class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default='7768'),
'local_log': PathProp(default='schedulerd.log'),
})
# Create the shinken class:
# Create a Pyro server (port = arvg 1)
# then create the interface for arbiter
# Then, it wait for a first configuration
def __init__(self, config_file, is_daemon, do_replace, debug, debug_file, profile=''):
BaseSatellite.__init__(self, 'scheduler', config_file, is_daemon, do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.istats = IStats(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
self.brokers = {}
def do_stop(self):
if self.http_daemon:
if self.ibroks:
self.http_daemon.unregister(self.ibroks)
if self.ichecks:
self.http_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.warning("A system time change of %d has been detected. Compensating..." % difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
if not hasattr(self.sched, "conf"):
# Race condition where time change before getting conf
return
# Then we compensate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == 'scheduled' and c.t_to_go is not None:
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
if ref.check_period is not None:
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == 'notification':
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
logger.warning("Received a SIGNAL %s" % sig)
# If we got USR1, just dump memory
if sig == signal.SIGUSR1:
self.sched.need_dump_memory = True
elif sig == signal.SIGUSR2: #usr2, dump objects
self.sched.need_objects_dump = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
logger.info("New configuration received")
self.setup_new_conf()
logger.info("New configuration loaded")
self.sched.run()
def setup_new_conf(self):
pk = self.new_conf
conf_raw = pk['conf']
override_conf = pk['override_conf']
modules = pk['modules']
satellites = pk['satellites']
instance_name = pk['instance_name']
push_flavor = pk['push_flavor']
skip_initial_broks = pk['skip_initial_broks']
accept_passive_unknown_check_results = pk['accept_passive_unknown_check_results']
# horay, we got a name, we can set it in our stats objects
statsmgr.register(instance_name, 'scheduler')
t0 = time.time()
conf = cPickle.loads(conf_raw)
logger.debug("Conf received at %d. Unserialized in %d secs" % (t0, time.time() - t0))
self.new_conf = None
# Tag the conf with our data
self.conf = conf
self.conf.push_flavor = push_flavor
self.conf.instance_name = instance_name
self.conf.skip_initial_broks = skip_initial_broks
self.conf.accept_passive_unknown_check_results = accept_passive_unknown_check_results
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
#self.pollers = self.app.pollers
if self.conf.human_timestamp_log:
logger.set_human_format()
# Now We create our pollers
for pol_id in satellites['pollers']:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites['pollers'][pol_id]
self.pollers[pol_id] = p
if p['name'] in override_conf['satellitemap']:
p = dict(p) # make a copy
p.update(override_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]['last_connection'] = 0
# First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
#print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != '':
logger.debug("Setting our timezone to %s" % str(self.conf.use_timezone))
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
if len(self.modules) != 0:
logger.debug("I've got %s modules" % str(self.modules))
# TODO: if scheduler had previous modules instanciated it must clean them!
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
logger.debug("Deconnecting previous Check Interface")
self.http_daemon.unregister(self.ichecks)
# Now create and connect it
self.ichecks = IChecks(self.sched)
self.http_daemon.register(self.ichecks)
logger.debug("The Scheduler Interface uri is: %s" % self.uri)
# Same for Broks
if self.ibroks is not None:
logger.debug("Deconnecting previous Broks Interface")
self.http_daemon.unregister(self.ibroks)
# Create and connect it
self.ibroks = IBroks(self.sched)
self.http_daemon.register(self.ibroks)
logger.info("Loading configuration.")
self.conf.explode_global_conf()
# we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
# We must update our Config dict macro with good value
# from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
#print "DBG: got macros", self.sched.conf.macros
# Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
#self.conf.dump()
#self.conf.quick_debug()
# Now create the external commander
# it's a applyer: it role is not to dispatch commands,
# but to apply them
e = ExternalCommandManager(self.conf, 'applyer')
# Scheduler need to know about external command to
# activate it if necessary
self.sched.load_external_command(e)
# External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# We clear our schedulers managed (it's us :) )
# and set ourself in it
self.schedulers = {self.conf.instance_id: self.sched}
# Give the arbiter the data about what I manage
# for me it's just my instance_id and my push flavor
def what_i_managed(self):
if hasattr(self, 'conf'):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
# our main function, launch after the init
def main(self):
try:
self.load_config_file()
self.look_for_early_exit()
self.do_daemon_init_and_start()
self.load_modules_manager()
self.http_daemon.register(self.interface)
self.http_daemon.register(self.istats)
#self.inject = Injector(self.sched)
#self.http_daemon.register(self.inject)
self.http_daemon.unregister(self.interface)
self.uri = self.http_daemon.uri
logger.info("[scheduler] General interface is at: %s" % self.uri)
self.do_mainloop()
except Exception, exp:
logger.critical("I got an unrecoverable error. I have to exit")
logger.critical("You can log a bug ticket at https://github.com/naparuba/shinken/issues/new to get help")
logger.critical("Back trace of it: %s" % (traceback.format_exc()))
raise
class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default='7768'),
'local_log': PathProp(default='schedulerd.log'),
})
# Create the shinken class:
# Create a Pyro server (port = arvg 1)
# then create the interface for arbiter
# Then, it wait for a first configuration
def __init__(self, config_file, is_daemon, do_replace, debug, debug_file):
BaseSatellite.__init__(self, 'scheduler', config_file, is_daemon,
do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
self.brokers = {}
def do_stop(self):
if self.pyro_daemon:
if self.ibroks:
self.pyro_daemon.unregister(self.ibroks)
if self.ichecks:
self.pyro_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.warning(
"A system time change of %d has been detected. Compensating..." %
difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
# Then we compensate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
if ref.check_period is not None:
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == 'notification':
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(
new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
logger.warning("Received a SIGNAL %s" % sig)
# If we got USR1, just dump memory
if sig == 10:
self.sched.need_dump_memory = True
elif sig == 12: #usr2, dump objects
self.sched.need_objects_dump = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
logger.info("New configuration received")
self.setup_new_conf()
logger.info("New configuration loaded")
self.sched.run()
def setup_new_conf(self):
pk = self.new_conf
conf_raw = pk['conf']
override_conf = pk['override_conf']
modules = pk['modules']
satellites = pk['satellites']
instance_name = pk['instance_name']
push_flavor = pk['push_flavor']
skip_initial_broks = pk['skip_initial_broks']
t0 = time.time()
conf = cPickle.loads(conf_raw)
logger.debug("Conf received at %d. Unserialized in %d secs" %
(t0, time.time() - t0))
self.new_conf = None
# Tag the conf with our data
self.conf = conf
self.conf.push_flavor = push_flavor
self.conf.instance_name = instance_name
self.conf.skip_initial_broks = skip_initial_broks
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
#self.pollers = self.app.pollers
if self.conf.human_timestamp_log:
logger.set_human_format()
# Now We create our pollers
for pol_id in satellites['pollers']:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites['pollers'][pol_id]
self.pollers[pol_id] = p
if p['name'] in override_conf['satellitemap']:
p = dict(p) # make a copy
p.update(override_conf['satellitemap'][p['name']])
uri = pyro.create_uri(p['address'], p['port'], 'Schedulers',
self.use_ssl)
self.pollers[pol_id]['uri'] = uri
self.pollers[pol_id]['last_connection'] = 0
# First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
#print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != '':
logger.debug("Setting our timezone to %s" %
str(self.conf.use_timezone))
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
if len(self.modules) != 0:
logger.debug("I've got %s modules" % str(self.modules))
# TODO: if scheduler had previous modules instanciated it must clean them!
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
logger.debug(
"Deconnecting previous Check Interface from pyro_daemon")
self.pyro_daemon.unregister(self.ichecks)
# Now create and connect it
self.ichecks = IChecks(self.sched)
self.uri = self.pyro_daemon.register(self.ichecks, "Checks")
logger.debug("The Checks Interface uri is: %s" % self.uri)
# Same for Broks
if self.ibroks is not None:
logger.debug(
"Deconnecting previous Broks Interface from pyro_daemon")
self.pyro_daemon.unregister(self.ibroks)
# Create and connect it
self.ibroks = IBroks(self.sched)
self.uri2 = self.pyro_daemon.register(self.ibroks, "Broks")
logger.debug("The Broks Interface uri is: %s" % self.uri2)
logger.info("Loading configuration.")
self.conf.explode_global_conf()
# we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
# We must update our Config dict macro with good value
# from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
#print "DBG: got macros", self.sched.conf.macros
# Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
#self.conf.dump()
#self.conf.quick_debug()
# Now create the external commander
# it's a applyer: it role is not to dispatch commands,
# but to apply them
e = ExternalCommandManager(self.conf, 'applyer')
# Scheduler need to know about external command to
# activate it if necessary
self.sched.load_external_command(e)
# External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# We clear our schedulers managed (it's us :) )
# and set ourself in it
self.schedulers = {self.conf.instance_id: self.sched}
# Give the arbiter the data about what I manage
# for me it's just my instance_id and my push flavor
def what_i_managed(self):
if hasattr(self, 'conf'):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
# our main function, launch after the init
def main(self):
try:
self.load_config_file()
self.look_for_early_exit()
self.do_daemon_init_and_start()
self.load_modules_manager()
self.uri2 = self.pyro_daemon.register(self.interface, "ForArbiter")
logger.info("[scheduler] General interface is at: %s" % self.uri2)
self.do_mainloop()
except Exception, exp:
logger.critical("I got an unrecoverable error. I have to exit")
logger.critical(
"You can log a bug ticket at https://github.com/naparuba/shinken/issues/new to get help"
)
logger.critical("Back trace of it: %s" % (traceback.format_exc()))
raise
class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='/usr/local/shinken/var/schedulerd.pid'),
'port': IntegerProp(default='7768'),
'local_log': PathProp(default='/usr/local/shinken/var/schedulerd.log'),
})
#Create the shinken class:
#Create a Pyro server (port = arvg 1)
#then create the interface for arbiter
#Then, it wait for a first configuration
def __init__(self, config_file, is_daemon, do_replace, debug, debug_file):
BaseSatellite.__init__(self, 'scheduler', config_file, is_daemon, do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
def do_stop(self):
self.pyro_daemon.unregister(self.ibroks)
self.pyro_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.log('Warning: A system time change of %d has been detected. Compensating...' % difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
# Then we compasate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == 'notification':
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
# If we got USR1, just dump memory
if sig == 10:
self.sched.need_dump_memory = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
print "Ok we've got conf"
self.setup_new_conf()
print "Configuration Loaded"
self.sched.run()
def setup_new_conf(self):
#self.use_ssl = self.app.use_ssl
(conf, override_conf, modules, satellites) = self.new_conf
self.new_conf = None
# In fact it make the scheduler just DIE as a bad guy.
# Must manage it better or not manage it at all!
#if self.cur_conf and self.cur_conf.magic_hash == conf.magic_hash:
# print("I received a conf with same hash than me, I skip it.")
# return
self.conf = conf
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
#self.pollers = self.app.pollers
# Now We create our pollers
for pol_id in satellites['pollers']:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites['pollers'][pol_id]
self.pollers[pol_id] = p
uri = pyro.create_uri(p['address'], p['port'], 'Schedulers', self.use_ssl)
self.pollers[pol_id]['uri'] = uri
self.pollers[pol_id]['last_connexion'] = 0
print "Got a poller", p
#First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != 'NOTSET':
print "Setting our timezone to", self.conf.use_timezone
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
print "I've got modules", self.modules
# TODO: if scheduler had previous modules instanciated it must clean them !
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# And start external ones too
self.modules_manager.start_external_instances()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
print "Deconnecting previous Check Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ichecks)
#Now create and connect it
self.ichecks = IChecks(self.sched)
self.uri = self.pyro_daemon.register(self.ichecks, "Checks")
print "The Checks Interface uri is:", self.uri
#Same for Broks
if self.ibroks is not None:
print "Deconnecting previous Broks Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ibroks)
#Create and connect it
self.ibroks = IBroks(self.sched)
self.uri2 = self.pyro_daemon.register(self.ibroks, "Broks")
print "The Broks Interface uri is:", self.uri2
print("Loading configuration..")
self.conf.explode_global_conf()
#we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
#We must update our Config dict macro with good value
#from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
#print "DBG: got macors", self.sched.conf.macros
#Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
#self.conf.dump()
#self.conf.quick_debug()
#Now create the external commander
#it's a applyer : it role is not to dispatch commands,
#but to apply them
e = ExternalCommandManager(self.conf, 'applyer')
#Scheduler need to know about external command to
#activate it if necessery
self.sched.load_external_command(e)
#External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# our main function, launch after the init
def main(self):
self.load_config_file()
self.do_daemon_init_and_start()
self.uri2 = self.pyro_daemon.register(self.interface, "ForArbiter")
print "The Arbiter Interface is at:", self.uri2
self.do_mainloop()
class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='/usr/local/shinken/var/schedulerd.pid'),
'port': IntegerProp(default='7768'),
'local_log': PathProp(default='/usr/local/shinken/var/schedulerd.log'),
})
#Create the shinken class:
#Create a Pyro server (port = arvg 1)
#then create the interface for arbiter
#Then, it wait for a first configuration
def __init__(self, config_file, is_daemon, do_replace, debug, debug_file):
BaseSatellite.__init__(self, 'scheduler', config_file, is_daemon, do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
def do_stop(self):
self.pyro_daemon.unregister(self.ibroks)
self.pyro_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.log('Warning: A system time change of %d has been detected. Compensating...' % difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
# Then we compasate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == 'notification':
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
# If we got USR1, just dump memory
if sig == 10:
self.sched.need_dump_memory = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
print "Ok we've got conf"
self.setup_new_conf()
print "Configuration Loaded"
self.sched.run()
def setup_new_conf(self):
#self.use_ssl = self.app.use_ssl
(conf, override_conf, modules, satellites) = self.new_conf
self.new_conf = None
# In fact it make the scheduler just DIE as a bad guy.
# Must manage it better or not manage it at all!
#if self.cur_conf and self.cur_conf.magic_hash == conf.magic_hash:
# print("I received a conf with same hash than me, I skip it.")
# return
self.conf = conf
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
#self.pollers = self.app.pollers
# Now We create our pollers
for pol_id in satellites['pollers']:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites['pollers'][pol_id]
self.pollers[pol_id] = p
uri = pyro.create_uri(p['address'], p['port'], 'Schedulers', self.use_ssl)
self.pollers[pol_id]['uri'] = uri
self.pollers[pol_id]['last_connexion'] = 0
print "Got a poller", p
#First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != 'NOTSET':
print "Setting our timezone to", self.conf.use_timezone
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
print "I've got modules", self.modules
# TODO: if scheduler had previous modules instanciated it must clean them !
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# And start external ones too
self.modules_manager.start_external_instances()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
print "Deconnecting previous Check Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ichecks)
#Now create and connect it
self.ichecks = IChecks(self.sched)
self.uri = self.pyro_daemon.register(self.ichecks, "Checks")
print "The Checks Interface uri is:", self.uri
#Same for Broks
if self.ibroks is not None:
print "Deconnecting previous Broks Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ibroks)
#Create and connect it
self.ibroks = IBroks(self.sched)
self.uri2 = self.pyro_daemon.register(self.ibroks, "Broks")
print "The Broks Interface uri is:", self.uri2
print("Loading configuration..")
self.conf.explode_global_conf()
#we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
#We must update our Config dict macro with good value
#from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
#print "DBG: got macors", self.sched.conf.macros
#Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
#self.conf.dump()
#self.conf.quick_debug()
#Now create the external commander
#it's a applyer : it role is not to dispatch commands,
#but to apply them
e = ExternalCommandManager(self.conf, 'applyer')
#Scheduler need to know about external command to
#activate it if necessery
self.sched.load_external_command(e)
#External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# our main function, launch after the init
def main(self):
self.load_config_file()
self.do_daemon_init_and_start()
self.uri2 = self.pyro_daemon.register(self.interface, "ForArbiter")
print "The Arbiter Interface is at:", self.uri2
self.do_mainloop()
class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update(
{
"pidfile": PathProp(default="schedulerd.pid"),
"port": IntegerProp(default="7768"),
"local_log": PathProp(default="schedulerd.log"),
}
)
# Create the shinken class:
# Create a Pyro server (port = arvg 1)
# then create the interface for arbiter
# Then, it wait for a first configuration
def __init__(self, config_file, is_daemon, do_replace, debug, debug_file):
BaseSatellite.__init__(self, "scheduler", config_file, is_daemon, do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
def do_stop(self):
if self.pyro_daemon:
if self.ibroks:
self.pyro_daemon.unregister(self.ibroks)
if self.ichecks:
self.pyro_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.log("Warning: A system time change of %d has been detected. Compensating..." % difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
# Then we compasate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == "scheduled":
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
if ref.check_period is not None:
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = "waitconsume"
c.exit_status = 2
c.output = "(Error: there is no available check time after time change!)"
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == "scheduled":
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, "ref", None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == "notification":
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = "waitconsume"
c.exit_status = 2
c.output = "(Error: there is no available check time after time change!)"
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
# If we got USR1, just dump memory
if sig == 10:
self.sched.need_dump_memory = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
print "Ok we've got conf"
self.setup_new_conf()
print "Configuration Loaded"
self.sched.run()
def setup_new_conf(self):
# self.use_ssl = self.app.use_ssl
(conf, override_conf, modules, satellites) = self.new_conf
self.new_conf = None
# In fact it make the scheduler just DIE as a bad guy.
# Must manage it better or not manage it at all!
# if self.cur_conf and self.cur_conf.magic_hash == conf.magic_hash:
# print("I received a conf with same hash than me, I skip it.")
# return
self.conf = conf
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
# self.pollers = self.app.pollers
if self.conf.human_timestamp_log:
logger.set_human_format()
# Now We create our pollers
for pol_id in satellites["pollers"]:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites["pollers"][pol_id]
self.pollers[pol_id] = p
uri = pyro.create_uri(p["address"], p["port"], "Schedulers", self.use_ssl)
self.pollers[pol_id]["uri"] = uri
self.pollers[pol_id]["last_connection"] = 0
# First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
# print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != "":
print "Setting our timezone to", self.conf.use_timezone
os.environ["TZ"] = self.conf.use_timezone
time.tzset()
if len(self.modules) != 0:
print "I've got modules", self.modules
# TODO: if scheduler had previous modules instanciated it must clean them !
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# And start external ones too
self.modules_manager.start_external_instances()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
print "Deconnecting previous Check Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ichecks)
# Now create and connect it
self.ichecks = IChecks(self.sched)
self.uri = self.pyro_daemon.register(self.ichecks, "Checks")
print "The Checks Interface uri is:", self.uri
# Same for Broks
if self.ibroks is not None:
print "Deconnecting previous Broks Interface from pyro_daemon"
self.pyro_daemon.unregister(self.ibroks)
# Create and connect it
self.ibroks = IBroks(self.sched)
self.uri2 = self.pyro_daemon.register(self.ibroks, "Broks")
print "The Broks Interface uri is:", self.uri2
print ("Loading configuration..")
self.conf.explode_global_conf()
# we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
# We must update our Config dict macro with good value
# from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
# print "DBG: got macors", self.sched.conf.macros
# Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
# self.conf.dump()
# self.conf.quick_debug()
# Now create the external commander
# it's a applyer : it role is not to dispatch commands,
# but to apply them
e = ExternalCommandManager(self.conf, "applyer")
# Scheduler need to know about external command to
# activate it if necessery
self.sched.load_external_command(e)
# External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# We clear our schedulers managed (it's us :) )
# and set ourself in it
self.schedulers = {self.conf.instance_id: self.sched}
# Give the arbiter the data about what I manage
# for me it's just my instance_id and my push flavor
def what_i_managed(self):
if hasattr(self, "conf"):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
# our main function, launch after the init
def main(self):
try:
self.load_config_file()
self.do_daemon_init_and_start()
self.uri2 = self.pyro_daemon.register(self.interface, "ForArbiter")
logger.log("[scheduler] General interface is at: %s" % self.uri2)
self.do_mainloop()
except Exception, exp:
logger.log("CRITICAL ERROR: I got an unrecoverable error. I have to exit")
logger.log("You can log a bug ticket at https://github.com/naparuba/shinken/issues/new to get help")
logger.log("Back trace of it: %s" % (traceback.format_exc()))
raise
class Shinken(BaseSatellite):
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default=7768),
'local_log': PathProp(default='schedulerd.log'),
})
# Create the shinken class:
# Create a Pyro server (port = arvg 1)
# then create the interface for arbiter
# Then, it wait for a first configuration
def __init__(self,
config_file,
is_daemon,
do_replace,
debug,
debug_file,
profile=''):
BaseSatellite.__init__(self, 'scheduler', config_file, is_daemon,
do_replace, debug, debug_file)
self.interface = IForArbiter(self)
self.istats = IStats(self)
self.sched = Scheduler(self)
self.ichecks = None
self.ibroks = None
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
self.brokers = {}
def do_stop(self):
if self.http_daemon:
if self.ibroks:
self.http_daemon.unregister(self.ibroks)
if self.ichecks:
self.http_daemon.unregister(self.ichecks)
super(Shinken, self).do_stop()
def compensate_system_time_change(self, difference):
""" Compensate a system time change of difference for all hosts/services/checks/notifs """
logger.warning(
"A system time change of %d has been detected. Compensating...",
difference)
# We only need to change some value
self.program_start = max(0, self.program_start + difference)
if not hasattr(self.sched, "conf"):
# Race condition where time change before getting conf
return
# Then we compensate all host/services
for h in self.sched.hosts:
h.compensate_system_time_change(difference)
for s in self.sched.services:
s.compensate_system_time_change(difference)
# Now all checks and actions
for c in self.sched.checks.values():
# Already launch checks should not be touch
if c.status == 'scheduled' and c.t_to_go is not None:
t_to_go = c.t_to_go
ref = c.ref
new_t = max(0, t_to_go + difference)
if ref.check_period is not None:
# But it's no so simple, we must match the timeperiod
new_t = ref.check_period.get_next_valid_time_from_t(new_t)
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for c in self.sched.actions.values():
# Already launch checks should not be touch
if c.status == 'scheduled':
t_to_go = c.t_to_go
# Event handler do not have ref
ref = getattr(c, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if c.is_a == 'notification':
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
new_t = ref.notification_period.get_next_valid_time_from_t(
new_t)
# And got a creation_time variable too
c.creation_time = c.creation_time + difference
# But maybe no there is no more new value! Not good :(
# Say as error, with error output
if new_t is None:
c.state = 'waitconsume'
c.exit_status = 2
c.output = '(Error: there is no available check time after time change!)'
c.check_time = time.time()
c.execution_time = 0
else:
c.t_to_go = new_t
def manage_signal(self, sig, frame):
logger.warning("Received a SIGNAL %s", sig)
# If we got USR1, just dump memory
if sig == signal.SIGUSR1:
self.sched.need_dump_memory = True
elif sig == signal.SIGUSR2: # usr2, dump objects
self.sched.need_objects_dump = True
else: # if not, die :)
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
# Ok, now the conf
self.wait_for_initial_conf()
if not self.new_conf:
return
logger.info("New configuration received")
self.setup_new_conf()
logger.info("New configuration loaded")
self.sched.run()
if self.new_conf and self.graceful_enabled:
self.switch_process()
def setup_new_conf(self):
pk = self.new_conf
conf_raw = pk['conf']
override_conf = pk['override_conf']
modules = pk['modules']
satellites = pk['satellites']
instance_name = pk['instance_name']
push_flavor = pk['push_flavor']
skip_initial_broks = pk['skip_initial_broks']
accept_passive_unknown_check_results = pk[
'accept_passive_unknown_check_results']
api_key = pk['api_key']
secret = pk['secret']
http_proxy = pk['http_proxy']
statsd_host = pk['statsd_host']
statsd_port = pk['statsd_port']
statsd_prefix = pk['statsd_prefix']
statsd_enabled = pk['statsd_enabled']
statsd_interval = pk['statsd_interval']
statsd_types = pk['statsd_types']
statsd_pattern = pk['statsd_pattern']
harakiri_threshold = parse_memory_expr(pk['harakiri_threshold'])
# horay, we got a name, we can set it in our stats objects
statsmgr.register(self.sched,
instance_name,
'scheduler',
api_key=api_key,
secret=secret,
http_proxy=http_proxy,
statsd_host=statsd_host,
statsd_port=statsd_port,
statsd_prefix=statsd_prefix,
statsd_enabled=statsd_enabled,
statsd_interval=statsd_interval,
statsd_types=statsd_types,
statsd_pattern=statsd_pattern)
t0 = time.time()
conf = cPickle.loads(conf_raw)
logger.debug("Conf received at %d. Unserialized in %d secs", t0,
time.time() - t0)
if harakiri_threshold is not None:
self.raw_conf = self.new_conf
else:
self.raw_conf = None
self.new_conf = None
if self.aggressive_memory_management:
free_memory()
# Tag the conf with our data
self.conf = conf
self.conf.push_flavor = push_flavor
self.conf.instance_name = instance_name
self.conf.skip_initial_broks = skip_initial_broks
self.conf.accept_passive_unknown_check_results = accept_passive_unknown_check_results
self.cur_conf = conf
self.override_conf = override_conf
self.modules = modules
self.satellites = satellites
self.harakiri_threshold = harakiri_threshold
# self.pollers = self.app.pollers
if self.conf.human_timestamp_log:
logger.set_human_format()
# Now We create our pollers
for pol_id in satellites['pollers']:
# Must look if we already have it
already_got = pol_id in self.pollers
p = satellites['pollers'][pol_id]
self.pollers[pol_id] = p
if p['name'] in override_conf['satellitemap']:
p = dict(p) # make a copy
p.update(override_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]['last_connection'] = 0
# Now We create our reactionners
for reac_id in satellites['reactionners']:
# Must look if we already have it
already_got = reac_id in self.reactionners
reac = satellites['reactionners'][reac_id]
self.reactionners[reac_id] = reac
if reac['name'] in override_conf['satellitemap']:
reac = dict(reac) # make a copy
reac.update(override_conf['satellitemap'][reac['name']])
proto = 'http'
if p['use_ssl']:
proto = 'https'
uri = '%s://%s:%s/' % (proto, reac['address'], reac['port'])
self.reactionners[reac_id]['uri'] = uri
self.reactionners[reac_id]['last_connection'] = 0
# First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
# print "Overriding the property %s with value %s" % (prop, self.override_conf[prop])
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != '':
logger.debug("Setting our timezone to %s",
str(self.conf.use_timezone))
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
if len(self.modules) != 0:
logger.debug("I've got %s modules", str(self.modules))
# TODO: if scheduler had previous modules instanciated it must clean them!
self.modules_manager.set_modules(self.modules)
self.do_load_modules()
# give it an interface
# But first remove previous interface if exists
if self.ichecks is not None:
logger.debug("Deconnecting previous Check Interface")
self.http_daemon.unregister(self.ichecks)
# Now create and connect it
self.ichecks = IChecks(self.sched)
self.http_daemon.register(self.ichecks)
logger.debug("The Scheduler Interface uri is: %s", self.uri)
# Same for Broks
if self.ibroks is not None:
logger.debug("Deconnecting previous Broks Interface")
self.http_daemon.unregister(self.ibroks)
# Create and connect it
self.ibroks = IBroks(self.sched)
self.http_daemon.register(self.ibroks)
logger.info("Loading configuration.")
self.conf.explode_global_conf()
# we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners)
# We must update our Config dict macro with good value
# from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
# print "DBG: got macros", self.sched.conf.macros
# Creating the Macroresolver Class & unique instance
m = MacroResolver()
m.init(self.conf)
# self.conf.dump()
# self.conf.quick_debug()
# Now create the external commander
# it's a applyer: it role is not to dispatch commands,
# but to apply them
e = ExternalCommandManager(self.conf, 'applyer')
# Scheduler need to know about external command to
# activate it if necessary
self.sched.load_external_command(e)
# External command need the sched because he can raise checks
e.load_scheduler(self.sched)
# We clear our schedulers managed (it's us :) )
# and set ourself in it
self.schedulers = {self.conf.instance_id: self.sched}
# Suicide service if memory threshold has been exceeded
def check_memory_usage(self):
super(Shinken, self).check_memory_usage()
if self.new_conf is not None:
self.sched.die()
# Give the arbiter the data about what I manage
# for me it's just my instance_id and my push flavor
def what_i_managed(self):
if hasattr(self, 'conf'):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
# our main function, launch after the init
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')
self.look_for_early_exit()
self.load_parent_config()
self.do_daemon_init_and_start()
self.load_modules_manager()
self.http_daemon.register(self.interface)
self.http_daemon.register(self.istats)
# self.inject = Injector(self.sched)
# self.http_daemon.register(self.inject)
self.http_daemon.unregister(self.interface)
self.uri = self.http_daemon.uri
logger.info("[scheduler] General interface is at: %s", self.uri)
self.do_mainloop()
except Exception, exp:
self.print_unrecoverable(traceback.format_exc())
raise