def restore(self, data):
"""Restore state of this service from datafile"""
# The state of a service is more easier than for jobs. There are just a few things we want to guarantee:
# 1. If service instances are up, they can continue to be up. We'll just start monitoring from where we left off.
# 2. Failures are maintained and have to be cleared.
# Start our machine from where it left off
self.machine.state = state.named_event_by_name(Service.STATE_DOWN, data['state'])
if self.machine.state in (Service.STATE_DOWN, Service.STATE_FAILED):
self.event_recorder.emit_info("restored")
return
# Restore all the instances
# We're going to just indicate they are up and start a monitor
for instance in data['instances']:
try:
node = self.node_pool[instance['node']]
except KeyError:
log.error("Failed to find node %s in pool for %s", instance['node'], self.name)
continue
service_instance = self._create_instance(node, instance['instance_number'])
service_instance.machine.state = ServiceInstance.STATE_MONITORING
service_instance._run_monitor()
self.instances.sort(key=lambda i:i.instance_number)
self.event_recorder.emit_info("restored")
def restore_state(self, state_data):
self.id = state_data['id']
self.machine.state = state.named_event_by_name(self.STATE_SCHEDULED, state_data['state'])
self.run_time = state_data['run_time']
self.start_time = state_data['start_time']
self.end_time = state_data['end_time']
self.rendered_command = state_data['command']
# We were running when the state file was built, so we have no idea
# what happened now.
if self.is_running:
self.machine.transition('fail_unknown')
def from_state(
cls, state_data, parent_context, output_path,
job_run_node, cleanup=False,
):
"""Restore the state of this ActionRun from a serialized state."""
pool_repo = node.NodePoolRepository.get_instance()
# Support state from older version
if 'id' in state_data:
job_run_id, action_name = state_data['id'].rsplit('.', 1)
else:
job_run_id = state_data['job_run_id']
action_name = state_data['action_name']
job_run_node = pool_repo.get_node(
state_data.get('node_name'), job_run_node,
)
rendered_command = state_data.get('rendered_command')
run = cls(
job_run_id,
action_name,
job_run_node,
parent_context=parent_context,
output_path=output_path,
rendered_command=rendered_command,
bare_command=state_data['command'],
cleanup=cleanup,
start_time=state_data['start_time'],
end_time=state_data['end_time'],
run_state=state.named_event_by_name(
cls.STATE_SCHEDULED, state_data['state'],
),
exit_status=state_data.get('exit_status'),
)
# Transition running to fail unknown because exit status was missed
if run.is_running:
run._done('fail_unknown')
if run.is_starting:
run.fail(None)
return run
def from_state(cls, state_data, parent_context, output_path,
job_run_node, cleanup=False):
"""Restore the state of this ActionRun from a serialized state."""
pool_repo = node.NodePoolRepository.get_instance()
# Support state from older version
if 'id' in state_data:
job_run_id, action_name = state_data['id'].rsplit('.', 1)
else:
job_run_id = state_data['job_run_id']
action_name = state_data['action_name']
job_run_node = pool_repo.get_node(
state_data.get('node_name'), job_run_node)
rendered_command = state_data.get('rendered_command')
run = cls(
job_run_id,
action_name,
job_run_node,
parent_context=parent_context,
output_path=output_path,
rendered_command=rendered_command,
bare_command=state_data['command'],
cleanup=cleanup,
start_time=state_data['start_time'],
end_time=state_data['end_time'],
run_state=state.named_event_by_name(
cls.STATE_SCHEDULED, state_data['state']),
exit_status=state_data.get('exit_status')
)
# Transition running to fail unknown because exit status was missed
if run.is_running:
run._done('fail_unknown')
if run.is_starting:
run.fail(None)
return run
def test(self):
assert_equal(state.named_event_by_name(self.start, "c"), self.end)
def test_match(self):
assert_equal(state.named_event_by_name(self.start, "c"), self.end)
