Source code for shinken.daemons.brokerdaemon

#!/usr/bin/python
# -*- coding: utf-8 -*-

# Copyright (C) 2009-2014:
#    Gabes Jean, naparuba@gmail.com
#    Gerhard Lausser, Gerhard.Lausser@consol.de
#    Gregory Starck, g.starck@gmail.com
#    Hartmut Goebel, h.goebel@goebel-consult.de
#
# This file is part of Shinken.
#
# Shinken is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# Shinken is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with Shinken.  If not, see <http://www.gnu.org/licenses/>.

import os
import sys
import time
import traceback
import cPickle
import base64
import zlib
import threading
from multiprocessing import active_children

from shinken.satellite import BaseSatellite
from shinken.property import PathProp, IntegerProp
from shinken.util import sort_by_ids
from shinken.log import logger
from shinken.stats import statsmgr
from shinken.external_command import ExternalCommand
from shinken.http_client import HTTPClient, HTTPExceptions
from shinken.daemon import Daemon, Interface

[docs]class IStats(Interface): """ Interface for various stats about broker activity """ doc = 'Get raw stats from the daemon'
[docs] def get_raw_stats(self): app = self.app res = [] insts = [inst for inst in app.modules_manager.instances if inst.is_external] for inst in insts: try: res.append({'module_name': inst.get_name(), 'queue_size': inst.to_q.qsize()}) except Exception, exp: res.append({'module_name': inst.get_name(), 'queue_size': 0}) return res
get_raw_stats.doc = doc # Our main APP class
[docs]class Broker(BaseSatellite): properties = BaseSatellite.properties.copy() properties.update({ 'pidfile': PathProp(default='brokerd.pid'), 'port': IntegerProp(default=7772), 'local_log': PathProp(default='brokerd.log'), }) def __init__(self, config_file, is_daemon, do_replace, debug, debug_file, profile=''): super(Broker, self).__init__('broker', config_file, is_daemon, do_replace, debug, debug_file) # Our arbiters self.arbiters = {} # Our pollers, reactionners and receivers self.pollers = {} self.reactionners = {} self.receivers = {} # Modules are load one time self.have_modules = False # Can have a queue of external_commands given by modules # will be processed by arbiter self.external_commands = [] # All broks to manage self.broks = [] # broks to manage # broks raised this turn and that needs to be put in self.broks self.broks_internal_raised = [] # broks raised by the arbiters, we need a lock so the push can be in parallel # to our current activities and won't lock the arbiter self.arbiter_broks = [] self.arbiter_broks_lock = threading.RLock() self.timeout = 1.0 self.istats = IStats(self) # 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
[docs] def add(self, elt): 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']) # The module death will be looked for elsewhere and restarted. # Get the good tabs for links by the kind. If unknown, return None
[docs] def is_connection_try_too_close(self, elt): now = time.time() last_connection = elt['last_connection'] if now - last_connection < 5: return True return False # wrapper function for the real function do_ # just for timing the connection
[docs] def pynag_con_init(self, id, type='scheduler'): _t = time.time() r = self.do_pynag_con_init(id, type) statsmgr.incr('con-init.%s' % type, time.time() - _t) return r # initialize or re-initialize connection with scheduler or # arbiter if type == arbiter
[docs] def do_pynag_con_init(self, id, type='scheduler'): # Get the good links tab for looping.. links = self.get_links_from_type(type) if links is None: logger.debug('Type unknown for connection! %s', type) return # default timeout for daemons like pollers/reactionners/... timeout = 3 data_timeout = 120 if type == 'scheduler': # If sched is not active, I do not try to init # it is just useless is_active = links[id]['active'] if not is_active: return # schedulers also got real timeout to respect timeout = links[id]['timeout'] data_timeout = links[id]['data_timeout'] # If we try to connect too much, we slow down our tests if self.is_connection_try_too_close(links[id]): return # Ok, we can now update it links[id]['last_connection'] = time.time() # DBG: print "Init connection with", links[id]['uri'] running_id = links[id]['running_id'] # DBG: print "Running id before connection", running_id uri = links[id]['uri'] try: con = links[id]['con'] = HTTPClient(uri=uri, strong_ssl=links[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", type, links[id]['name'], str(exp)) links[id]['con'] = None return try: # initial ping must be quick con.get('ping') new_run_id = con.get('get_running_id') new_run_id = float(new_run_id) # data transfer can be longer # The schedulers have been restarted: it has a new run_id. # So we clear all verifs, they are obsolete now. if new_run_id != running_id: logger.debug("[%s] New running id for the %s %s: %s (was %s)", self.name, type, links[id]['name'], new_run_id, running_id) links[id]['broks'].clear() # we must ask for a new full broks if # it's a scheduler if type == 'scheduler': logger.debug("[%s] I ask for a broks generation to the scheduler %s", self.name, links[id]['name']) con.get('fill_initial_broks', {'bname': self.name}, wait='long') # Ok all is done, we can save this new running id links[id]['running_id'] = new_run_id except HTTPExceptions, exp: logger.info("Connection problem to the %s %s: %s", type, links[id]['name'], str(exp)) links[id]['con'] = None return except KeyError, exp: logger.info("the %s '%s' is not initialized: %s", type, links[id]['name'], str(exp)) links[id]['con'] = None traceback.print_stack() return logger.info("Connection OK to the %s %s", type, links[id]['name']) # Get a brok. Our role is to put it in the modules # DO NOT CHANGE data of b!!! # REF: doc/broker-modules.png (4-5)
[docs] def manage_brok(self, b): # Call all modules if they catch the call for mod in self.modules_manager.get_internal_instances(): try: mod.manage_brok(b) 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) # Add broks (a tab) to different queues for # internal and external modules
[docs] def add_broks_to_queue(self, broks): # Ok now put in queue broks to be managed by # internal modules self.broks.extend(broks) # Each turn we get all broks from # self.broks_internal_raised and we put them in # self.broks
[docs] def interger_internal_broks(self): self.add_broks_to_queue(self.broks_internal_raised) self.broks_internal_raised = [] # We will get in the broks list the broks from the arbiters, # but as the arbiter_broks list can be push by arbiter without Global lock, # we must protect this with he list lock
[docs] def interger_arbiter_broks(self): with self.arbiter_broks_lock: self.add_broks_to_queue(self.arbiter_broks) self.arbiter_broks = [] # We get new broks from schedulers # REF: doc/broker-modules.png (2)
[docs] def get_new_broks(self, type='scheduler'): # Get the good links tab for looping.. links = self.get_links_from_type(type) if links is None: logger.debug('Type unknown for connection! %s', type) return # We check for new check in each schedulers and put # the result in new_checks for sched_id in links: try: con = links[sched_id]['con'] if con is not None: # None = not initialized t0 = time.time() # Before ask a call that can be long, do a simple ping to be sure it is alive con.get('ping') tmp_broks = con.get('get_broks', {'bname': self.name}, wait='long') try: _t = base64.b64decode(tmp_broks) _t = zlib.decompress(_t) tmp_broks = cPickle.loads(_t) except (TypeError, zlib.error, cPickle.PickleError), exp: logger.error('Cannot load broks data from %s : %s', links[sched_id]['name'], exp) links[sched_id]['con'] = None continue logger.debug("%s Broks get in %s", len(tmp_broks), time.time() - t0) for b in tmp_broks.values(): b.instance_id = links[sched_id]['instance_id'] # Ok, we can add theses broks to our queues self.add_broks_to_queue(tmp_broks.values()) else: # no con? make the connection self.pynag_con_init(sched_id, type=type) # Ok, con is not known, so we create it except KeyError, exp: logger.debug("Key error for get_broks : %s", str(exp)) self.pynag_con_init(sched_id, type=type) except HTTPExceptions, exp: logger.warning("Connection problem to the %s %s: %s", type, links[sched_id]['name'], str(exp)) links[sched_id]['con'] = None # scheduler must not #be initialized except AttributeError, exp: logger.warning("The %s %s should not be initialized: %s", type, links[sched_id]['name'], str(exp)) # scheduler must not have checks # What the F**k? We do not know what happened, # so.. bye bye :) except Exception, x: logger.error(str(x)) logger.error(traceback.format_exc()) sys.exit(1) # Helper function for module, will give our broks
[docs] def get_retention_data(self): return self.broks # Get back our broks from a retention module
[docs] def restore_retention_data(self, data): self.broks.extend(data)
[docs] def do_stop(self): act = active_children() for a in act: a.terminate() a.join(1) super(Broker, self).do_stop()
[docs] def setup_new_conf(self): conf = self.new_conf self.new_conf = None self.cur_conf = conf # Got our name from the globals g_conf = conf['global'] if 'broker_name' in g_conf: name = g_conf['broker_name'] else: name = 'Unnamed broker' self.name = name self.api_key = g_conf['api_key'] self.secret = g_conf['secret'] self.http_proxy = g_conf['http_proxy'] self.statsd_host = g_conf['statsd_host'] self.statsd_port = g_conf['statsd_port'] self.statsd_prefix = g_conf['statsd_prefix'] self.statsd_enabled = g_conf['statsd_enabled'] # We got a name so we can update the logger and the stats global objects logger.load_obj(self, name) statsmgr.register(self, name, 'broker', api_key=self.api_key, secret=self.secret, http_proxy=self.http_proxy, statsd_host=self.statsd_host, statsd_port=self.statsd_port, statsd_prefix=self.statsd_prefix, statsd_enabled=self.statsd_enabled) logger.debug("[%s] Sending us configuration %s", self.name, conf) # If we've got something in the schedulers, we do not # want it anymore # self.schedulers.clear() for sched_id in conf['schedulers']: # Must look if we already have it to do not overdie our broks already_got = False # We can already got this conf id, but with another address if sched_id in self.schedulers: new_addr = conf['schedulers'][sched_id]['address'] old_addr = self.schedulers[sched_id]['address'] new_port = conf['schedulers'][sched_id]['port'] old_port = self.schedulers[sched_id]['port'] # Should got all the same to be ok :) if new_addr == old_addr and new_port == old_port: already_got = True if already_got: broks = self.schedulers[sched_id]['broks'] running_id = self.schedulers[sched_id]['running_id'] else: broks = {} running_id = 0 s = conf['schedulers'][sched_id] self.schedulers[sched_id] = s # replacing scheduler address and port by those defined in satellitemap if s['name'] in g_conf['satellitemap']: s = dict(s) # make a copy s.update(g_conf['satellitemap'][s['name']]) proto = 'http' if s['use_ssl']: proto = 'https' uri = '%s://%s:%s/' % (proto, s['address'], s['port']) self.schedulers[sched_id]['uri'] = uri self.schedulers[sched_id]['broks'] = broks self.schedulers[sched_id]['instance_id'] = s['instance_id'] self.schedulers[sched_id]['running_id'] = running_id self.schedulers[sched_id]['active'] = s['active'] self.schedulers[sched_id]['last_connection'] = 0 self.schedulers[sched_id]['timeout'] = s['timeout'] self.schedulers[sched_id]['data_timeout'] = s['data_timeout'] logger.info("We have our schedulers: %s ", self.schedulers) # Now get arbiter for arb_id in conf['arbiters']: # Must look if we already have it already_got = arb_id in self.arbiters if already_got: broks = self.arbiters[arb_id]['broks'] else: broks = {} a = conf['arbiters'][arb_id] self.arbiters[arb_id] = a # replacing arbiter address and port by those defined in satellitemap if a['name'] in g_conf['satellitemap']: a = dict(a) # make a copy a.update(g_conf['satellitemap'][a['name']]) proto = 'http' if a['use_ssl']: proto = 'https' uri = '%s://%s:%s/' % (proto, a['address'], a['port']) self.arbiters[arb_id]['uri'] = uri self.arbiters[arb_id]['broks'] = broks self.arbiters[arb_id]['instance_id'] = 0 # No use so all to 0 self.arbiters[arb_id]['running_id'] = 0 self.arbiters[arb_id]['last_connection'] = 0 # We do not connect to the arbiter. Connection hangs logger.info("We have our arbiters: %s ", self.arbiters) # Now for pollers for pol_id in conf['pollers']: # Must look if we already have it already_got = pol_id in self.pollers if already_got: broks = self.pollers[pol_id]['broks'] running_id = self.schedulers[sched_id]['running_id'] else: broks = {} running_id = 0 p = conf['pollers'][pol_id] self.pollers[pol_id] = p # replacing poller address and port by those defined in satellitemap if p['name'] in g_conf['satellitemap']: p = dict(p) # make a copy p.update(g_conf['satellitemap'][p['name']]) proto = 'http' if p['use_ssl']: proto = 'https' uri = '%s://%s:%s/' % (proto, p['address'], p['port']) self.pollers[pol_id]['uri'] = uri self.pollers[pol_id]['broks'] = broks self.pollers[pol_id]['instance_id'] = 0 # No use so all to 0 self.pollers[pol_id]['running_id'] = running_id self.pollers[pol_id]['last_connection'] = 0 # #And we connect to it # self.app.pynag_con_init(pol_id, 'poller') logger.info("We have our pollers: %s", self.pollers) # Now reactionners for rea_id in conf['reactionners']: # Must look if we already have it already_got = rea_id in self.reactionners if already_got: broks = self.reactionners[rea_id]['broks'] running_id = self.schedulers[sched_id]['running_id'] else: broks = {} running_id = 0 r = conf['reactionners'][rea_id] self.reactionners[rea_id] = r # replacing reactionner address and port by those defined in satellitemap if r['name'] in g_conf['satellitemap']: r = dict(r) # make a copy r.update(g_conf['satellitemap'][r['name']]) proto = 'http' if r['use_ssl']: proto = 'https' uri = '%s://%s:%s/' % (proto, r['address'], r['port']) self.reactionners[rea_id]['uri'] = uri self.reactionners[rea_id]['broks'] = broks self.reactionners[rea_id]['instance_id'] = 0 # No use so all to 0 self.reactionners[rea_id]['running_id'] = running_id self.reactionners[rea_id]['last_connection'] = 0 # #And we connect to it # self.app.pynag_con_init(rea_id, 'reactionner') logger.info("We have our reactionners: %s", self.reactionners) # Now receivers for rec_id in conf['receivers']: # Must look if we already have it already_got = rec_id in self.receivers if already_got: broks = self.receivers[rec_id]['broks'] running_id = self.schedulers[sched_id]['running_id'] else: broks = {} running_id = 0 r = conf['receivers'][rec_id] self.receivers[rec_id] = r # replacing reactionner address and port by those defined in satellitemap if r['name'] in g_conf['satellitemap']: r = dict(r) # make a copy r.update(g_conf['satellitemap'][r['name']]) proto = 'http' if r['use_ssl']: proto = 'https' uri = '%s://%s:%s/' % (proto, r['address'], r['port']) self.receivers[rec_id]['uri'] = uri self.receivers[rec_id]['broks'] = broks self.receivers[rec_id]['instance_id'] = 0 # No use so all to 0 self.receivers[rec_id]['running_id'] = running_id self.receivers[rec_id]['last_connection'] = 0 if not self.have_modules: self.modules = mods = conf['global']['modules'] self.have_modules = True logger.info("We received modules %s ", mods) # Ok now start, or restart them! # Set modules, init them and start external ones self.modules_manager.set_modules(self.modules) self.do_load_modules() self.modules_manager.start_external_instances() # Set our giving timezone from arbiter use_timezone = conf['global']['use_timezone'] if use_timezone != 'NOTSET': logger.info("Setting our timezone to %s", use_timezone) os.environ['TZ'] = use_timezone time.tzset() # Connection init with Schedulers for sched_id in self.schedulers: self.pynag_con_init(sched_id, type='scheduler') for pol_id in self.pollers: self.pynag_con_init(pol_id, type='poller') for rea_id in self.reactionners: self.pynag_con_init(rea_id, type='reactionner') # An arbiter ask us to wait for a new conf, so we must clean # all our mess we did, and close modules too
[docs] def clean_previous_run(self): # Clean all lists self.schedulers.clear() self.pollers.clear() self.reactionners.clear() self.broks = self.broks[:] self.broks_internal_raised = self.broks_internal_raised[:] with self.arbiter_broks_lock: self.arbiter_broks = self.arbiter_broks[:] self.external_commands = self.external_commands[:] # And now modules self.have_modules = False self.modules_manager.clear_instances() # stats threads is asking us a main structure for stats
[docs] def get_stats_struct(self): now = int(time.time()) # call the daemon one res = super(Broker, self).get_stats_struct() res.update({'name': self.name, 'type': 'broker'}) metrics = res['metrics'] # metrics specific metrics.append('broker.%s.external-commands.queue %d %d' % ( self.name, len(self.external_commands), now)) metrics.append('broker.%s.broks.queue %d %d' % (self.name, len(self.broks), now)) return res
[docs] def do_loop_turn(self): logger.debug("Begin Loop: managing old broks (%d)", len(self.broks)) # Dump modules Queues size insts = [inst for inst in self.modules_manager.instances if inst.is_external] for inst in insts: try: logger.debug("External Queue len (%s): %s", inst.get_name(), inst.to_q.qsize()) except Exception, exp: logger.debug("External Queue len (%s): Exception! %s", inst.get_name(), exp) # Begin to clean modules self.check_and_del_zombie_modules() # Maybe the arbiter ask us to wait for a new conf # If true, we must restart all... if self.cur_conf is None: # Clean previous run from useless objects # and close modules self.clean_previous_run() self.wait_for_initial_conf() # we may have been interrupted or so; then # just return from this loop turn if not self.new_conf: return self.setup_new_conf() # Now we check if arbiter speak to us in the pyro_daemon. # If so, we listen for it # When it pushes conf to us, we reinit connections self.watch_for_new_conf(0.0) if self.new_conf: self.setup_new_conf() # Maybe the last loop we raised some broks internally # we should integrate them in broks self.interger_internal_broks() # Also reap broks sent from the arbiters self.interger_arbiter_broks() # Main job, go get broks in our distants daemons types = ['scheduler', 'poller', 'reactionner', 'receiver'] for _type in types: _t = time.time() # And from schedulers self.get_new_broks(type=_type) statsmgr.incr('get-new-broks.%s' % _type, time.time() - _t) # Sort the brok list by id self.broks.sort(sort_by_ids) # and for external queues # REF: doc/broker-modules.png (3) # We put to external queues broks that was not already send t0 = time.time() # We are sending broks as a big list, more efficient than one by one ext_modules = self.modules_manager.get_external_instances() to_send = [b for b in self.broks if getattr(b, 'need_send_to_ext', True)] # Send our pack to all external modules to_q queue so they can get the wole packet # beware, the sub-process/queue can be die/close, so we put to restart the whole module # instead of killing ourself :) for mod in ext_modules: try: mod.to_q.put(to_send) except Exception, exp: # first we must find the modules logger.debug(str(exp.__dict__)) logger.warning("The mod %s queue 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) # No more need to send them for b in to_send: b.need_send_to_ext = False statsmgr.incr('core.put-to-external-queue', time.time() - t0) logger.debug("Time to send %s broks (%d secs)", len(to_send), time.time() - t0) # We must had new broks at the end of the list, so we reverse the list self.broks.reverse() start = time.time() while len(self.broks) != 0: now = time.time() # Do not 'manage' more than 1s, we must get new broks # every 1s if now - start > 1: break b = self.broks.pop() # Ok, we can get the brok, and doing something with it # REF: doc/broker-modules.png (4-5) # We un serialize the brok before consume it b.prepare() _t = time.time() self.manage_brok(b) statsmgr.incr('core.manage-brok', time.time() - _t) nb_broks = len(self.broks) # Ok we manage brok, but we still want to listen to arbiter self.watch_for_new_conf(0.0) # if we got new broks here from arbiter, we should break the loop # because such broks will not be managed by the # external modules before this loop (we pop them!) if len(self.broks) != nb_broks: break # Maybe external modules raised 'objects' # we should get them self.get_objects_from_from_queues() # Maybe we do not have something to do, so we wait a little # TODO: redone the diff management.... if len(self.broks) == 0: while self.timeout > 0: begin = time.time() self.watch_for_new_conf(1.0) end = time.time() self.timeout = self.timeout - (end - begin) self.timeout = 1.0 # print "get new broks watch new conf 1: end", len(self.broks) # Say to modules it's a new tick :) self.hook_point('tick') # Main function, will loop forever
[docs] def main(self): try: 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') for line in self.get_header(): logger.info(line) logger.info("[Broker] Using working directory: %s", os.path.abspath(self.workdir)) # 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.load_modules_manager() self.uri2 = self.http_daemon.register(self.interface) logger.debug("The Arbiter uri it at %s", self.uri2) self.uri3 = self.http_daemon.register(self.istats) # We wait for initial conf self.wait_for_initial_conf() if not self.new_conf: return self.setup_new_conf() # Do the modules part, we have our modules in self.modules # REF: doc/broker-modules.png (1) self.hook_point('load_retention') # Now the main loop self.do_mainloop() except Exception, exp: self.print_unrecoverable(traceback.format_exc()) raise