = []
# Send a bulk of all metrics at once
for (metric, value) in couples:
lines.append("%s.%s %s %d" % (path, metric, value, check_time))
lines.append("\n")
packet = '\n'.join(lines)
self.send_packet(packet)
def wait_new_conf(self):
"""Ask to scheduler to wait for new conf (HTTP GET from arbiter)
:return: None
"""
with self.app.conf_lock:
logger.debug("Arbiter wants me to wait for a new configuration")
self.app.sched.die()
super(SchedulerInterface, self).wait_new_conf()
def what_i_managed(self):
"""Arbiter ask me which scheduler id I manage
:return: managed configuration ids
:rtype: dict
"""
print "The arbiter asked me what I manage. It's %s", self.app.what_i_managed()
logger.debug("The arbiter asked me what I manage. It's %s", self.app.what_i_managed())
return self.app.what_i_managed()
def wait_new_conf(self):
"""Ask to scheduler to wait for new conf (HTTP GET from arbiter)
:return: None
"""
with self.app.conf_lock:
logger.debug("Arbiter wants me to wait for a new configuration")
self.app.sched.die()
super(SchedulerInterface, self).wait_new_conf()
def register(self,
app,
name,
_type,
api_key='',
secret='',
http_proxy='',
statsd_host='localhost',
statsd_port=8125,
statsd_prefix='alignak',
statsd_enabled=False):
"""Init statsd instance with real values
:param app: application (arbiter, scheduler..)
:type app: alignak.daemon.Daemon
:param name: daemon name
:type name: str
:param _type: daemon type
:type _type:
:param api_key: api_key to post data
:type api_key: str
:param secret: secret to post data
:type secret: str
:param http_proxy: proxy http if necessary
:type http_proxy: str
:param statsd_host: host to post data
:type statsd_host: str
:param statsd_port: port to post data
:type statsd_port: int
:param statsd_prefix: prefix to add to metric
:type statsd_prefix: str
:param statsd_enabled: bool to enable statsd
:type statsd_enabled: bool
:return: None
"""
self.app = app
self.name = name
self.type = _type
# kernel.io part
self.api_key = api_key
self.secret = secret
self.http_proxy = http_proxy
# local statsd part
self.statsd_host = statsd_host
self.statsd_port = statsd_port
self.statsd_prefix = statsd_prefix
self.statsd_enabled = statsd_enabled
if self.statsd_enabled:
logger.debug('Loading statsd communication with %s:%s.%s',
self.statsd_host, self.statsd_port,
self.statsd_prefix)
self.load_statsd()
# Also load the proxy if need
self.con.set_proxy(self.http_proxy)
def wait_new_conf(self):
"""Ask the daemon to drop its configuration and wait for a new one
:return: None
"""
with self.app.conf_lock:
logger.debug("Arbiter wants me to wait for a new configuration")
# Clear can occur while setting up a new conf and lead to error.
self.app.schedulers.clear()
self.app.cur_conf = None
def wait_new_conf(self):
"""Ask the daemon to drop its configuration and wait for a new one
:return: None
"""
with self.app.conf_lock:
logger.debug("Arbiter wants me to wait for a new configuration")
# Clear can occur while setting up a new conf and lead to error.
self.app.schedulers.clear()
self.app.cur_conf = None
def execute_query(self, query):
"""Just run the query
:param query: the query
:type query: str
:return: None
"""
logger.debug("[SqliteDB] Info: I run query '%s'", query)
self.db_cursor.execute(query)
self.db.commit()
def reaper(self):
"""Get data from daemon and send it to the statsd daemon
:return: None
"""
try:
from Crypto.Cipher import AES
except ImportError:
logger.error("Cannot find python lib crypto: stats export is not available")
AES = None # pylint: disable=C0103
while True:
now = int(time.time())
stats = self.stats
self.stats = {}
if len(stats) != 0:
string = ", ".join(["%s:%s" % (key, v) for (key, v) in stats.iteritems()])
# If we are not in an initializer daemon we skip, we cannot have a real name, it sucks
# to find the data after this
if not self.name or not self.api_key or not self.secret:
time.sleep(60)
continue
metrics = []
for (key, elem) in stats.iteritems():
namekey = "%s.%s.%s" % (self.type, self.name, key)
_min, _max, number, _sum = elem
_avg = float(_sum) / number
# nb can't be 0 here and _min_max can't be None too
string = "%s.avg %f %d" % (namekey, _avg, now)
metrics.append(string)
string = "%s.min %f %d" % (namekey, _min, now)
metrics.append(string)
string = "%s.max %f %d" % (namekey, _max, now)
metrics.append(string)
string = "%s.count %f %d" % (namekey, number, now)
metrics.append(string)
# logger.debug('REAPER metrics to send %s (%d)' % (metrics, len(str(metrics))) )
# get the inner data for the daemon
struct = self.app.get_stats_struct()
struct["metrics"].extend(metrics)
# logger.debug('REAPER whole struct %s' % struct)
j = json.dumps(struct)
if AES is not None and self.secret != "":
logger.debug("Stats PUT to kernel.alignak.io/api/v1/put/ with %s %s", self.api_key, self.secret)
# assume a %16 length messagexs
encrypted_text = self._encrypt(j)
try:
self.con.put("/api/v1/put/?api_key=%s" % (self.api_key), encrypted_text)
except HTTPException, exp:
logger.error("Stats REAPER cannot put to the metric server %s", exp)
time.sleep(60)
def declared(function):
""" Decorator to add function in trigger environment
:param function: function to add to trigger environment
:type function: types.FunctionType
:return : the function itself only update TRIGGER_FUNCTIONS variable
"""
name = function.func_name
TRIGGER_FUNCTIONS[name] = function
logger.debug("Added %s to trigger functions list ", name)
return function
def declared(function):
""" Decorator to add function in trigger environment
:param function: function to add to trigger environment
:type function: types.FunctionType
:return : the function itself only update TRIGGER_FUNCTIONS variable
"""
name = function.func_name
TRIGGER_FUNCTIONS[name] = function
logger.debug("Added %s to trigger functions list ", name)
return function
def what_i_managed(self):
"""Arbiter ask me which scheduler id I manage
:return: managed configuration ids
:rtype: dict
"""
print "The arbiter asked me what I manage. It's %s", self.app.what_i_managed(
)
logger.debug("The arbiter asked me what I manage. It's %s",
self.app.what_i_managed())
return self.app.what_i_managed()
def allperfs(obj_ref):
""" Get all perfdatas from a service or a host
:param obj_ref:
:type obj_ref: object
:return: dictionary with perfdatas
:rtype: dict
"""
obj = get_object(obj_ref)
perfdata = PerfDatas(obj.perf_data)
logger.debug("[trigger] I get all perfdatas")
return dict([(metric.name, perfdata[metric.name]) for metric in perfdata])
def allperfs(obj_ref):
""" Get all perfdatas from a service or a host
:param obj_ref:
:type obj_ref: object
:return: dictionary with perfdatas
:rtype: dict
"""
obj = get_object(obj_ref)
perfdata = PerfDatas(obj.perf_data)
logger.debug("[trigger] I get all perfdatas")
return dict([(metric.name, perfdata[metric.name]) for metric in perfdata])
def set_proxy(self, proxy):
"""Set HTTP proxy
:param proxy: proxy url
:type proxy: str
:return: None
"""
if proxy:
logger.debug('PROXY SETTING PROXY %s', proxy)
self._requests_con.proxies = {
'http': proxy,
'https': proxy,
}
def do_pynag_con_init(self, s_id, i_type='scheduler'):
"""Initialize or re-initialize connection with scheduler or arbiter if type == arbiter
:param s_id: s_id
:type s_id: int
:param i_type: type of item
:type i_type: str
:return: None
"""
# Get the good links tab for looping..
links = self.get_links_from_type(i_type)
if links is None:
logger.debug('Type unknown for connection! %s', i_type)
return
# default timeout for daemons like pollers/reactionners/...
timeout = 3
data_timeout = 120
if i_type == 'scheduler':
# If sched is not active, I do not try to init
# it is just useless
is_active = links[s_id]['active']
if not is_active:
return
# schedulers also got real timeout to respect
timeout = links[s_id]['timeout']
data_timeout = links[s_id]['data_timeout']
# If we try to connect too much, we slow down our tests
if self.is_connection_try_too_close(links[s_id]):
return
# Ok, we can now update it
links[s_id]['last_connection'] = time.time()
# DBG: print "Init connection with", links[s_id]['uri']
running_id = links[s_id]['running_id']
# DBG: print "Running id before connection", running_id
uri = links[s_id]['uri']
try:
con = links[s_id]['con'] = HTTPClient(uri=uri,
strong_ssl=links[s_id]['hard_ssl_name_check'],
timeout=timeout, data_timeout=data_timeout)
except HTTPEXCEPTIONS, exp:
# But the multiprocessing module is not compatible with it!
# so we must disable it immediately after
logger.info("Connection problem to the %s %s: %s",
i_type, links[s_id]['name'], str(exp))
links[s_id]['con'] = None
return
def execute_query(self, query):
""" Execute a query against an Oracle database.
:param query: the query
:type query: str
:return: None
"""
logger.debug("[DBOracle] Execute Oracle query %s\n", query)
try:
self.db_cursor.execute(query)
self.db.commit()
except IntegrityError_exp, exp:
logger.warning("[DBOracle] Warning: a query raise an integrity error: %s, %s",
query, exp)
def get_nb_of_must_have_satellites(self, s_type):
"""Generic function to access one of the number satellite attribute
ie : self.nb_pollers, self.nb_reactionners ...
:param s_type: satellite type wanted
:type s_type: str
:return: self.nb_*type*s
:rtype: int
"""
if hasattr(self, 'nb_' + s_type + 's'):
return getattr(self, 'nb_' + s_type + 's')
else:
logger.debug("[realm] do not have this kind of satellites: %s", s_type)
return 0
def get_potential_satellites_by_type(self, s_type):
"""Generic function to access one of the potential satellite attribute
ie : self.potential_pollers, self.potential_reactionners ...
:param s_type: satellite type wanted
:type s_type: str
:return: self.potential_*type*s
:rtype: list
"""
if hasattr(self, 'potential_' + s_type + 's'):
return getattr(self, 'potential_' + s_type + 's')
else:
logger.debug("[realm] do not have this kind of satellites: %s", s_type)
return []
def add(self, elt):
"""Add elt to this broker
Original comment : Schedulers have some queues. We can simplify the call by adding
elements into the proper queue just by looking at their type Brok -> self.broks
TODO: better tag ID?
External commands -> self.external_commands
:param elt: object to add
:type elt: object
:return: None
"""
cls_type = elt.__class__.my_type
if cls_type == 'brok':
# For brok, we TAG brok with our instance_id
elt.instance_id = 0
self.broks_internal_raised.append(elt)
return
elif cls_type == 'externalcommand':
logger.debug("Enqueuing an external command '%s'", str(ExternalCommand.__dict__))
self.external_commands.append(elt)
# Maybe we got a Message from the modules, it's way to ask something
# like from now a full data from a scheduler for example.
elif cls_type == 'message':
# We got a message, great!
logger.debug(str(elt.__dict__))
if elt.get_type() == 'NeedData':
data = elt.get_data()
# Full instance id means: I got no data for this scheduler
# so give me all dumbass!
if 'full_instance_id' in data:
c_id = data['full_instance_id']
source = elt.source
logger.info('The module %s is asking me to get all initial data '
'from the scheduler %d',
source, c_id)
# so we just reset the connection and the running_id,
# it will just get all new things
try:
self.schedulers[c_id]['con'] = None
self.schedulers[c_id]['running_id'] = 0
except KeyError: # maybe this instance was not known, forget it
logger.warning("the module %s ask me a full_instance_id "
"for an unknown ID (%d)!", source, c_id)
# Maybe a module tells me that it's dead, I must log it's last words...
if elt.get_type() == 'ICrash':
data = elt.get_data()
logger.error('the module %s just crash! Please look at the traceback:',
data['name'])
logger.error(data['trace'])
def linkify_hg_by_realms(self, realms):
"""
More than an explode function, but we need to already
have members so... Will be really linkify just after
And we explode realm in ours members, but do not override
a host realm value if it's already set
:param realms: object Realms
:type realms: object
:return: None
"""
# Now we explode the realm value if we've got one
# The group realm must not override a host one (warning?)
for hostgroup in self:
if not hasattr(hostgroup, "realm"):
continue
# Maybe the value is void?
if not hostgroup.realm.strip():
continue
realm = realms.find_by_name(hostgroup.realm.strip())
if realm is not None:
hostgroup.realm = realm
logger.debug("[hostgroups] %s is in %s realm", hostgroup.get_name(), realm.get_name())
else:
err = "the hostgroup %s got an unknown realm '%s'" % (hostgroup.get_name(), hostgroup.realm)
hostgroup.configuration_errors.append(err)
hostgroup.realm = None
continue
for host in hostgroup:
if host is None:
continue
if host.realm is None or host.got_default_realm: # default not hasattr(h, 'realm'):
logger.debug(
"[hostgroups] apply a realm %s to host %s from a hostgroup " "rule (%s)",
hostgroup.realm.get_name(),
host.get_name(),
hostgroup.get_name(),
)
host.realm = hostgroup.realm
else:
if host.realm != hostgroup.realm:
logger.warning(
"[hostgroups] host %s it not in the same realm than it's " "hostgroup %s",
host.get_name(),
hostgroup.get_name(),
)
def get_nb_of_must_have_satellites(self, s_type):
"""Generic function to access one of the number satellite attribute
ie : self.nb_pollers, self.nb_reactionners ...
:param s_type: satellite type wanted
:type s_type: str
:return: self.nb_*type*s
:rtype: int
"""
if hasattr(self, 'nb_' + s_type + 's'):
return getattr(self, 'nb_' + s_type + 's')
else:
logger.debug("[realm] do not have this kind of satellites: %s",
s_type)
return 0
def get_potential_satellites_by_type(self, s_type):
"""Generic function to access one of the potential satellite attribute
ie : self.potential_pollers, self.potential_reactionners ...
:param s_type: satellite type wanted
:type s_type: str
:return: self.potential_*type*s
:rtype: list
"""
if hasattr(self, 'potential_' + s_type + 's'):
return getattr(self, 'potential_' + s_type + 's')
else:
logger.debug("[realm] do not have this kind of satellites: %s",
s_type)
return []
def do_not_run(self):
"""Master tells to slave to not run (HTTP GET)
Master will ignore this call
:return: None
"""
# If I'm the master, ignore the command
if self.app.is_master:
logger.debug("Received message to not run. "
"I am the Master, ignore and continue to run.")
# Else, I'm just a spare, so I listen to my master
else:
logger.debug("Received message to not run. I am the spare, stopping.")
self.app.last_master_speack = time.time()
self.app.must_run = False
def linkify_hg_by_realms(self, realms):
"""
More than an explode function, but we need to already
have members so... Will be really linkify just after
And we explode realm in ours members, but do not override
a host realm value if it's already set
:param realms: object Realms
:type realms: object
:return: None
"""
# Now we explode the realm value if we've got one
# The group realm must not override a host one (warning?)
for hostgroup in self:
if not hasattr(hostgroup, 'realm'):
continue
# Maybe the value is void?
if not hostgroup.realm.strip():
continue
realm = realms.find_by_name(hostgroup.realm.strip())
if realm is not None:
hostgroup.realm = realm
logger.debug("[hostgroups] %s is in %s realm",
hostgroup.get_name(), realm.get_name())
else:
err = "the hostgroup %s got an unknown realm '%s'" % \
(hostgroup.get_name(), hostgroup.realm)
hostgroup.configuration_errors.append(err)
hostgroup.realm = None
continue
for host in hostgroup:
if host is None:
continue
if host.realm is None or host.got_default_realm: # default not hasattr(h, 'realm'):
logger.debug(
"[hostgroups] apply a realm %s to host %s from a hostgroup "
"rule (%s)", hostgroup.realm.get_name(),
host.get_name(), hostgroup.get_name())
host.realm = hostgroup.realm
else:
if host.realm != hostgroup.realm:
logger.warning(
"[hostgroups] host %s it not in the same realm than it's "
"hostgroup %s", host.get_name(),
hostgroup.get_name())
def get_metric_and_value(self, service, perf_data):
"""
Get the metric and the value associated
:param service:
:type service:
:param perf_data:
:type perf_data:
:return:
:rtype; list
"""
result = []
metrics = PerfDatas(perf_data)
for e in metrics:
logger.debug("[Graphite] service: %s, metric: %s", e.name, e.value)
if service in self.filtered_metrics:
if e.name in self.filtered_metrics[service]:
logger.debug("[Graphite] Ignore metric '%s' for filtered service: %s", e.name,
service)
continue
name = self.illegal_char_metric.sub('_', e.name)
name = self.multival.sub(r'.\1', name)
# get metric value and its thresholds values if they exist
name_value = {name: e.value}
# bailout if no value
if name_value[name] == '':
continue
# Get or ignore extra values depending upon module configuration
if e.warning and self.send_warning:
name_value[name + '_warn'] = e.warning
if e.critical and self.send_critical:
name_value[name + '_crit'] = e.critical
if e.min and self.send_min:
name_value[name + '_min'] = e.min
if e.max and self.send_max:
name_value[name + '_max'] = e.max
for key, value in name_value.items():
result.append((key, value))
return result
def do_not_run(self):
"""Master tells to slave to not run (HTTP GET)
Master will ignore this call
:return: None
"""
# If I'm the master, ignore the command
if self.app.is_master:
logger.debug("Received message to not run. "
"I am the Master, ignore and continue to run.")
# Else, I'm just a spare, so I listen to my master
else:
logger.debug(
"Received message to not run. I am the spare, stopping.")
self.app.last_master_speack = time.time()
self.app.must_run = False
def perf(obj_ref, metric_name):
""" Get perf data from a service
:param obj_ref:
:type obj_ref: object
:param metric_name:
:type metric_name: str
:return: None
"""
obj = get_object(obj_ref)
perfdata = PerfDatas(obj.perf_data)
if metric_name in perfdata:
logger.debug("[trigger] I found the perfdata")
return perfdata[metric_name].value
logger.debug("[trigger] I am in perf command")
return None
def perf(obj_ref, metric_name):
""" Get perf data from a service
:param obj_ref:
:type obj_ref: object
:param metric_name:
:type metric_name: str
:return: None
"""
obj = get_object(obj_ref)
perfdata = PerfDatas(obj.perf_data)
if metric_name in perfdata:
logger.debug("[trigger] I found the perfdata")
return perfdata[metric_name].value
logger.debug("[trigger] I am in perf command")
return None
def try_to_restart_deads(self):
"""Try to reinit and restart dead instances
:return: None
"""
to_restart = self.to_restart[:]
del self.to_restart[:]
for inst in to_restart:
logger.debug("I should try to reinit %s", inst.get_name())
if self.try_instance_init(inst):
logger.debug("Good, I try to restart %s", inst.get_name())
# If it's an external, it will start it
inst.start()
# Ok it's good now :)
else:
self.to_restart.append(inst)
def try_to_restart_deads(self):
"""Try to reinit and restart dead instances
:return: None
"""
to_restart = self.to_restart[:]
del self.to_restart[:]
for inst in to_restart:
logger.debug("I should try to reinit %s", inst.get_name())
if self.try_instance_init(inst):
logger.debug("Good, I try to restart %s", inst.get_name())
# If it's an external, it will start it
inst.start()
# Ok it's good now :)
else:
self.to_restart.append(inst)
def add(self, elt):
"""Add an object to the receiver one
Handles brok and externalcommand
:param elt: object to add
:type elt: object
:return: None
"""
cls_type = elt.__class__.my_type
if cls_type == 'brok':
# For brok, we TAG brok with our instance_id
elt.instance_id = 0
self.broks[elt._id] = elt
return
elif cls_type == 'externalcommand':
logger.debug("Enqueuing an external command: %s", str(ExternalCommand.__dict__))
self.unprocessed_external_commands.append(elt)
def remove_instance(self, inst):
"""Request to cleanly remove the given instance.
If instance is external also shutdown it cleanly
:param inst: instance to remove
:type inst: object
:return: None
"""
# External instances need to be close before (process + queues)
if inst.is_external:
logger.debug("Ask stop process for %s", inst.get_name())
inst.stop_process()
logger.debug("Stop process done")
inst.clear_queues(self.manager)
# Then do not listen anymore about it
self.instances.remove(inst)
def manage_initial_host_status_brok(self, b):
"""
Initialize the cache for hosts
:param b:
:type b:
:return: None
"""
host_name = b.data['host_name']
logger.info("[Graphite] got initial host status: %s", host_name)
self.hosts_cache[host_name] = {}
if '_GRAPHITE_PRE' in b.data['customs']:
self.hosts_cache[host_name]['_GRAPHITE_PRE'] = b.data['customs']['_GRAPHITE_PRE']
if '_GRAPHITE_GROUP' in b.data['customs']:
self.hosts_cache[host_name]['_GRAPHITE_GROUP'] = b.data['customs']['_GRAPHITE_GROUP']
logger.debug("[Graphite] initial host status received: %s", host_name)
def add(self, elt):
"""Add an object to the receiver one
Handles brok and externalcommand
:param elt: object to add
:type elt: object
:return: None
"""
cls_type = elt.__class__.my_type
if cls_type == 'brok':
# For brok, we TAG brok with our instance_id
elt.instance_id = 0
self.broks[elt._id] = elt
return
elif cls_type == 'externalcommand':
logger.debug("Enqueuing an external command: %s",
str(ExternalCommand.__dict__))
self.unprocessed_external_commands.append(elt)
def main(self):
"""
Main function where send queue to manage_brok function
:return: None
"""
self.set_proctitle(self.name)
self.set_exit_handler()
while not self.interrupted:
logger.debug("[Alignak Backend Brok] queue length: %s", self.to_q.qsize())
start = time.time()
l = self.to_q.get()
for b in l:
b.prepare()
self.manage_brok(b)
logger.debug("[Alignak Backend Brok] time to manage %s broks (%d secs)", len(l),
time.time() - start)
def remove_instance(self, inst):
"""Request to cleanly remove the given instance.
If instance is external also shutdown it cleanly
:param inst: instance to remove
:type inst: object
:return: None
"""
# External instances need to be close before (process + queues)
if inst.is_external:
logger.debug("Ask stop process for %s", inst.get_name())
inst.stop_process()
logger.debug("Stop process done")
inst.clear_queues(self.manager)
# Then do not listen anymore about it
self.instances.remove(inst)
def add(self, elt):
"""Add an object to the satellite one
Handles brok and externalcommand
:param elt: object to add
:type elt: object
:return: None
"""
cls_type = elt.__class__.my_type
if cls_type == 'brok':
# For brok, we TAG brok with our instance_id
elt.instance_id = 0
self.broks[elt._id] = elt
return
elif cls_type == 'externalcommand':
logger.debug("Enqueuing an external command '%s'", str(elt.__dict__))
with self.external_commands_lock:
self.external_commands.append(elt)
def add(self, elt):
"""Add an object to the satellite one
Handles brok and externalcommand
:param elt: object to add
:type elt: object
:return: None
"""
cls_type = elt.__class__.my_type
if cls_type == 'brok':
# For brok, we TAG brok with our instance_id
elt.instance_id = 0
self.broks[elt._id] = elt
return
elif cls_type == 'externalcommand':
logger.debug("Enqueuing an external command '%s'",
str(elt.__dict__))
with self.external_commands_lock:
self.external_commands.append(elt)
def get_return_for_passive(self, sched_id):
"""Get returns of passive actions for a specific scheduler
:param sched_id: scheduler id
:type sched_id: int
:return: Action list
:rtype: list
"""
# I do not know this scheduler?
sched = self.schedulers.get(sched_id)
if sched is None:
logger.debug("I do not know this scheduler: %s", sched_id)
return []
ret, sched['wait_homerun'] = sched['wait_homerun'], {}
logger.debug("Preparing to return %s results", len(ret))
return ret.values()
def manage_brok(self, brok):
"""Get a brok.
We put brok data to the modules
:param brok: object with data
:type brok: object
:return: None
"""
# Call all modules if they catch the call
for mod in self.modules_manager.get_internal_instances():
try:
mod.manage_brok(brok)
except Exception, exp:
logger.debug(str(exp.__dict__))
logger.warning("The mod %s raise an exception: %s, I'm tagging it to restart later",
mod.get_name(), str(exp))
logger.warning("Exception type: %s", type(exp))
logger.warning("Back trace of this kill: %s", traceback.format_exc())
self.modules_manager.set_to_restart(mod)
def get_outputs(self, out, max_plugins_output_length):
"""Get outputs from single output (split perfdata etc).
Edit output, perf_data and long_output attributes.
:param out: output data of a check
:type out: str
:param max_plugins_output_length: max plugin data length
:type max_plugins_output_length: int
:return: None
"""
# Squeeze all output after max_plugins_output_length
out = out[:max_plugins_output_length]
# manage escaped pipes
out = out.replace(r'\|', '___PROTECT_PIPE___')
# Then cuts by lines
elts = out.split('\n')
# For perf data
elts_line1 = elts[0].split('|')
# First line before | is output, and strip it
self.output = elts_line1[0].strip().replace('___PROTECT_PIPE___', '|')
# Init perfdata as void
self.perf_data = ''
# After | is perfdata, and strip it
if len(elts_line1) > 1:
self.perf_data = elts_line1[1].strip().replace('___PROTECT_PIPE___', '|')
# Now manage others lines. Before the | it's long_output
# And after it's all perf_data, \n join
long_output = []
in_perfdata = False
for line in elts[1:]:
# if already in perfdata, direct append
if in_perfdata:
self.perf_data += ' ' + line.strip().replace('___PROTECT_PIPE___', '|')
else: # not already in? search for the | part :)
elts = line.split('|', 1)
# The first part will always be long_output
long_output.append(elts[0].strip().replace('___PROTECT_PIPE___', '|'))
if len(elts) > 1:
in_perfdata = True
self.perf_data += ' ' + elts[1].strip().replace('___PROTECT_PIPE___', '|')
# long_output is all non output and perfline, join with \n
self.long_output = '\n'.join(long_output)
logger.debug("Command result for '%s': %s", self.command, self.output)
def get_return_for_passive(self, sched_id):
"""Get returns of passive actions for a specific scheduler
:param sched_id: scheduler id
:type sched_id: int
:return: Action list
:rtype: list
"""
# I do not know this scheduler?
sched = self.schedulers.get(sched_id)
if sched is None:
logger.debug("I do not know this scheduler: %s", sched_id)
return []
ret, sched['wait_homerun'] = sched['wait_homerun'], {}
logger.debug("Preparing to return %s results", len(ret))
return ret.values()
def get_all_states(self):
"""Return all the data of satellites
:return: dict containing satellites data
Output looks like this ::
{'arbiter' : [{'schedproperty1':'value1' ..}, {'pollerproperty1', 'value11' ..}, ..],
'scheduler': [..],
'poller': [..],
'reactionner': [..],
'receiver': [..],
'broker: [..]'
}
:rtype: dict
"""
res = {}
for s_type in [
'arbiter', 'scheduler', 'poller', 'reactionner', 'receiver',
'broker'
]:
lst = []
res[s_type] = lst
for daemon in getattr(self.app.conf, s_type + 's'):
cls = daemon.__class__
env = {}
all_props = [cls.properties, cls.running_properties]
for props in all_props:
for prop in props:
if hasattr(daemon, prop):
val = getattr(daemon, prop)
if prop == "realm":
if hasattr(val, "realm_name"):
env[prop] = val.realm_name
# give a try to a json able object
try:
json.dumps(val)
env[prop] = val
except Exception, exp:
logger.debug('%s', exp)
lst.append(env)
def get_objects_properties(self, table):
"""'Dump all objects of the type in
[hosts, services, contacts, commands, hostgroups, servicegroups]
:param table: table name
:type table: str
:return: list all properties of all objects
:rtype: list
"""
with self.app.conf_lock:
logger.debug('ASK:: table= %s', str(table))
objs = getattr(self.app.conf, table, None)
logger.debug("OBJS:: %s", str(objs))
if objs is None or len(objs) == 0:
return []
res = []
for obj in objs:
j_obj = jsonify_r(obj)
res.append(j_obj)
return res
def put_results(self, results):
"""Put results to scheduler, used by poller and reactionners
:param results: results to handle
:type results:
:return: True or ?? (if lock acquire fails)
:rtype: bool
"""
nb_received = len(results)
self.app.sched.nb_check_received += nb_received
if nb_received != 0:
logger.debug("Received %d results", nb_received)
for result in results:
result.set_type_active()
with self.app.sched.waiting_results_lock:
self.app.sched.waiting_results.extend(results)
# for c in results:
# self.sched.put_results(c)
return True
def __setstate_deprecated__(self, state):
"""In 1.0 we move to a dict save.
:param state: it's the state
:type state: dict
:return: None
"""
cls = self.__class__
# Check if the len of this state is like the previous,
# if not, we will do errors!
# -1 because of the '_id' prop
if len(cls.properties) != (len(state) - 1):
logger.debug("Passing comment")
return
self._id = state.pop()
for prop in cls.properties:
val = state.pop()
setattr(self, prop, val)
if self._id >= cls._id:
cls._id = self._id + 1
def execute(self):
"""Start this action command. The command will be executed in a
subprocess.
:return: None or str 'toomanyopenfiles'
:rtype: None | str
"""
self.status = 'launched'
self.check_time = time.time()
self.wait_time = 0.0001
self.last_poll = self.check_time
# Get a local env variables with our additional values
self.local_env = self.get_local_environnement()
# Initialize stdout and stderr. we will read them in small parts
# if the fcntl is available
self.stdoutdata = ''
self.stderrdata = ''
logger.debug("Launch command: %s", self.command)
return self.execute__() # OS specific part