def del_host(self, deregister=True):
if deregister:
try:
self.dbapi.unregister_wampagent(self.host)
LOG.info(
_LI('Successfully stopped wampagent with hostname '
'%(hostname)s.'), {'hostname': self.host})
except exception.WampAgentNotFound:
pass
else:
LOG.info(
_LI('Not deregistering wampagent with hostname '
'%(hostname)s.'), {'hostname': self.host})
def del_host(self, deregister=True):
if deregister:
try:
self.dbapi.unregister_conductor(self.host)
LOG.info(_LI('Successfully stopped conductor with hostname '
'%(hostname)s.'),
{'hostname': self.host})
except exception.ConductorNotFound:
pass
else:
LOG.info(_LI('Not deregistering conductor with hostname '
'%(hostname)s.'),
{'hostname': self.host})
def del_host(self, deregister=True):
if deregister:
try:
self.dbapi.unregister_wampagent(self.host)
LOG.info(_LI('Successfully stopped wampagent with hostname '
'%(hostname)s.'),
{'hostname': self.host})
except exception.WampAgentNotFound:
pass
else:
LOG.info(_LI('Not deregistering wampagent with hostname '
'%(hostname)s.'),
{'hostname': self.host})
def init_host(self):
self.dbapi = dbapi.get_instance()
self._keepalive_evt = threading.Event()
"""Event for the keepalive thread."""
self._worker_pool = greenpool.GreenPool(
size=CONF.conductor.workers_pool_size)
"""GreenPool of background workers for performing tasks async."""
try:
# Register this conductor with the cluster
cdr = self.dbapi.register_conductor({'hostname': self.host})
except exception.ConductorAlreadyRegistered:
LOG.warn(
_LW("A conductor with hostname %(hostname)s "
"was previously registered. Updating registration"),
{'hostname': self.host})
cdr = self.dbapi.register_conductor({'hostname': self.host},
update_existing=True)
self.conductor = cdr
# Spawn a dedicated greenthread for the keepalive
try:
self._spawn_worker(self._conductor_service_record_keepalive)
LOG.info(
_LI('Successfully started conductor with hostname '
'%(hostname)s.'), {'hostname': self.host})
except exception.NoFreeConductorWorker:
with excutils.save_and_reraise_exception():
LOG.critical(_LC('Failed to start keepalive'))
self.del_host()
def _handle_signal(self, signo, frame):
LOG.info(
_LI('Got signal SIGUSR1. Not deregistering on next shutdown '
'of service %(service)s on host %(host)s.'), {
'service': self.topic,
'host': self.host
})
self.deregister = False
def del_host(self, deregister=True):
self._keepalive_evt.set()
if deregister:
try:
# Inform the cluster that this conductor is shutting down.
# Note that rebalancing will not occur immediately, but when
# the periodic sync takes place.
self.dbapi.unregister_conductor(self.host)
LOG.info(
_LI('Successfully stopped conductor with hostname '
'%(hostname)s.'), {'hostname': self.host})
except exception.ConductorNotFound:
pass
else:
LOG.info(
_LI('Not deregistering conductor with hostname '
'%(hostname)s.'), {'hostname': self.host})
# Waiting here to give workers the chance to finish. This has the
# benefit of releasing locks workers placed on nodes, as well as
# having work complete normally.
self._worker_pool.waitall()
def stop(self):
try:
self.rpcserver.stop()
self.rpcserver.wait()
except Exception as e:
LOG.exception(_LE('Service error occurred when stopping the '
'RPC server. Error: %s'), e)
try:
self.manager.del_host(deregister=self.deregister)
except Exception as e:
LOG.exception(_LE('Service error occurred when cleaning up '
'the RPC manager. Error: %s'), e)
super(RPCService, self).stop(graceful=True)
LOG.info(_LI('Stopped RPC server for service %(service)s on host '
'%(host)s.'),
{'service': self.topic, 'host': self.host})
def stop(self):
try:
self.rpcserver.stop()
self.rpcserver.wait()
except Exception as e:
LOG.exception(_LE('Service error occurred when stopping the '
'RPC server. Error: %s'), e)
try:
self.manager.del_host(deregister=self.deregister)
except Exception as e:
LOG.exception(_LE('Service error occurred when cleaning up '
'the RPC manager. Error: %s'), e)
super(RPCService, self).stop(graceful=True)
LOG.info(_LI('Stopped RPC server for service %(service)s on host '
'%(host)s.'),
{'service': self.topic, 'host': self.host})
def start(self):
super(RPCService, self).start()
admin_context = context.RequestContext('admin', 'admin', is_admin=True)
target = messaging.Target(topic=self.topic, server=self.host)
endpoints = [self.manager]
serializer = objects_base.IotronicObjectSerializer()
self.rpcserver = rpc.get_server(target, endpoints, serializer)
self.rpcserver.start()
self.handle_signal()
self.manager.init_host()
self.tg.add_dynamic_timer(
self.manager.periodic_tasks,
periodic_interval_max=cfg.CONF.periodic_interval,
context=admin_context)
LOG.info(_LI('Created RPC server for service %(service)s on host '
'%(host)s.'),
{'service': self.topic, 'host': self.host})
def start(self):
super(RPCService, self).start()
admin_context = context.RequestContext('admin', 'admin', is_admin=True)
target = messaging.Target(topic=self.topic, server=self.host)
endpoints = [self.manager]
serializer = objects_base.IotronicObjectSerializer()
self.rpcserver = rpc.get_server(target, endpoints, serializer)
self.rpcserver.start()
self.handle_signal()
self.manager.init_host()
self.tg.add_dynamic_timer(
self.manager.periodic_tasks,
periodic_interval_max=cfg.CONF.periodic_interval,
context=admin_context)
LOG.info(_LI('Created RPC server for service %(service)s on host '
'%(host)s.'),
{'service': self.topic, 'host': self.host})
def destroy_node(self, context, node_id):
"""Delete a node.
:param context: request context.
:param node_id: node id or uuid.
:raises: NodeLocked if node is locked by another conductor.
:raises: NodeNotConnected if the node is not connected.
"""
with task_manager.acquire(context, node_id) as task:
node = task.node
r = WampResponse()
r.clearConfig()
response = self.wamp.rpc_call(
'stack4things.' + node.uuid + '.configure', r.getResponse())
if response['result'] == 0:
node.destroy()
LOG.info(_LI('Successfully deleted node %(node)s.'),
{'node': node.uuid})
else:
raise exception.NodeNotConnected(node=node.uuid)
def _handle_signal(self, signo, frame):
LOG.info(_LI('Got signal SIGUSR1. Not deregistering on next shutdown '
'of service %(service)s on host %(host)s.'),
{'service': self.topic, 'host': self.host})
self.deregister = False
def node_power_action(task, new_state):
"""Change power state or reset for a node.
Perform the requested power action if the transition is required.
:param task: a TaskManager instance containing the node to act on.
:param new_state: Any power state from iotronic.common.states. If the
state is 'REBOOT' then a reboot will be attempted, otherwise
the node power state is directly set to 'state'.
:raises: InvalidParameterValue when the wrong state is specified
or the wrong driver info is specified.
:raises: other exceptions by the node's power driver if something
wrong occurred during the power action.
"""
node = task.node
target_state = states.POWER_ON if new_state == states.REBOOT else new_state
if new_state != states.REBOOT:
try:
curr_state = task.driver.power.get_power_state(task)
except Exception as e:
with excutils.save_and_reraise_exception():
node['last_error'] = _(
"Failed to change power state to '%(target)s'. "
"Error: %(error)s") % {
'target': new_state,
'error': e
}
node['target_power_state'] = states.NOSTATE
node.save()
if curr_state == new_state:
# Neither the iotronic service nor the hardware has erred. The
# node is, for some reason, already in the requested state,
# though we don't know why. eg, perhaps the user previously
# requested the node POWER_ON, the network delayed those IPMI
# packets, and they are trying again -- but the node finally
# responds to the first request, and so the second request
# gets to this check and stops.
# This isn't an error, so we'll clear last_error field
# (from previous operation), log a warning, and return.
node['last_error'] = None
# NOTE(dtantsur): under rare conditions we can get out of sync here
node['power_state'] = new_state
node['target_power_state'] = states.NOSTATE
node.save()
LOG.warn(
_LW("Not going to change_node_power_state because "
"current state = requested state = '%(state)s'."),
{'state': curr_state})
return
if curr_state == states.ERROR:
# be optimistic and continue action
LOG.warn(_LW("Driver returns ERROR power state for node %s."),
node.uuid)
# Set the target_power_state and clear any last_error, if we're
# starting a new operation. This will expose to other processes
# and clients that work is in progress.
if node['target_power_state'] != target_state:
node['target_power_state'] = target_state
node['last_error'] = None
node.save()
# take power action
try:
if new_state != states.REBOOT:
task.driver.power.set_power_state(task, new_state)
else:
task.driver.power.reboot(task)
except Exception as e:
with excutils.save_and_reraise_exception():
node['last_error'] = _(
"Failed to change power state to '%(target)s'. "
"Error: %(error)s") % {
'target': target_state,
'error': e
}
else:
# success!
node['power_state'] = target_state
LOG.info(
_LI('Successfully set node %(node)s power state to '
'%(state)s.'), {
'node': node.uuid,
'state': target_state
})
finally:
node['target_power_state'] = states.NOSTATE
node.save()