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.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
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 __init__(self,
config_file,
is_daemon,
do_replace,
debug,
debug_file,
port=None,
local_log=None,
daemon_name=None):
self.daemon_name = 'scheduler'
if daemon_name:
self.daemon_name = daemon_name
BaseSatellite.__init__(self, self.daemon_name, config_file, is_daemon,
do_replace, debug, debug_file, port, local_log)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_broks_send = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
self.brokers = {}
def __init__(self, **kwargs):
"""Scheduler daemon initialisation
:param kwargs: command line arguments
"""
super(Alignak,
self).__init__(kwargs.get('daemon_name', 'Default-scheduler'),
**kwargs)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
self.brokers = {}
self.pollers = {}
self.reactionners = {}
self.receivers = {}
# This because it is the Satellite that has thes properties and I am a Satellite
# todo: change this?
# Broks are stored in each broker link, not locally
# self.broks = []
self.broks_lock = threading.RLock()
# Modules are only loaded one time
self.have_modules = False
self.first_scheduling = False
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.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
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 __init__(self, **kwargs):
"""Scheduler daemon initialisation
:param kwargs: command line arguments
"""
super(Alignak, self).__init__(kwargs.get('daemon_name', 'Default-scheduler'), **kwargs)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
self.brokers = {}
self.pollers = {}
self.reactionners = {}
self.receivers = {}
# This because it is the Satellite that has thes properties and I am a Satellite
# todo: change this?
# Broks are stored in each broker link, not locally
# self.broks = []
self.broks_lock = threading.RLock()
# Modules are only loaded one time
self.have_modules = False
self.first_scheduling = False
class Alignak(BaseSatellite):
# pylint: disable=too-many-instance-attributes
"""Scheduler class. Referenced as "app" in most Interface
"""
properties = BaseSatellite.properties.copy()
properties.update({
'type': StringProp(default='scheduler'),
'port': IntegerProp(default=7768)
})
def __init__(self, **kwargs):
"""Scheduler daemon initialisation
:param kwargs: command line arguments
"""
super(Alignak,
self).__init__(kwargs.get('daemon_name', 'Default-scheduler'),
**kwargs)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
self.brokers = {}
self.pollers = {}
self.reactionners = {}
self.receivers = {}
# This because it is the Satellite that has thes properties and I am a Satellite
# todo: change this?
# Broks are stored in each broker link, not locally
# self.broks = []
self.broks_lock = threading.RLock()
# Modules are only loaded one time
self.have_modules = False
self.first_scheduling = False
def get_broks(self, broker_name):
"""Send broks to a specific broker
:param broker_name: broker name to send broks
:type broker_name: str
:greturn: dict of brok for this broker
:rtype: dict[alignak.brok.Brok]
"""
logger.debug("Broker %s requests my broks list", broker_name)
res = []
if not broker_name:
return res
for broker_link in list(self.brokers.values()):
if broker_name == broker_link.name:
for brok in sorted(broker_link.broks,
key=lambda x: x.creation_time):
# Only provide broks that did not yet sent to our external modules
if getattr(brok, 'sent_to_externals', False):
res.append(brok)
brok.got = True
broker_link.broks = [
b for b in broker_link.broks
if not getattr(b, 'got', False)
]
logger.debug("Providing %d broks to %s", len(res), broker_name)
break
else:
logger.warning("Got a brok request from an unknown broker: %s",
broker_name)
return res
def compensate_system_time_change(self, difference): # pragma: no cover,
# pylint: disable=too-many-branches
# not with unit tests
"""Compensate a system time change of difference for all hosts/services/checks/notifs
:param difference: difference in seconds
:type difference: int
:return: None
"""
super(Alignak, self).compensate_system_time_change(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 host in self.sched.hosts:
host.compensate_system_time_change(difference)
for serv in self.sched.services:
serv.compensate_system_time_change(difference)
# Now all checks and actions
for chk in list(self.sched.checks.values()):
# Already launch checks should not be touch
if chk.status == u'scheduled' and chk.t_to_go is not None:
t_to_go = chk.t_to_go
ref = self.sched.find_item_by_id(chk.ref)
new_t = max(0, t_to_go + difference)
timeperiod = self.sched.timeperiods[ref.check_period]
if timeperiod is not None:
# But it's no so simple, we must match the timeperiod
new_t = timeperiod.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:
chk.state = u'waitconsume'
chk.exit_status = 2
chk.output = '(Error: there is no available check time after time change!)'
chk.check_time = time.time()
chk.execution_time = 0
else:
chk.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for act in list(self.sched.actions.values()):
# Already launch checks should not be touch
if act.status == u'scheduled':
t_to_go = act.t_to_go
# Event handler do not have ref
ref_id = getattr(act, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if act.is_a == u'notification':
ref = self.sched.find_item_by_id(ref_id)
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
notification_period = self.sched.timeperiods[
ref.notification_period]
new_t = notification_period.get_next_valid_time_from_t(
new_t)
# And got a creation_time variable too
act.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:
act.state = 'waitconsume'
act.exit_status = 2
act.output = '(Error: there is no available check time after time change!)'
act.check_time = time.time()
act.execution_time = 0
else:
act.t_to_go = new_t
def do_before_loop(self):
"""Stop the scheduling process"""
if self.sched:
self.sched.stop_scheduling()
def do_loop_turn(self):
"""Scheduler loop turn
Simply run the Alignak scheduler loop
This is called when a configuration got received by the scheduler daemon. As of it,
check if the first scheduling has been done... and manage this.
:return: None
"""
if not self.first_scheduling:
# Ok, now all is initialized, we can make the initial broks
logger.info("First scheduling launched")
_t0 = time.time()
# Program start brok
self.sched.initial_program_status()
# First scheduling
self.sched.schedule()
statsmgr.timer('first_scheduling', time.time() - _t0)
logger.info("First scheduling done")
# Connect to our passive satellites if needed
for satellite in [
s for s in list(self.pollers.values()) if s.passive
]:
if not self.daemon_connection_init(satellite):
logger.error("Passive satellite connection failed: %s",
satellite)
for satellite in [
s for s in list(self.reactionners.values()) if s.passive
]:
if not self.daemon_connection_init(satellite):
logger.error("Passive satellite connection failed: %s",
satellite)
# Ticks are for recurrent function call like consume, del zombies etc
self.sched.ticks = 0
self.first_scheduling = True
# Each loop turn, execute the daemon specific treatment...
# only if the daemon has a configuration to manage
if self.sched.pushed_conf:
# If scheduling is not yet enabled, enable scheduling
if not self.sched.must_schedule:
self.sched.start_scheduling()
self.sched.before_run()
self.sched.run()
else:
logger.warning("#%d - No monitoring configuration to scheduler...",
self.loop_count)
def get_managed_configurations(self):
"""Get the configurations managed by this scheduler
The configuration managed by a scheduler is the self configuration got
by the scheduler during the dispatching.
:return: a dict of scheduler links with instance_id as key and
hash, push_flavor and configuration identifier as values
:rtype: dict
"""
# for scheduler_link in list(self.schedulers.values()):
# res[scheduler_link.instance_id] = {
# 'hash': scheduler_link.hash,
# 'push_flavor': scheduler_link.push_flavor,
# 'managed_conf_id': scheduler_link.managed_conf_id
# }
res = {}
if self.sched.pushed_conf and self.cur_conf and 'instance_id' in self.cur_conf:
res[self.cur_conf['instance_id']] = {
'hash': self.cur_conf['hash'],
'push_flavor': self.cur_conf['push_flavor'],
'managed_conf_id': self.cur_conf['managed_conf_id']
}
logger.debug("Get managed configuration: %s", res)
return res
def setup_new_conf(self):
# pylint: disable=too-many-statements, too-many-branches, too-many-locals
"""Setup new conf received for scheduler
:return: None
"""
# Execute the base class treatment...
super(Alignak, self).setup_new_conf()
# ...then our own specific treatment!
with self.conf_lock:
# self_conf is our own configuration from the alignak environment
# self_conf = self.cur_conf['self_conf']
logger.debug("Got config: %s", self.cur_conf)
if 'conf_part' not in self.cur_conf:
self.cur_conf['conf_part'] = None
conf_part = self.cur_conf['conf_part']
# Ok now we can save the retention data
if self.sched.pushed_conf is not None:
self.sched.update_retention()
# Get the monitored objects configuration
t00 = time.time()
received_conf_part = None
try:
received_conf_part = unserialize(conf_part)
assert received_conf_part is not None
except AssertionError as exp:
# This to indicate that no configuration is managed by this scheduler...
logger.warning(
"No managed configuration received from arbiter")
except AlignakClassLookupException as exp: # pragma: no cover
# This to indicate that the new configuration is not managed...
self.new_conf = {
"_status":
"Cannot un-serialize configuration received from arbiter",
"_error": str(exp)
}
logger.error(self.new_conf)
logger.error("Back trace of the error:\n%s",
traceback.format_exc())
return
except Exception as exp: # pylint: disable=broad-except
# This to indicate that the new configuration is not managed...
self.new_conf = {
"_status":
"Cannot un-serialize configuration received from arbiter",
"_error": str(exp)
}
logger.error(self.new_conf)
self.exit_on_exception(exp, str(self.new_conf))
# if not received_conf_part:
# return
logger.info(
"Monitored configuration %s received at %d. Un-serialized in %d secs",
received_conf_part, t00,
time.time() - t00)
logger.info("Scheduler received configuration : %s",
received_conf_part)
# Now we create our pollers, reactionners and brokers
for link_type in ['pollers', 'reactionners', 'brokers']:
if link_type not in self.cur_conf['satellites']:
logger.error("Missing %s in the configuration!", link_type)
continue
my_satellites = getattr(self, link_type, {})
received_satellites = self.cur_conf['satellites'][link_type]
for link_uuid in received_satellites:
rs_conf = received_satellites[link_uuid]
logger.debug("- received %s - %s: %s",
rs_conf['instance_id'], rs_conf['type'],
rs_conf['name'])
# Must look if we already had a configuration and save our broks
already_got = rs_conf['instance_id'] in my_satellites
broks = []
actions = {}
wait_homerun = {}
external_commands = {}
running_id = 0
if already_got:
logger.warning("I already got: %s",
rs_conf['instance_id'])
# Save some information
running_id = my_satellites[link_uuid].running_id
(broks, actions,
wait_homerun, external_commands) = \
my_satellites[link_uuid].get_and_clear_context()
# Delete the former link
del my_satellites[link_uuid]
# My new satellite link...
new_link = SatelliteLink.get_a_satellite_link(
link_type[:-1], rs_conf)
my_satellites[new_link.uuid] = new_link
logger.info("I got a new %s satellite: %s", link_type[:-1],
new_link)
new_link.running_id = running_id
new_link.external_commands = external_commands
new_link.broks = broks
new_link.wait_homerun = wait_homerun
new_link.actions = actions
# Replacing the satellite address and port by those defined in satellite_map
if new_link.name in self.cur_conf['override_conf'].get(
'satellite_map', {}):
override_conf = self.cur_conf['override_conf']
overriding = override_conf.get('satellite_map')[
new_link.name]
logger.warning(
"Do not override the configuration for: %s, with: %s. "
"Please check whether this is necessary!",
new_link.name, overriding)
# First mix conf and override_conf to have our definitive conf
for prop in getattr(self.cur_conf, 'override_conf', []):
logger.debug("Overriden: %s / %s ", prop,
getattr(received_conf_part, prop, None))
logger.debug("Overriding: %s / %s ", prop,
self.cur_conf['override_conf'])
setattr(received_conf_part, prop,
self.cur_conf['override_conf'].get(prop, None))
# Scheduler modules
if not self.have_modules:
try:
logger.debug("Modules configuration: %s",
self.cur_conf['modules'])
self.modules = unserialize(self.cur_conf['modules'],
no_load=True)
except AlignakClassLookupException as exp: # pragma: no cover, simple protection
logger.error(
'Cannot un-serialize modules configuration '
'received from arbiter: %s', exp)
if self.modules:
logger.debug("I received some modules configuration: %s",
self.modules)
self.have_modules = True
self.do_load_modules(self.modules)
# and start external modules too
self.modules_manager.start_external_instances()
else:
logger.info("I do not have modules")
if received_conf_part:
logger.info("Loading configuration...")
# Propagate the global parameters to the configuration items
received_conf_part.explode_global_conf()
# We give the configuration to our scheduler
self.sched.reset()
self.sched.load_conf(self.cur_conf['instance_id'],
self.cur_conf['instance_name'],
received_conf_part)
# Once loaded, the scheduler has an inner pushed_conf object
logger.info("Loaded: %s", self.sched.pushed_conf)
# Update the scheduler ticks according to the daemon configuration
self.sched.update_recurrent_works_tick(self)
# We must update our pushed configuration macros with correct values
# from the configuration parameters
# self.sched.pushed_conf.fill_resource_macros_names_macros()
# Creating the Macroresolver Class & unique instance
m_solver = MacroResolver()
m_solver.init(received_conf_part)
# Now create the external commands manager
# We are an applyer: our role is not to dispatch commands, but to apply them
ecm = ExternalCommandManager(
received_conf_part, 'applyer', self.sched,
received_conf_part.accept_passive_unknown_check_results,
received_conf_part.log_external_commands)
# Scheduler needs to know about this external command manager to use it if necessary
self.sched.external_commands_manager = ecm
# Ok now we can load the retention data
self.sched.retention_load()
# Log hosts/services initial states
self.sched.log_initial_states()
# Create brok new conf
brok = Brok({'type': 'new_conf', 'data': {}})
self.sched.add_brok(brok)
# Initialize connection with all our satellites
logger.info("Initializing connection with my satellites:")
my_satellites = self.get_links_of_type(s_type='')
for satellite in list(my_satellites.values()):
logger.info("- : %s/%s", satellite.type, satellite.name)
if not self.daemon_connection_init(satellite):
logger.error("Satellite connection failed: %s", satellite)
if received_conf_part:
# Enable the scheduling process
logger.info("Loaded: %s", self.sched.pushed_conf)
self.sched.start_scheduling()
# Now I have a configuration!
self.have_conf = True
def clean_previous_run(self):
"""Clean variables from previous configuration
:return: None
"""
# Execute the base class treatment...
super(Alignak, self).clean_previous_run()
# Clean all lists
self.pollers.clear()
self.reactionners.clear()
self.brokers.clear()
def get_daemon_stats(self, details=False):
"""Increase the stats provided by the Daemon base class
:return: stats dictionary
:rtype: dict
"""
# Call the base Daemon one
res = super(Alignak, self).get_daemon_stats(details=details)
res.update({
'name': self.name,
'type': self.type,
'monitored_objects': {}
})
counters = res['counters']
# Satellites counters
counters['brokers'] = len(self.brokers)
counters['pollers'] = len(self.pollers)
counters['reactionners'] = len(self.reactionners)
counters['receivers'] = len(self.receivers)
if not self.sched:
return res
# # Hosts/services problems counters
# m_solver = MacroResolver()
# counters['hosts_problems'] = m_solver._get_total_host_problems()
# counters['hosts_unhandled_problems'] = m_solver._get_total_host_problems_unhandled()
# counters['services_problems'] = m_solver._get_total_service_problems()
# counters['services_unhandled_problems'] = m_solver._get_total_service_problems_unhandled()
# Get statistics from the scheduler
scheduler_stats = self.sched.get_scheduler_stats(details=details)
res['counters'].update(scheduler_stats['counters'])
scheduler_stats.pop('counters')
res.update(scheduler_stats)
return res
def get_monitoring_problems(self):
"""Get the current scheduler livesynthesis
:return: live synthesis and problems dictionary
:rtype: dict
"""
res = {}
if not self.sched:
return res
# Get statistics from the scheduler
scheduler_stats = self.sched.get_scheduler_stats(details=True)
if 'livesynthesis' in scheduler_stats:
res['livesynthesis'] = scheduler_stats['livesynthesis']
if 'problems' in scheduler_stats:
res['problems'] = scheduler_stats['problems']
return res
def main(self):
"""Main function for Scheduler, launch after the init::
* Init daemon
* Load module manager
* Launch main loop
* Catch any Exception that occurs
:return: None
"""
try:
# Start the daemon mode
if not self.do_daemon_init_and_start():
self.exit_on_error(message="Daemon initialization error",
exit_code=3)
# We wait for initial conf
self.wait_for_initial_conf()
if self.new_conf:
# Setup the received configuration
self.setup_new_conf()
# Now the main loop
self.do_main_loop()
logger.info("Exited from the main loop.")
# On main loop exit, call the scheduler after run process
self.sched.after_run()
self.request_stop()
except Exception: # pragma: no cover, this should never happen indeed ;)
self.exit_on_exception(traceback.format_exc())
raise
class Alignak(BaseSatellite):
# pylint: disable=too-many-instance-attributes
"""Scheduler class. Referenced as "app" in most Interface
"""
properties = BaseSatellite.properties.copy()
properties.update({
'type':
StringProp(default='scheduler'),
'port':
IntegerProp(default=7768)
})
def __init__(self, **kwargs):
"""Scheduler daemon initialisation
:param kwargs: command line arguments
"""
super(Alignak, self).__init__(kwargs.get('daemon_name', 'Default-scheduler'), **kwargs)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
self.brokers = {}
self.pollers = {}
self.reactionners = {}
self.receivers = {}
# This because it is the Satellite that has thes properties and I am a Satellite
# todo: change this?
# Broks are stored in each broker link, not locally
# self.broks = []
self.broks_lock = threading.RLock()
# Modules are only loaded one time
self.have_modules = False
self.first_scheduling = False
def get_broks(self, broker_name):
"""Send broks to a specific broker
:param broker_name: broker name to send broks
:type broker_name: str
:greturn: dict of brok for this broker
:rtype: dict[alignak.brok.Brok]
"""
logger.debug("Broker %s requests my broks list", broker_name)
res = []
if not broker_name:
return res
for broker_link in list(self.brokers.values()):
if broker_name == broker_link.name:
for brok in sorted(broker_link.broks, key=lambda x: x.creation_time):
# Only provide broks that did not yet sent to our external modules
if getattr(brok, 'sent_to_externals', False):
res.append(brok)
brok.got = True
broker_link.broks = [b for b in broker_link.broks if not getattr(b, 'got', False)]
logger.debug("Providing %d broks to %s", len(res), broker_name)
break
else:
logger.warning("Got a brok request from an unknown broker: %s", broker_name)
return res
def compensate_system_time_change(self, difference): # pragma: no cover,
# pylint: disable=too-many-branches
# not with unit tests
"""Compensate a system time change of difference for all hosts/services/checks/notifs
:param difference: difference in seconds
:type difference: int
:return: None
"""
super(Alignak, self).compensate_system_time_change(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 host in self.sched.hosts:
host.compensate_system_time_change(difference)
for serv in self.sched.services:
serv.compensate_system_time_change(difference)
# Now all checks and actions
for chk in list(self.sched.checks.values()):
# Already launch checks should not be touch
if chk.status == u'scheduled' and chk.t_to_go is not None:
t_to_go = chk.t_to_go
ref = self.sched.find_item_by_id(chk.ref)
new_t = max(0, t_to_go + difference)
timeperiod = self.sched.timeperiods[ref.check_period]
if timeperiod is not None:
# But it's no so simple, we must match the timeperiod
new_t = timeperiod.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:
chk.state = u'waitconsume'
chk.exit_status = 2
chk.output = '(Error: there is no available check time after time change!)'
chk.check_time = time.time()
chk.execution_time = 0
else:
chk.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for act in list(self.sched.actions.values()):
# Already launch checks should not be touch
if act.status == u'scheduled':
t_to_go = act.t_to_go
# Event handler do not have ref
ref_id = getattr(act, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if act.is_a == u'notification':
ref = self.sched.find_item_by_id(ref_id)
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
notification_period = self.sched.timeperiods[ref.notification_period]
new_t = notification_period.get_next_valid_time_from_t(new_t)
# And got a creation_time variable too
act.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:
act.state = 'waitconsume'
act.exit_status = 2
act.output = '(Error: there is no available check time after time change!)'
act.check_time = time.time()
act.execution_time = 0
else:
act.t_to_go = new_t
def do_before_loop(self):
"""Stop the scheduling process"""
if self.sched:
self.sched.stop_scheduling()
def do_loop_turn(self):
"""Scheduler loop turn
Simply run the Alignak scheduler loop
This is called when a configuration got received by the scheduler daemon. As of it,
check if the first scheduling has been done... and manage this.
:return: None
"""
if not self.first_scheduling:
# Ok, now all is initialized, we can make the initial broks
logger.info("First scheduling launched")
_t0 = time.time()
# Program start brok
self.sched.initial_program_status()
# First scheduling
self.sched.schedule()
statsmgr.timer('first_scheduling', time.time() - _t0)
logger.info("First scheduling done")
# Connect to our passive satellites if needed
for satellite in [s for s in list(self.pollers.values()) if s.passive]:
if not self.daemon_connection_init(satellite):
logger.error("Passive satellite connection failed: %s", satellite)
for satellite in [s for s in list(self.reactionners.values()) if s.passive]:
if not self.daemon_connection_init(satellite):
logger.error("Passive satellite connection failed: %s", satellite)
# Ticks are for recurrent function call like consume, del zombies etc
self.sched.ticks = 0
self.first_scheduling = True
# Each loop turn, execute the daemon specific treatment...
# only if the daemon has a configuration to manage
if self.sched.pushed_conf:
# If scheduling is not yet enabled, enable scheduling
if not self.sched.must_schedule:
self.sched.start_scheduling()
self.sched.before_run()
self.sched.run()
else:
logger.warning("#%d - No monitoring configuration to scheduler...",
self.loop_count)
def get_managed_configurations(self):
"""Get the configurations managed by this scheduler
The configuration managed by a scheduler is the self configuration got
by the scheduler during the dispatching.
:return: a dict of scheduler links with instance_id as key and
hash, push_flavor and configuration identifier as values
:rtype: dict
"""
# for scheduler_link in list(self.schedulers.values()):
# res[scheduler_link.instance_id] = {
# 'hash': scheduler_link.hash,
# 'push_flavor': scheduler_link.push_flavor,
# 'managed_conf_id': scheduler_link.managed_conf_id
# }
res = {}
if self.sched.pushed_conf and self.cur_conf and 'instance_id' in self.cur_conf:
res[self.cur_conf['instance_id']] = {
'hash': self.cur_conf['hash'],
'push_flavor': self.cur_conf['push_flavor'],
'managed_conf_id': self.cur_conf['managed_conf_id']
}
logger.debug("Get managed configuration: %s", res)
return res
def setup_new_conf(self):
# pylint: disable=too-many-statements, too-many-branches, too-many-locals
"""Setup new conf received for scheduler
:return: None
"""
# Execute the base class treatment...
super(Alignak, self).setup_new_conf()
# ...then our own specific treatment!
with self.conf_lock:
# self_conf is our own configuration from the alignak environment
# self_conf = self.cur_conf['self_conf']
logger.debug("Got config: %s", self.cur_conf)
if 'conf_part' not in self.cur_conf:
self.cur_conf['conf_part'] = None
conf_part = self.cur_conf['conf_part']
# Ok now we can save the retention data
if self.sched.pushed_conf is not None:
self.sched.update_retention()
# Get the monitored objects configuration
t00 = time.time()
received_conf_part = None
try:
received_conf_part = unserialize(conf_part)
assert received_conf_part is not None
except AssertionError as exp:
# This to indicate that no configuration is managed by this scheduler...
logger.warning("No managed configuration received from arbiter")
except AlignakClassLookupException as exp: # pragma: no cover
# This to indicate that the new configuration is not managed...
self.new_conf = {
"_status": "Cannot un-serialize configuration received from arbiter",
"_error": str(exp)
}
logger.error(self.new_conf)
logger.error("Back trace of the error:\n%s", traceback.format_exc())
return
except Exception as exp: # pylint: disable=broad-except
# This to indicate that the new configuration is not managed...
self.new_conf = {
"_status": "Cannot un-serialize configuration received from arbiter",
"_error": str(exp)
}
logger.error(self.new_conf)
self.exit_on_exception(exp, str(self.new_conf))
# if not received_conf_part:
# return
logger.info("Monitored configuration %s received at %d. Un-serialized in %d secs",
received_conf_part, t00, time.time() - t00)
logger.info("Scheduler received configuration : %s", received_conf_part)
# Now we create our pollers, reactionners and brokers
for link_type in ['pollers', 'reactionners', 'brokers']:
if link_type not in self.cur_conf['satellites']:
logger.error("Missing %s in the configuration!", link_type)
continue
my_satellites = getattr(self, link_type, {})
received_satellites = self.cur_conf['satellites'][link_type]
for link_uuid in received_satellites:
rs_conf = received_satellites[link_uuid]
logger.debug("- received %s - %s: %s", rs_conf['instance_id'],
rs_conf['type'], rs_conf['name'])
# Must look if we already had a configuration and save our broks
already_got = rs_conf['instance_id'] in my_satellites
broks = []
actions = {}
wait_homerun = {}
external_commands = {}
running_id = 0
if already_got:
logger.warning("I already got: %s", rs_conf['instance_id'])
# Save some information
running_id = my_satellites[link_uuid].running_id
(broks, actions,
wait_homerun, external_commands) = \
my_satellites[link_uuid].get_and_clear_context()
# Delete the former link
del my_satellites[link_uuid]
# My new satellite link...
new_link = SatelliteLink.get_a_satellite_link(link_type[:-1],
rs_conf)
my_satellites[new_link.uuid] = new_link
logger.info("I got a new %s satellite: %s", link_type[:-1], new_link)
new_link.running_id = running_id
new_link.external_commands = external_commands
new_link.broks = broks
new_link.wait_homerun = wait_homerun
new_link.actions = actions
# Replacing the satellite address and port by those defined in satellite_map
if new_link.name in self.cur_conf['override_conf'].get('satellite_map', {}):
override_conf = self.cur_conf['override_conf']
overriding = override_conf.get('satellite_map')[new_link.name]
logger.warning("Do not override the configuration for: %s, with: %s. "
"Please check whether this is necessary!",
new_link.name, overriding)
# First mix conf and override_conf to have our definitive conf
for prop in getattr(self.cur_conf, 'override_conf', []):
logger.debug("Overriden: %s / %s ", prop, getattr(received_conf_part, prop, None))
logger.debug("Overriding: %s / %s ", prop, self.cur_conf['override_conf'])
setattr(received_conf_part, prop, self.cur_conf['override_conf'].get(prop, None))
# Scheduler modules
if not self.have_modules:
try:
logger.debug("Modules configuration: %s", self.cur_conf['modules'])
self.modules = unserialize(self.cur_conf['modules'], no_load=True)
except AlignakClassLookupException as exp: # pragma: no cover, simple protection
logger.error('Cannot un-serialize modules configuration '
'received from arbiter: %s', exp)
if self.modules:
logger.debug("I received some modules configuration: %s", self.modules)
self.have_modules = True
self.do_load_modules(self.modules)
# and start external modules too
self.modules_manager.start_external_instances()
else:
logger.info("I do not have modules")
if received_conf_part:
logger.info("Loading configuration...")
# Propagate the global parameters to the configuration items
received_conf_part.explode_global_conf()
# We give the configuration to our scheduler
self.sched.reset()
self.sched.load_conf(self.cur_conf['instance_id'],
self.cur_conf['instance_name'],
received_conf_part)
# Once loaded, the scheduler has an inner pushed_conf object
logger.info("Loaded: %s", self.sched.pushed_conf)
# Update the scheduler ticks according to the daemon configuration
self.sched.update_recurrent_works_tick(self)
# We must update our pushed configuration macros with correct values
# from the configuration parameters
# self.sched.pushed_conf.fill_resource_macros_names_macros()
# Creating the Macroresolver Class & unique instance
m_solver = MacroResolver()
m_solver.init(received_conf_part)
# Now create the external commands manager
# We are an applyer: our role is not to dispatch commands, but to apply them
ecm = ExternalCommandManager(
received_conf_part, 'applyer', self.sched,
received_conf_part.accept_passive_unknown_check_results,
received_conf_part.log_external_commands)
# Scheduler needs to know about this external command manager to use it if necessary
self.sched.external_commands_manager = ecm
# Ok now we can load the retention data
self.sched.retention_load()
# Log hosts/services initial states
self.sched.log_initial_states()
# Create brok new conf
brok = Brok({'type': 'new_conf', 'data': {}})
self.sched.add_brok(brok)
# Initialize connection with all our satellites
logger.info("Initializing connection with my satellites:")
my_satellites = self.get_links_of_type(s_type='')
for satellite in list(my_satellites.values()):
logger.info("- : %s/%s", satellite.type, satellite.name)
if not self.daemon_connection_init(satellite):
logger.error("Satellite connection failed: %s", satellite)
if received_conf_part:
# Enable the scheduling process
logger.info("Loaded: %s", self.sched.pushed_conf)
self.sched.start_scheduling()
# Now I have a configuration!
self.have_conf = True
def clean_previous_run(self):
"""Clean variables from previous configuration
:return: None
"""
# Execute the base class treatment...
super(Alignak, self).clean_previous_run()
# Clean all lists
self.pollers.clear()
self.reactionners.clear()
self.brokers.clear()
def get_daemon_stats(self, details=False):
"""Increase the stats provided by the Daemon base class
:return: stats dictionary
:rtype: dict
"""
# Call the base Daemon one
res = super(Alignak, self).get_daemon_stats(details=details)
res.update({'name': self.name, 'type': self.type, 'monitored_objects': {}})
counters = res['counters']
# Satellites counters
counters['brokers'] = len(self.brokers)
counters['pollers'] = len(self.pollers)
counters['reactionners'] = len(self.reactionners)
counters['receivers'] = len(self.receivers)
if not self.sched:
return res
# # Hosts/services problems counters
# m_solver = MacroResolver()
# counters['hosts_problems'] = m_solver._get_total_host_problems()
# counters['hosts_unhandled_problems'] = m_solver._get_total_host_problems_unhandled()
# counters['services_problems'] = m_solver._get_total_service_problems()
# counters['services_unhandled_problems'] = m_solver._get_total_service_problems_unhandled()
# Get statistics from the scheduler
scheduler_stats = self.sched.get_scheduler_stats(details=details)
res['counters'].update(scheduler_stats['counters'])
scheduler_stats.pop('counters')
res.update(scheduler_stats)
return res
def get_monitoring_problems(self):
"""Get the current scheduler livesynthesis
:return: live synthesis and problems dictionary
:rtype: dict
"""
res = {}
if not self.sched:
return res
# Get statistics from the scheduler
scheduler_stats = self.sched.get_scheduler_stats(details=True)
if 'livesynthesis' in scheduler_stats:
res['livesynthesis'] = scheduler_stats['livesynthesis']
if 'problems' in scheduler_stats:
res['problems'] = scheduler_stats['problems']
return res
def main(self):
"""Main function for Scheduler, launch after the init::
* Init daemon
* Load module manager
* Launch main loop
* Catch any Exception that occurs
:return: None
"""
try:
# Start the daemon mode
if not self.do_daemon_init_and_start():
self.exit_on_error(message="Daemon initialization error", exit_code=3)
# We wait for initial conf
self.wait_for_initial_conf()
if self.new_conf:
# Setup the received configuration
self.setup_new_conf()
# Now the main loop
self.do_main_loop()
logger.info("Exited from the main loop.")
# On main loop exit, call the scheduler after run process
self.sched.after_run()
self.request_stop()
except Exception: # pragma: no cover, this should never happen indeed ;)
self.exit_on_exception(traceback.format_exc())
raise
class Alignak(BaseSatellite):
"""Scheduler class. Referenced as "app" in most Interface
"""
properties = BaseSatellite.properties.copy()
properties.update({
'daemon_type': StringProp(default='scheduler'),
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default=7768),
'local_log': PathProp(default='schedulerd.log'),
})
def __init__(self,
config_file,
is_daemon,
do_replace,
debug,
debug_file,
port=None,
local_log=None,
daemon_name=None):
self.daemon_name = 'scheduler'
if daemon_name:
self.daemon_name = daemon_name
BaseSatellite.__init__(self, self.daemon_name, config_file, is_daemon,
do_replace, debug, debug_file, port, local_log)
self.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
self.must_run = True
# Now the interface
self.uri = None
self.uri2 = None
# stats part
# --- copied from scheduler.py
self.nb_pulled_checks = 0
self.nb_pulled_actions = 0
# self.nb_checks_send = 0
self.nb_pushed_checks = 0
self.nb_pushed_actions = 0
self.nb_broks_send = 0
self.nb_pulled_broks = 0
# ---
# And possible links for satellites
# from now only pollers
self.pollers = {}
self.reactionners = {}
self.brokers = {}
def compensate_system_time_change(self, difference,
timeperiods): # pragma: no cover,
# not with unit tests
"""Compensate a system time change of difference for all hosts/services/checks/notifs
:param difference: difference in seconds
:type difference: int
:return: None
"""
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 host in self.sched.hosts:
host.compensate_system_time_change(difference)
for serv in self.sched.services:
serv.compensate_system_time_change(difference)
# Now all checks and actions
for chk in self.sched.checks.values():
# Already launch checks should not be touch
if chk.status == 'scheduled' and chk.t_to_go is not None:
t_to_go = chk.t_to_go
ref = self.sched.find_item_by_id(chk.ref)
new_t = max(0, t_to_go + difference)
timeperiod = timeperiods[ref.check_period]
if timeperiod is not None:
# But it's no so simple, we must match the timeperiod
new_t = timeperiod.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:
chk.state = 'waitconsume'
chk.exit_status = 2
chk.output = '(Error: there is no available check time after time change!)'
chk.check_time = time.time()
chk.execution_time = 0
else:
chk.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for act in self.sched.actions.values():
# Already launch checks should not be touch
if act.status == 'scheduled':
t_to_go = act.t_to_go
# Event handler do not have ref
ref_id = getattr(act, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if act.is_a == 'notification':
ref = self.sched.find_item_by_id(ref_id)
if ref.notification_period:
# But it's no so simple, we must match the timeperiod
notification_period = self.sched.timeperiods[
ref.notification_period]
new_t = notification_period.get_next_valid_time_from_t(
new_t)
# And got a creation_time variable too
act.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:
act.state = 'waitconsume'
act.exit_status = 2
act.output = '(Error: there is no available check time after time change!)'
act.check_time = time.time()
act.execution_time = 0
else:
act.t_to_go = new_t
def manage_signal(self, sig, frame):
"""Manage signals caught by the daemon
signal.SIGUSR1 : dump_memory
signal.SIGUSR2 : dump_object (nothing)
signal.SIGTERM, signal.SIGINT : terminate process
:param sig: signal caught by daemon
:type sig: str
:param frame: current stack frame
:type frame:
:return: None
TODO: Refactor with Daemon one
"""
logger.info("scheduler process %d received a signal: %s", os.getpid(),
str(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 :)
logger.info("scheduler process %d is dying...", os.getpid())
self.sched.die()
self.must_run = False
Daemon.manage_signal(self, sig, frame)
def do_loop_turn(self):
"""Scheduler loop turn
Basically wait initial conf and run
:return: None
"""
# 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(
"[%s] New configuration loaded, scheduling for Alignak: %s",
self.name, self.sched.alignak_name)
self.sched.run()
def setup_new_conf(self): # pylint: disable=too-many-statements
"""Setup new conf received for scheduler
:return: None
"""
with self.conf_lock:
self.clean_previous_run()
new_conf = self.new_conf
logger.info("[%s] Sending us a configuration", self.name)
conf_raw = new_conf['conf']
override_conf = new_conf['override_conf']
modules = new_conf['modules']
satellites = new_conf['satellites']
instance_name = new_conf['instance_name']
# Ok now we can save the retention data
if hasattr(self.sched, 'conf'):
self.sched.update_retention_file(forced=True)
# horay, we got a name, we can set it in our stats objects
statsmgr.register(instance_name,
'scheduler',
statsd_host=new_conf['statsd_host'],
statsd_port=new_conf['statsd_port'],
statsd_prefix=new_conf['statsd_prefix'],
statsd_enabled=new_conf['statsd_enabled'])
t00 = time.time()
try:
conf = unserialize(conf_raw)
except AlignakClassLookupException as exp: # pragma: no cover, simple protection
logger.error(
'Cannot un-serialize configuration received from arbiter: %s',
exp)
logger.debug("Conf received at %d. Un-serialized in %d secs", t00,
time.time() - t00)
self.new_conf = None
if 'scheduler_name' in new_conf:
name = new_conf['scheduler_name']
else:
name = instance_name
self.name = name
# Set my own process title
self.set_proctitle(self.name)
logger.info("[%s] Received a new configuration, containing: ",
self.name)
for key in new_conf:
logger.info("[%s] - %s", self.name, key)
logger.info("[%s] configuration identifiers: %s (%s)", self.name,
new_conf['conf_uuid'], new_conf['push_flavor'])
# Tag the conf with our data
self.conf = conf
self.conf.push_flavor = new_conf['push_flavor']
self.conf.alignak_name = new_conf['alignak_name']
self.conf.instance_name = instance_name
self.conf.skip_initial_broks = new_conf['skip_initial_broks']
self.conf.accept_passive_unknown_check_results = \
new_conf['accept_passive_unknown_check_results']
self.cur_conf = conf
self.override_conf = override_conf
self.modules = unserialize(modules, True)
self.satellites = satellites
# Now We create our pollers, reactionners and brokers
for sat_type in ['pollers', 'reactionners', 'brokers']:
if sat_type not in satellites:
continue
for sat_id in satellites[sat_type]:
# Must look if we already have it
sats = getattr(self, sat_type)
sat = satellites[sat_type][sat_id]
sats[sat_id] = sat
if sat['name'] in override_conf['satellitemap']:
sat = dict(sat) # make a copy
sat.update(override_conf['satellitemap'][sat['name']])
proto = 'http'
if sat['use_ssl']:
proto = 'https'
uri = '%s://%s:%s/' % (proto, sat['address'], sat['port'])
sats[sat_id]['uri'] = uri
sats[sat_id]['con'] = None
sats[sat_id]['running_id'] = 0
sats[sat_id]['last_connection'] = 0
sats[sat_id]['connection_attempt'] = 0
sats[sat_id]['max_failed_connections'] = 3
setattr(self, sat_type, sats)
logger.debug("We have our %s: %s ", sat_type,
satellites[sat_type])
logger.info("We have our %s:", sat_type)
for daemon in satellites[sat_type].values():
logger.info(" - %s ", daemon['name'])
# First mix conf and override_conf to have our definitive conf
for prop in self.override_conf:
val = self.override_conf[prop]
setattr(self.conf, prop, val)
if self.conf.use_timezone != '':
logger.info("Setting our timezone to %s",
str(self.conf.use_timezone))
os.environ['TZ'] = self.conf.use_timezone
time.tzset()
self.do_load_modules(self.modules)
logger.info("Loading configuration.")
self.conf.explode_global_conf() # pylint: disable=E1101
# we give sched it's conf
self.sched.reset()
self.sched.load_conf(self.conf)
self.sched.load_satellites(self.pollers, self.reactionners,
self.brokers)
# We must update our Config dict macro with good value
# from the config parameters
self.sched.conf.fill_resource_macros_names_macros()
# Creating the Macroresolver Class & unique instance
m_solver = MacroResolver()
m_solver.init(self.conf)
# self.conf.dump()
# self.conf.quick_debug()
# Now create the external commands manager
# We are an applyer: our role is not to dispatch commands, but to apply them
ecm = ExternalCommandManager(self.conf, 'applyer', self.sched)
# Scheduler needs to know about this external command manager to use it if necessary
self.sched.set_external_commands_manager(ecm)
# Update External Commands Manager
self.sched.external_commands_manager.accept_passive_unknown_check_results = \
self.sched.conf.accept_passive_unknown_check_results
# We clear our schedulers managed (it's us :) )
# and set ourselves in it
self.schedulers = {self.conf.uuid: self.sched} # pylint: disable=E1101
# Ok now we can load the retention data
self.sched.retention_load()
# Create brok new conf
brok = Brok({'type': 'new_conf', 'data': {}})
self.sched.add_brok(brok)
def what_i_managed(self):
# pylint: disable=no-member
"""Get my managed dict (instance id and push_flavor)
:return: dict containing instance_id key and push flavor value
:rtype: dict
"""
if hasattr(self, 'conf'):
return {self.conf.uuid: self.conf.push_flavor} # pylint: disable=E1101
return {}
def clean_previous_run(self):
"""Clean variables from previous configuration
:return: None
"""
# Clean all lists
self.pollers.clear()
self.reactionners.clear()
self.brokers.clear()
def main(self):
"""Main function for Scheduler, launch after the init::
* Init daemon
* Load module manager
* Launch main loop
* Catch any Exception that occurs
:return: None
"""
try:
self.setup_alignak_logger()
# Look if we are enabled or not. If ok, start the daemon mode
self.look_for_early_exit()
# todo:
# This function returns False if some problem is detected during initialization
# (eg. communication port not free)
# Perharps we should stop the initialization process and exit?
if not self.do_daemon_init_and_start():
return
self.load_modules_manager(self.name)
self.uri = self.http_daemon.uri
logger.info("[Scheduler] General interface is at: %s", self.uri)
self.do_mainloop()
except Exception:
self.print_unrecoverable(traceback.format_exc())
raise
class Alignak(BaseSatellite):
"""Scheduler class. Referenced as "app" in most Interface
"""
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default=7768),
'local_log': PathProp(default='schedulerd.log'),
})
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.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
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 compensate_system_time_change(self, difference):
"""Compensate a system time change of difference for all hosts/services/checks/notifs
:param difference: difference in seconds
:type difference: int
:return: None
"""
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 host in self.sched.hosts:
host.compensate_system_time_change(difference)
for serv in self.sched.services:
serv.compensate_system_time_change(difference)
# Now all checks and actions
for chk in self.sched.checks.values():
# Already launch checks should not be touch
if chk.status == 'scheduled' and chk.t_to_go is not None:
t_to_go = chk.t_to_go
ref = chk.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:
chk.state = 'waitconsume'
chk.exit_status = 2
chk.output = '(Error: there is no available check time after time change!)'
chk.check_time = time.time()
chk.execution_time = 0
else:
chk.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for act in self.sched.actions.values():
# Already launch checks should not be touch
if act.status == 'scheduled':
t_to_go = act.t_to_go
# Event handler do not have ref
ref = getattr(act, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if act.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
act.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:
act.state = 'waitconsume'
act.exit_status = 2
act.output = '(Error: there is no available check time after time change!)'
act.check_time = time.time()
act.execution_time = 0
else:
act.t_to_go = new_t
def manage_signal(self, sig, frame):
"""Manage signals caught by the daemon
signal.SIGUSR1 : dump_memory
signal.SIGUSR2 : dump_object (nothing)
signal.SIGTERM, signal.SIGINT : terminate process
:param sig: signal caught by daemon
:type sig: str
:param frame: current stack frame
:type frame:
:return: None
TODO: Refactor with Daemon one
"""
logger.warning("%s > Received a SIGNAL %s", process.current_process(),
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):
"""Scheduler loop turn
Basically wait initial conf and run
:return: None
"""
# 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):
"""Setup new conf received for scheduler
:return: None
"""
with self.conf_lock:
new_c = self.new_conf
conf_raw = new_c['conf']
override_conf = new_c['override_conf']
modules = new_c['modules']
satellites = new_c['satellites']
instance_name = new_c['instance_name']
push_flavor = new_c['push_flavor']
skip_initial_broks = new_c['skip_initial_broks']
accept_passive_unknown_chk_res = new_c[
'accept_passive_unknown_check_results']
api_key = new_c['api_key']
secret = new_c['secret']
http_proxy = new_c['http_proxy']
statsd_host = new_c['statsd_host']
statsd_port = new_c['statsd_port']
statsd_prefix = new_c['statsd_prefix']
statsd_enabled = new_c['statsd_enabled']
# 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)
t00 = time.time()
conf = cPickle.loads(conf_raw)
logger.debug("Conf received at %d. Unserialized in %d secs", t00,
time.time() - t00)
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_chk_res
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:
# pylint: disable=E1101
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
poll = satellites['pollers'][pol_id]
self.pollers[pol_id] = poll
if poll['name'] in override_conf['satellitemap']:
poll = dict(poll) # make a copy
poll.update(override_conf['satellitemap'][poll['name']])
proto = 'http'
if poll['use_ssl']:
proto = 'https'
uri = '%s://%s:%s/' % (proto, poll['address'], poll['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 poll['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:
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.do_load_modules(self.modules)
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_solver = MacroResolver()
m_solver.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
ecm = ExternalCommandManager(self.conf, 'applyer')
# Scheduler need to know about external command to
# activate it if necessary
self.sched.load_external_command(ecm)
# External command need the sched because he can raise checks
ecm.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}
def what_i_managed(self):
"""Get my managed dict (instance id and push_flavor)
:return: dict containing instance_id key and push flavor value
:rtype: dict
"""
if hasattr(self, 'conf'):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
def main(self):
"""Main function for Scheduler, launch after the init::
* Init daemon
* Load module manager
* Launch main loop
* Catch any Exception that occurs
:return: None
"""
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.do_daemon_init_and_start()
self.load_modules_manager()
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
class Alignak(BaseSatellite):
"""Scheduler class. Referenced as "app" in most Interface
"""
properties = BaseSatellite.properties.copy()
properties.update({
'pidfile': PathProp(default='schedulerd.pid'),
'port': IntegerProp(default=7768),
'local_log': PathProp(default='schedulerd.log'),
})
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.http_interface = SchedulerInterface(self)
self.sched = Scheduler(self)
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 compensate_system_time_change(self, difference):
"""Compensate a system time change of difference for all hosts/services/checks/notifs
:param difference: difference in seconds
:type difference: int
:return: None
"""
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 host in self.sched.hosts:
host.compensate_system_time_change(difference)
for serv in self.sched.services:
serv.compensate_system_time_change(difference)
# Now all checks and actions
for chk in self.sched.checks.values():
# Already launch checks should not be touch
if chk.status == 'scheduled' and chk.t_to_go is not None:
t_to_go = chk.t_to_go
ref = chk.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:
chk.state = 'waitconsume'
chk.exit_status = 2
chk.output = '(Error: there is no available check time after time change!)'
chk.check_time = time.time()
chk.execution_time = 0
else:
chk.t_to_go = new_t
ref.next_chk = new_t
# Now all checks and actions
for act in self.sched.actions.values():
# Already launch checks should not be touch
if act.status == 'scheduled':
t_to_go = act.t_to_go
# Event handler do not have ref
ref = getattr(act, 'ref', None)
new_t = max(0, t_to_go + difference)
# Notification should be check with notification_period
if act.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
act.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:
act.state = 'waitconsume'
act.exit_status = 2
act.output = '(Error: there is no available check time after time change!)'
act.check_time = time.time()
act.execution_time = 0
else:
act.t_to_go = new_t
def manage_signal(self, sig, frame):
"""Manage signals caught by the daemon
signal.SIGUSR1 : dump_memory
signal.SIGUSR2 : dump_object (nothing)
signal.SIGTERM, signal.SIGINT : terminate process
:param sig: signal caught by daemon
:type sig: str
:param frame: current stack frame
:type frame:
:return: None
TODO: Refactor with Daemon one
"""
logger.warning("%s > Received a SIGNAL %s", process.current_process(), 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):
"""Scheduler loop turn
Basically wait initial conf and run
:return: None
"""
# 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):
"""Setup new conf received for scheduler
:return: None
"""
with self.conf_lock:
new_c = self.new_conf
conf_raw = new_c['conf']
override_conf = new_c['override_conf']
modules = new_c['modules']
satellites = new_c['satellites']
instance_name = new_c['instance_name']
push_flavor = new_c['push_flavor']
skip_initial_broks = new_c['skip_initial_broks']
accept_passive_unknown_chk_res = new_c['accept_passive_unknown_check_results']
api_key = new_c['api_key']
secret = new_c['secret']
http_proxy = new_c['http_proxy']
statsd_host = new_c['statsd_host']
statsd_port = new_c['statsd_port']
statsd_prefix = new_c['statsd_prefix']
statsd_enabled = new_c['statsd_enabled']
# 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)
t00 = time.time()
conf = cPickle.loads(conf_raw)
logger.debug("Conf received at %d. Unserialized in %d secs", t00, time.time() - t00)
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_chk_res
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
poll = satellites['pollers'][pol_id]
self.pollers[pol_id] = poll
if poll['name'] in override_conf['satellitemap']:
poll = dict(poll) # make a copy
poll.update(override_conf['satellitemap'][poll['name']])
proto = 'http'
if poll['use_ssl']:
proto = 'https'
uri = '%s://%s:%s/' % (proto, poll['address'], poll['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 poll['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:
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()
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_solver = MacroResolver()
m_solver.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
ecm = ExternalCommandManager(self.conf, 'applyer')
# Scheduler need to know about external command to
# activate it if necessary
self.sched.load_external_command(ecm)
# External command need the sched because he can raise checks
ecm.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}
def what_i_managed(self):
"""Get my managed dict (instance id and push_flavor)
:return: dict containing instance_id key and push flavor value
:rtype: dict
"""
if hasattr(self, 'conf'):
return {self.conf.instance_id: self.conf.push_flavor}
else:
return {}
def main(self):
"""Main function for Scheduler, launch after the init::
* Init daemon
* Load module manager
* Launch main loop
* Catch any Exception that occurs
:return: None
"""
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.do_daemon_init_and_start()
self.load_modules_manager()
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