def apply_async(self, *args, **kwargs):
"""
A wrapper around the Celery apply_async method. It allows us to accept a few more
parameters than Celery does for our own purposes, listed below. It also allows us
to create and update task status which can be used to track status of this task
during it's lifetime.
:param queue: The queue that the task has been placed into (optional, defaults to
the general Celery queue.)
:type queue: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching for
tasks by tag
:type tags: list
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
routing_key = kwargs.get('routing_key',
defaults.NAMESPACES['CELERY']['DEFAULT_ROUTING_KEY'].default)
tags = kwargs.pop('tags', [])
async_result = super(Task, self).apply_async(*args, **kwargs)
async_result.tags = tags
# Create a new task status with the task id and tags.
task_status = TaskStatus(
task_id=async_result.id, task_type=self.name,
state=constants.CALL_WAITING_STATE, worker_name=routing_key, tags=tags)
# To avoid the race condition where __call__ method below is called before
# this change is propagated to all db nodes, using an 'upsert' here and setting
# the task state to 'waiting' only on an insert.
task_status.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
return async_result
def apply_async(self, *args, **kwargs):
"""
A wrapper around the PulpTask apply_async method. It allows us to accept a few more
parameters than Celery does for our own purposes, listed below. It also allows us
to create and update task status which can be used to track status of this task
during it's lifetime.
:param queue: The queue that the task has been placed into (optional, defaults to
the general Celery queue.)
:type queue: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching for
tasks by tag
:type tags: list
:param group_id: The id that identifies which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
routing_key = kwargs.get('routing_key',
defaults.NAMESPACES['CELERY']['DEFAULT_ROUTING_KEY'].default)
tag_list = kwargs.pop('tags', [])
group_id = kwargs.pop('group_id', None)
async_result = super(Task, self).apply_async(*args, **kwargs)
async_result.tags = tag_list
# Create a new task status with the task id and tags.
task_status = TaskStatus(
task_id=async_result.id, task_type=self.name,
state=constants.CALL_WAITING_STATE, worker_name=routing_key, tags=tag_list,
group_id=group_id)
# To avoid the race condition where __call__ method below is called before
# this change is propagated to all db nodes, using an 'upsert' here and setting
# the task state to 'waiting' only on an insert.
task_status.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
return async_result
def test_save_update_with_set_on_insert(self):
"""
Test the save method with set on insert arguments when the object is already in the
database.
"""
task_id = str(uuid4())
worker_name = 'worker_name'
tags = ['tag_1', 'tag_2']
state = constants.CALL_ACCEPTED_STATE
spawned_tasks = ['foo']
error = {'error': 'some_error'}
progress_report = {'what do we want?': 'progress!', 'when do we want it?': 'now!'}
task_type = 'some.task'
old_start_time = start_time = datetime.now()
finish_time = start_time + timedelta(minutes=5)
start_time = dateutils.format_iso8601_datetime(start_time)
finish_time = dateutils.format_iso8601_datetime(finish_time)
result = None
ts = TaskStatus(
task_id, worker_name, tags, state, spawned_tasks=spawned_tasks, error=error,
progress_report=progress_report, task_type=task_type, start_time=start_time,
finish_time=finish_time, result=result)
# Put the object in the database, and then change some of it settings.
ts.save()
new_worker_name = 'a different_worker'
new_state = constants.CALL_SUSPENDED_STATE
new_start_time = old_start_time + timedelta(minutes=10)
new_start_time = dateutils.format_iso8601_datetime(new_start_time)
ts.worker_name = new_worker_name
ts.state = new_state
ts.start_time = new_start_time
# This should update the worker_name on ts in the database, but should not update the state
# or start_time
ts.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
ts = TaskStatus.objects()
# There should only be one TaskStatus in the db
self.assertEqual(len(ts), 1)
ts = ts[0]
# Make sure all the attributes are correct
self.assertEqual(ts['task_id'], task_id)
# Queue should have been updated
self.assertEqual(ts['worker_name'], new_worker_name)
self.assertEqual(ts['tags'], tags)
# state should not have been updated
self.assertEqual(ts['state'], state)
self.assertEqual(ts['error'], error)
self.assertEqual(ts['spawned_tasks'], spawned_tasks)
self.assertEqual(ts['progress_report'], progress_report)
self.assertEqual(ts['task_type'], task_type)
# start_time should not have been updated
self.assertEqual(ts['start_time'], start_time)
self.assertEqual(ts['finish_time'], finish_time)
self.assertEqual(ts['result'], result)
# These are always None
self.assertEqual(ts['traceback'], None)
self.assertEqual(ts['exception'], None)
def apply_async_with_reservation(self, resource_type, resource_id, *args, **kwargs):
"""
This method allows the caller to schedule the ReservedTask to run asynchronously just like
Celery's apply_async(), while also making the named resource. No two tasks that claim the
same resource reservation can execute concurrently. It accepts type and id of a resource
and combines them to form a resource id.
This does not dispatch the task directly, but instead promises to dispatch it later by
encapsulating the desired task through a call to a _queue_reserved_task task. See the
docblock on _queue_reserved_task for more information on this.
This method creates a TaskStatus as a placeholder for later updates. Pulp expects to poll
on a task just after calling this method, so a TaskStatus entry needs to exist for it
before it returns.
For a list of parameters accepted by the *args and **kwargs parameters, please see the
docblock for the apply_async() method.
:param resource_type: A string that identifies type of a resource
:type resource_type: basestring
:param resource_id: A string that identifies some named resource, guaranteeing that only
one task reserving this same string can happen at a time.
:type resource_id: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching
for tasks by tag
:type tags: list
:param group_id: The id to identify which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
# Form a resource_id for reservation by combining given resource type and id. This way,
# two different resources having the same id will not block each other.
resource_id = ":".join((resource_type, resource_id))
inner_task_id = str(uuid.uuid4())
task_name = self.name
tag_list = kwargs.get('tags', [])
group_id = kwargs.get('group_id', None)
# Create a new task status with the task id and tags.
task_status = TaskStatus(task_id=inner_task_id, task_type=task_name,
state=constants.CALL_WAITING_STATE, tags=tag_list,
group_id=group_id)
# To avoid the race condition where __call__ method below is called before
# this change is propagated to all db nodes, using an 'upsert' here and setting
# the task state to 'waiting' only on an insert.
task_status.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
try:
_queue_reserved_task.apply_async(
args=[task_name, inner_task_id, resource_id, args, kwargs],
queue=RESOURCE_MANAGER_QUEUE
)
except Exception:
TaskStatus.objects(task_id=task_status.task_id).update(state=constants.CALL_ERROR_STATE)
raise
return AsyncResult(inner_task_id)
def apply_async_with_reservation(self, resource_type, resource_id, *args, **kwargs):
"""
This method allows the caller to schedule the ReservedTask to run asynchronously just like
Celery's apply_async(), while also making the named resource. No two tasks that claim the
same resource reservation can execute concurrently. It accepts type and id of a resource
and combines them to form a resource id.
This does not dispatch the task directly, but instead promises to dispatch it later by
encapsulating the desired task through a call to a _queue_reserved_task task. See the
docblock on _queue_reserved_task for more information on this.
This method creates a TaskStatus as a placeholder for later updates. Pulp expects to poll
on a task just after calling this method, so a TaskStatus entry needs to exist for it
before it returns.
For a list of parameters accepted by the *args and **kwargs parameters, please see the
docblock for the apply_async() method.
:param resource_type: A string that identifies type of a resource
:type resource_type: basestring
:param resource_id: A string that identifies some named resource, guaranteeing that only
one task reserving this same string can happen at a time.
:type resource_id: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching
for tasks by tag
:type tags: list
:param group_id: The id to identify which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
# Form a resource_id for reservation by combining given resource type and id. This way,
# two different resources having the same id will not block each other.
resource_id = ":".join((resource_type, resource_id))
inner_task_id = str(uuid.uuid4())
task_name = self.name
tag_list = kwargs.get('tags', [])
group_id = kwargs.get('group_id', None)
# Create a new task status with the task id and tags.
task_status = TaskStatus(task_id=inner_task_id, task_type=task_name,
state=constants.CALL_WAITING_STATE, tags=tag_list,
group_id=group_id)
# To avoid the race condition where __call__ method below is called before
# this change is propagated to all db nodes, using an 'upsert' here and setting
# the task state to 'waiting' only on an insert.
task_status.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
try:
_queue_reserved_task.apply_async(
args=[task_name, inner_task_id, resource_id, args, kwargs],
queue=RESOURCE_MANAGER_QUEUE
)
except Exception:
TaskStatus.objects(task_id=task_status.task_id).update(state=constants.CALL_ERROR_STATE)
raise
return AsyncResult(inner_task_id)
def apply_async(self, *args, **kwargs):
"""
A wrapper around the PulpTask apply_async method. It allows us to accept a few more
parameters than Celery does for our own purposes, listed below. It also allows us
to create and update task status which can be used to track status of this task
during it's lifetime.
:param queue: The queue that the task has been placed into (optional, defaults to
the general Celery queue.)
:type queue: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching for
tasks by tag
:type tags: list
:param group_id: The id that identifies which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
if celery_version.startswith('4'):
routing_key = kwargs.get(
'routing_key',
defaults.NAMESPACES['task']['default_routing_key'].default)
else:
routing_key = kwargs.get(
'routing_key',
defaults.NAMESPACES['CELERY']['DEFAULT_ROUTING_KEY'].default)
tag_list = kwargs.pop('tags', [])
group_id = kwargs.pop('group_id', None)
try:
async_result = super(Task, self).apply_async(*args, **kwargs)
except Exception:
if 'task_id' in kwargs:
TaskStatus.objects(task_id=kwargs['task_id']).update(
state=constants.CALL_ERROR_STATE)
raise
async_result.tags = tag_list
# Create a new task status with the task id and tags.
task_status = TaskStatus(task_id=async_result.id,
task_type=self.name,
state=constants.CALL_WAITING_STATE,
worker_name=routing_key,
tags=tag_list,
group_id=group_id)
# We're now racing with __call__, on_failure and on_success, any of which may
# have completed by now. To avoid overwriting TaskStatus updates from those callbacks,
# we'll do an upsert and only touch the fields listed below if we've inserted the object.
task_status.save_with_set_on_insert(fields_to_set_on_insert=[
'state', 'start_time', 'finish_time', 'result', 'error',
'spawned_tasks', 'traceback'
])
return async_result
def apply_async(self, *args, **kwargs):
"""
A wrapper around the PulpTask apply_async method. It allows us to accept a few more
parameters than Celery does for our own purposes, listed below. It also allows us
to create and update task status which can be used to track status of this task
during it's lifetime.
:param queue: The queue that the task has been placed into (optional, defaults to
the general Celery queue.)
:type queue: basestring
:param tags: A list of tags (strings) to place onto the task, used for searching for
tasks by tag
:type tags: list
:param group_id: The id that identifies which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
if celery_version.startswith('4'):
routing_key = kwargs.get('routing_key',
defaults.NAMESPACES['task']['default_routing_key'].default)
else:
routing_key = kwargs.get('routing_key',
defaults.NAMESPACES['CELERY']['DEFAULT_ROUTING_KEY'].default)
tag_list = kwargs.pop('tags', [])
group_id = kwargs.pop('group_id', None)
try:
async_result = super(Task, self).apply_async(*args, **kwargs)
except Exception:
if 'task_id' in kwargs:
TaskStatus.objects(task_id=kwargs['task_id']).update(
state=constants.CALL_ERROR_STATE
)
raise
async_result.tags = tag_list
# Create a new task status with the task id and tags.
task_status = TaskStatus(
task_id=async_result.id, task_type=self.name,
state=constants.CALL_WAITING_STATE, worker_name=routing_key, tags=tag_list,
group_id=group_id)
# We're now racing with __call__, on_failure and on_success, any of which may
# have completed by now. To avoid overwriting TaskStatus updates from those callbacks,
# we'll do an upsert and only touch the fields listed below if we've inserted the object.
task_status.save_with_set_on_insert(fields_to_set_on_insert=[
'state', 'start_time', 'finish_time', 'result', 'error',
'spawned_tasks', 'traceback'])
return async_result
def test_save_insert_with_set_on_insert(self):
"""
Test the save method with set on insert arguments when the object is not already in the
database.
"""
task_id = str(uuid4())
worker_name = 'some_worker'
tags = ['tag_1', 'tag_2']
state = constants.CALL_RUNNING_STATE
spawned_tasks = ['foo']
error = {'error': 'some_error'}
progress_report = {'what do we want?': 'progress!', 'when do we want it?': 'now!'}
task_type = 'some.task'
start_time = datetime.now()
finish_time = start_time + timedelta(minutes=5)
start_time = dateutils.format_iso8601_datetime(start_time)
finish_time = dateutils.format_iso8601_datetime(finish_time)
result = None
ts = TaskStatus(
task_id, worker_name, tags, state, spawned_tasks=spawned_tasks, error=error,
progress_report=progress_report, task_type=task_type, start_time=start_time,
finish_time=finish_time, result=result)
# This should cause ts to be in the database
ts.save_with_set_on_insert(fields_to_set_on_insert=['state', 'start_time'])
ts = TaskStatus.objects()
# There should only be one TaskStatus in the db
self.assertEqual(len(ts), 1)
ts = ts[0]
# Make sure all the attributes are correct
self.assertEqual(ts['task_id'], task_id)
self.assertEqual(ts['worker_name'], worker_name)
self.assertEqual(ts['tags'], tags)
self.assertEqual(ts['state'], state)
self.assertEqual(ts['error'], error)
self.assertEqual(ts['spawned_tasks'], spawned_tasks)
self.assertEqual(ts['progress_report'], progress_report)
self.assertEqual(ts['task_type'], task_type)
self.assertEqual(ts['start_time'], start_time)
self.assertEqual(ts['finish_time'], finish_time)
self.assertEqual(ts['result'], result)
# These are always None
self.assertEqual(ts['traceback'], None)
self.assertEqual(ts['exception'], None)
def _apply_async_inner(self, reservation, *args, **kwargs):
"""
This method allows the caller to schedule the ReservedTask to run asynchronously just like
Celery's apply_async(), while also locking named resource(s). No two tasks that claim the
same named-resource(s) can execute concurrently.
It can accept a list-of-strings, of the form 'resource-type:resource-id'. If only
asked for one resource (ie, list-len == 1), then call _queue_reserved_task, otherwise
let _queue_reserved_task_list do the deed.
This does not dispatch the task directly, but instead promises to dispatch it later. If the
agument 'is_list' is True, the desired task is encapsualted by a call to
_queue_reserved_task_list; otherwise, by a call to _queue_reserved_task.
See the docblock on _queue_reserved_task and _queue_reserved_task_list for more
information.
This method creates a TaskStatus as a placeholder for later updates. Pulp expects to poll
on a task just after calling this method, so a TaskStatus entry needs to exist for it
before it returns.
For a list of parameters accepted by the *args and **kwargs parameters, please see the
docblock for the apply_async() method.
:param reservation: A list-of-strings that identify a set of named resources,
guaranteeing that only one task reserving any resource-ids in this
list can happen at a time.
:type reservation: list
:param tags: A list of tags (strings) to place onto the task, used for searching
for tasks by tag
:type tags: list
:param group_id: The id to identify which group of tasks a task belongs to
:type group_id: uuid.UUID
:return: An AsyncResult instance as returned by Celery's apply_async
:rtype: celery.result.AsyncResult
"""
inner_task_id = str(uuid.uuid4())
task_name = self.name
tag_list = kwargs.get('tags', [])
group_id = kwargs.get('group_id', None)
# Create a new task status with the task id and tags.
task_status = TaskStatus(task_id=inner_task_id,
task_type=task_name,
state=constants.CALL_WAITING_STATE,
tags=tag_list,
group_id=group_id)
# To avoid the race condition where __call__ method below is called before
# this change is propagated to all db nodes, using an 'upsert' here and setting
# the task state to 'waiting' only on an insert.
task_status.save_with_set_on_insert(
fields_to_set_on_insert=['state', 'start_time'])
try:
# Decide what to call based on how many reservation(s) we are being asked to make
if len(reservation) == 1:
_queue_reserved_task.apply_async(args=[
task_name, inner_task_id, reservation[0], args, kwargs
],
queue=RESOURCE_MANAGER_QUEUE)
else:
_queue_reserved_task_list.apply_async(
args=[task_name, inner_task_id, reservation, args, kwargs],
queue=RESOURCE_MANAGER_QUEUE)
except Exception:
TaskStatus.objects(task_id=task_status.task_id).update(
state=constants.CALL_ERROR_STATE)
raise
return AsyncResult(inner_task_id)
