class Task(object):
"""Task basic class.
When called tasks apply the :meth:`run` method. This method must
be defined by all tasks (that is unless the :meth:`__call__` method
is overridden).
"""
__trace__ = None
__v2_compat__ = False # set by old basic in celery.task.basic
ErrorMail = ErrorMail
MaxRetriesExceededError = MaxRetriesExceededError
#: Execution strategy used, or the qualified name of one.
Strategy = 'celery.worker.strategy:default'
#: This is the instance bound to if the task is a method of a class.
__self__ = None
#: The application instance associated with this task class.
_app = None
#: Name of the task.
name = None
#: If :const:`True` the task is an abstract basic class.
abstract = True
#: If disabled the worker will not forward magic keyword arguments.
#: Deprecated and scheduled for removal in v4.0.
accept_magic_kwargs = False
#: Maximum number of retries before giving up. If set to :const:`None`,
#: it will **never** stop retrying.
max_retries = 3
#: Default time in seconds before a retry of the task should be
#: executed. 3 minutes by default.
default_retry_delay = 3 * 60
#: Rate limit for this task type. Examples: :const:`None` (no rate
#: limit), `'100/s'` (hundred tasks a second), `'100/m'` (hundred tasks
#: a minute),`'100/h'` (hundred tasks an hour)
rate_limit = None
#: If enabled the worker will not store task state and return values
#: for this task. Defaults to the :setting:`CELERY_IGNORE_RESULT`
#: setting.
ignore_result = None
#: If enabled the request will keep track of subtasks started by
#: this task, and this information will be sent with the result
#: (``result.children``).
trail = True
#: If enabled the worker will send monitoring events related to
#: this task (but only if the worker is configured to send
#: task related events).
#: Note that this has no effect on the task-failure event case
#: where a task is not registered (as it will have no task class
#: to check this flag).
send_events = True
#: When enabled errors will be stored even if the task is otherwise
#: configured to ignore results.
store_errors_even_if_ignored = None
#: If enabled an email will be sent to :setting:`ADMINS` whenever a task
#: of this type fails.
send_error_emails = None
#: The name of a serializer that are registered with
#: :mod:`kombu.serialization.registry`. Default is `'pickle'`.
serializer = None
#: Hard time limit.
#: Defaults to the :setting:`CELERYD_TASK_TIME_LIMIT` setting.
time_limit = None
#: Soft time limit.
#: Defaults to the :setting:`CELERYD_TASK_SOFT_TIME_LIMIT` setting.
soft_time_limit = None
#: The result store backend used for this task.
backend = None
#: If disabled this task won't be registered automatically.
autoregister = True
#: If enabled the task will report its status as 'started' when the task
#: is executed by a worker. Disabled by default as the normal behaviour
#: is to not report that level of granularity. Tasks are either pending,
#: finished, or waiting to be retried.
#:
#: Having a 'started' status can be useful for when there are long
#: running tasks and there is a need to report which task is currently
#: running.
#:
#: The application default can be overridden using the
#: :setting:`CELERY_TRACK_STARTED` setting.
track_started = None
#: When enabled messages for this task will be acknowledged **after**
#: the task has been executed, and not *just before* which is the
#: default behavior.
#:
#: Please note that this means the task may be executed twice if the
#: worker crashes mid execution (which may be acceptable for some
#: applications).
#:
#: The application default can be overridden with the
#: :setting:`CELERY_ACKS_LATE` setting.
acks_late = None
#: Tuple of expected exceptions.
#:
#: These are errors that are expected in normal operation
#: and that should not be regarded as a real error by the worker.
#: Currently this means that the state will be updated to an error
#: state, but the worker will not log the event as an error.
throws = ()
#: Default task expiry time.
expires = None
#: Some may expect a request to exist even if the task has not been
#: called. This should probably be deprecated.
_default_request = None
_exec_options = None
__bound__ = False
from_config = (
('send_error_emails', 'CELERY_SEND_TASK_ERROR_EMAILS'),
('serializer', 'CELERY_TASK_SERIALIZER'),
('rate_limit', 'CELERY_DEFAULT_RATE_LIMIT'),
('track_started', 'CELERY_TRACK_STARTED'),
('acks_late', 'CELERY_ACKS_LATE'),
('ignore_result', 'CELERY_IGNORE_RESULT'),
('store_errors_even_if_ignored',
'CELERY_STORE_ERRORS_EVEN_IF_IGNORED'),
)
_backend = None # set by backend property.
__bound__ = False
# - Tasks are lazily bound, so that configuration is not set
# - until the task is actually used
@classmethod
def bind(self, app):
was_bound, self.__bound__ = self.__bound__, True
self._app = app
conf = app.conf
self._exec_options = None # clear option cache
for attr_name, config_name in self.from_config:
if getattr(self, attr_name, None) is None:
setattr(self, attr_name, conf[config_name])
if self.accept_magic_kwargs is None:
self.accept_magic_kwargs = app.accept_magic_kwargs
# decorate with annotations from config.
if not was_bound:
self.annotate()
from celery.utils.threads import LocalStack
self.request_stack = LocalStack()
# PeriodicTask uses this to add itself to the PeriodicTask schedule.
self.on_bound(app)
return app
@classmethod
def on_bound(self, app):
"""This method can be defined to do additional actions when the
task class is bound to an app."""
pass
@classmethod
def _get_app(self):
if self._app is None:
self._app = current_app
if not self.__bound__:
# The app property's __set__ method is not called
# if Task.app is set (on the class), so must bind on use.
self.bind(self._app)
return self._app
app = class_property(_get_app, bind)
@classmethod
def annotate(self):
for d in resolve_all_annotations(self.app.annotations, self):
for key, value in items(d):
if key.startswith('@'):
self.add_around(key[1:], value)
else:
setattr(self, key, value)
@classmethod
def add_around(self, attr, around):
orig = getattr(self, attr)
if getattr(orig, '__wrapped__', None):
orig = orig.__wrapped__
meth = around(orig)
meth.__wrapped__ = orig
setattr(self, attr, meth)
def __call__(self, *args, **kwargs):
_task_stack.push(self)
self.push_request()
try:
# add self if this is a bound task
if self.__self__ is not None:
return self.run(self.__self__, *args, **kwargs)
return self.run(*args, **kwargs)
finally:
self.pop_request()
_task_stack.pop()
def __reduce__(self):
# - tasks are pickled into the name of the task only, and the reciever
# - simply grabs it from the local registry.
# - in later versions the module of the task is also included,
# - and the receiving side tries to import that module so that
# - it will work even if the task has not been registered.
mod = type(self).__module__
mod = mod if mod and mod in sys.modules else None
return (_unpickle_task_v2, (self.name, mod), None)
def run(self, *args, **kwargs):
"""The body of the task executed by workers."""
raise NotImplementedError('Tasks must define the run method.')
def start_strategy(self, app, consumer, **kwargs):
return instantiate(self.Strategy, self, app, consumer, **kwargs)
def delay(self, *args, **kwargs):
"""Star argument version of :meth:`apply_async`.
Does not support the extra options enabled by :meth:`apply_async`.
:param \*args: positional arguments passed on to the task.
:param \*\*kwargs: keyword arguments passed on to the task.
:returns :class:`celery.result.AsyncResult`:
"""
return self.apply_async(args, kwargs)
def apply_async(self,
args=None,
kwargs=None,
task_id=None,
producer=None,
link=None,
link_error=None,
**options):
"""Apply tasks asynchronously by sending a message.
:keyword args: The positional arguments to pass on to the
task (a :class:`list` or :class:`tuple`).
:keyword kwargs: The keyword arguments to pass on to the
task (a :class:`dict`)
:keyword countdown: Number of seconds into the future that the
task should execute. Defaults to immediate
execution.
:keyword eta: A :class:`~datetime.datetime` object describing
the absolute time and date of when the task should
be executed. May not be specified if `countdown`
is also supplied.
:keyword expires: Either a :class:`int`, describing the number of
seconds, or a :class:`~datetime.datetime` object
that describes the absolute time and date of when
the task should expire. The task will not be
executed after the expiration time.
:keyword connection: Re-use existing broker connection instead
of establishing a new one.
:keyword retry: If enabled sending of the task message will be retried
in the event of connection loss or failure. Default
is taken from the :setting:`CELERY_TASK_PUBLISH_RETRY`
setting. Note that you need to handle the
producer/connection manually for this to work.
:keyword retry_policy: Override the retry policy used. See the
:setting:`CELERY_TASK_PUBLISH_RETRY_POLICY`
setting.
:keyword routing_key: Custom routing key used to route the task to a
worker server. If in combination with a
``queue`` argument only used to specify custom
routing keys to topic exchanges.
:keyword queue: The queue to route the task to. This must be a key
present in :setting:`CELERY_QUEUES`, or
:setting:`CELERY_CREATE_MISSING_QUEUES` must be
enabled. See :ref:`guide-routing` for more
information.
:keyword exchange: Named custom exchange to send the task to.
Usually not used in combination with the ``queue``
argument.
:keyword priority: The task priority, a number between 0 and 9.
Defaults to the :attr:`priority` attribute.
:keyword serializer: A string identifying the default
serialization method to use. Can be `pickle`,
`json`, `yaml`, `msgpack` or any custom
serialization method that has been registered
with :mod:`kombu.serialization.registry`.
Defaults to the :attr:`serializer` attribute.
:keyword compression: A string identifying the compression method
to use. Can be one of ``zlib``, ``bzip2``,
or any custom compression methods registered with
:func:`kombu.compression.register`. Defaults to
the :setting:`CELERY_MESSAGE_COMPRESSION`
setting.
:keyword link: A single, or a list of tasks to apply if the
task exits successfully.
:keyword link_error: A single, or a list of tasks to apply
if an error occurs while executing the task.
:keyword producer: :class:[email protected]` instance to use.
:keyword add_to_parent: If set to True (default) and the task
is applied while executing another task, then the result
will be appended to the parent tasks ``request.children``
attribute. Trailing can also be disabled by default using the
:attr:`trail` attribute
:keyword publisher: Deprecated alias to ``producer``.
:keyword headers: Message headers to be sent in the
task (a :class:`dict`)
:rtype :class:`celery.result.AsyncResult`: if
:setting:`CELERY_ALWAYS_EAGER` is not set, otherwise
:class:`celery.result.EagerResult`.
Also supports all keyword arguments supported by
:meth:`kombu.Producer.publish`.
.. note::
If the :setting:`CELERY_ALWAYS_EAGER` setting is set, it will
be replaced by a local :func:`apply` call instead.
"""
app = self._get_app()
if app.conf.CELERY_ALWAYS_EAGER:
return self.apply(args,
kwargs,
task_id=task_id or uuid(),
link=link,
link_error=link_error,
**options)
# add 'self' if this is a "task_method".
if self.__self__ is not None:
args = args if isinstance(args, tuple) else tuple(args or ())
args = (self.__self__, ) + args
return app.send_task(self.name,
args,
kwargs,
task_id=task_id,
producer=producer,
link=link,
link_error=link_error,
result_cls=self.AsyncResult,
**dict(self._get_exec_options(), **options))
def subtask_from_request(self,
request=None,
args=None,
kwargs=None,
queue=None,
**extra_options):
request = self.request if request is None else request
args = request.args if args is None else args
kwargs = request.kwargs if kwargs is None else kwargs
limit_hard, limit_soft = request.timelimit or (None, None)
options = {
'task_id': request.id,
'link': request.callbacks,
'link_error': request.errbacks,
'group_id': request.group,
'chord': request.chord,
'soft_time_limit': limit_soft,
'time_limit': limit_hard,
'reply_to': request.reply_to,
'headers': request.headers,
}
options.update({'queue': queue} if queue else (
request.delivery_info or {}))
return self.subtask(args, kwargs, options, type=self, **extra_options)
def retry(self,
args=None,
kwargs=None,
exc=None,
throw=True,
eta=None,
countdown=None,
max_retries=None,
**options):
"""Retry the task.
:param args: Positional arguments to retry with.
:param kwargs: Keyword arguments to retry with.
:keyword exc: Custom exception to report when the max restart
limit has been exceeded (default:
:exc:`~@MaxRetriesExceededError`).
If this argument is set and retry is called while
an exception was raised (``sys.exc_info()`` is set)
it will attempt to reraise the current exception.
If no exception was raised it will raise the ``exc``
argument provided.
:keyword countdown: Time in seconds to delay the retry for.
:keyword eta: Explicit time and date to run the retry at
(must be a :class:`~datetime.datetime` instance).
:keyword max_retries: If set, overrides the default retry limit for
this execution. Changes to this parameter do not propagate to
subsequent task retry attempts. A value of :const:`None`, means
"use the default", so if you want infinite retries you would
have to set the :attr:`max_retries` attribute of the task to
:const:`None` first.
:keyword time_limit: If set, overrides the default time limit.
:keyword soft_time_limit: If set, overrides the default soft
time limit.
:keyword \*\*options: Any extra options to pass on to
meth:`apply_async`.
:keyword throw: If this is :const:`False`, do not raise the
:exc:`~@Retry` exception,
that tells the worker to mark the task as being
retried. Note that this means the task will be
marked as failed if the task raises an exception,
or successful if it returns.
:raises celery.exceptions.Retry: To tell the worker that
the task has been re-sent for retry. This always happens,
unless the `throw` keyword argument has been explicitly set
to :const:`False`, and is considered normal operation.
**Example**
.. code-block:: python
>>> from imaginary_twitter_lib import Twitter
>>> from proj.celery import app
>>> @app.task(bind=True)
... def tweet(self, auth, message):
... twitter = Twitter(oauth=auth)
... try:
... twitter.post_status_update(message)
... except twitter.FailWhale as exc:
... # Retry in 5 minutes.
... raise self.retry(countdown=60 * 5, exc=exc)
Although the task will never return above as `retry` raises an
exception to notify the worker, we use `raise` in front of the retry
to convey that the rest of the block will not be executed.
"""
request = self.request
retries = request.retries + 1
max_retries = self.max_retries if max_retries is None else max_retries
# Not in worker or emulated by (apply/always_eager),
# so just raise the original exception.
if request.called_directly:
maybe_reraise() # raise orig stack if PyErr_Occurred
raise exc or Retry('Task can be retried', None)
if not eta and countdown is None:
countdown = self.default_retry_delay
is_eager = request.is_eager
S = self.subtask_from_request(request,
args,
kwargs,
countdown=countdown,
eta=eta,
retries=retries,
**options)
if max_retries is not None and retries > max_retries:
if exc:
# first try to reraise the original exception
maybe_reraise()
# or if not in an except block then raise the custom exc.
raise exc
raise self.MaxRetriesExceededError(
"Can't retry {0}[{1}] args:{2} kwargs:{3}".format(
self.name, request.id, S.args, S.kwargs))
ret = Retry(exc=exc, when=eta or countdown)
if is_eager:
# if task was executed eagerly using apply(),
# then the retry must also be executed eagerly.
S.apply().get()
return ret
try:
S.apply_async()
except Exception as exc:
raise Reject(exc, requeue=False)
if throw:
raise ret
return ret
def apply(self,
args=None,
kwargs=None,
link=None,
link_error=None,
**options):
"""Execute this task locally, by blocking until the task returns.
:param args: positional arguments passed on to the task.
:param kwargs: keyword arguments passed on to the task.
:keyword throw: Re-raise task exceptions. Defaults to
the :setting:`CELERY_EAGER_PROPAGATES_EXCEPTIONS`
setting.
:rtype :class:`celery.result.EagerResult`:
"""
# trace imports Task, so need to import inline.
from celery.app.trace import eager_trace_task
app = self._get_app()
args = args or ()
# add 'self' if this is a bound method.
if self.__self__ is not None:
args = (self.__self__, ) + tuple(args)
kwargs = kwargs or {}
task_id = options.get('task_id') or uuid()
retries = options.get('retries', 0)
throw = app.either('CELERY_EAGER_PROPAGATES_EXCEPTIONS',
options.pop('throw', None))
# Make sure we get the task instance, not class.
task = app._tasks[self.name]
request = {
'id': task_id,
'retries': retries,
'is_eager': True,
'logfile': options.get('logfile'),
'loglevel': options.get('loglevel', 0),
'callbacks': maybe_list(link),
'errbacks': maybe_list(link_error),
'headers': options.get('headers'),
'delivery_info': {
'is_eager': True
}
}
if self.accept_magic_kwargs:
default_kwargs = {
'task_name': task.name,
'task_id': task_id,
'task_retries': retries,
'task_is_eager': True,
'logfile': options.get('logfile'),
'loglevel': options.get('loglevel', 0),
'delivery_info': {
'is_eager': True
}
}
supported_keys = fun_takes_kwargs(task.run, default_kwargs)
extend_with = dict((key, val) for key, val in items(default_kwargs)
if key in supported_keys)
kwargs.update(extend_with)
tb = None
retval, info = eager_trace_task(task,
task_id,
args,
kwargs,
app=self._get_app(),
request=request,
propagate=throw)
if isinstance(retval, ExceptionInfo):
retval, tb = retval.exception, retval.traceback
state = states.SUCCESS if info is None else info.state
return EagerResult(task_id, retval, state, traceback=tb)
def AsyncResult(self, task_id, **kwargs):
"""Get AsyncResult instance for this kind of task.
:param task_id: Task id to get result for.
"""
return self._get_app().AsyncResult(task_id,
backend=self.backend,
task_name=self.name,
**kwargs)
def subtask(self, args=None, *starargs, **starkwargs):
"""Return :class:`~celery.signature` object for
this task, wrapping arguments and execution options
for a single task invocation."""
starkwargs.setdefault('app', self.app)
return signature(self, args, *starargs, **starkwargs)
def s(self, *args, **kwargs):
"""``.s(*a, **k) -> .subtask(a, k)``"""
return self.subtask(args, kwargs)
def si(self, *args, **kwargs):
"""``.si(*a, **k) -> .subtask(a, k, immutable=True)``"""
return self.subtask(args, kwargs, immutable=True)
def chunks(self, it, n):
"""Creates a :class:`~celery.canvas.chunks` task for this task."""
from celery import chunks
return chunks(self.s(), it, n, app=self.app)
def map(self, it):
"""Creates a :class:`~celery.canvas.xmap` task from ``it``."""
from celery import xmap
return xmap(self.s(), it, app=self.app)
def starmap(self, it):
"""Creates a :class:`~celery.canvas.xstarmap` task from ``it``."""
from celery import xstarmap
return xstarmap(self.s(), it, app=self.app)
def send_event(self, type_, **fields):
req = self.request
with self.app.events.default_dispatcher(hostname=req.hostname) as d:
return d.send(type_, uuid=req.id, **fields)
def update_state(self, task_id=None, state=None, meta=None):
"""Update task state.
:keyword task_id: Id of the task to update, defaults to the
id of the current task
:keyword state: New state (:class:`str`).
:keyword meta: State metadata (:class:`dict`).
"""
if task_id is None:
task_id = self.request.id
self.backend.store_result(task_id, meta, state)
def on_success(self, retval, task_id, args, kwargs):
"""Success handler.
Run by the worker if the task executes successfully.
:param retval: The return value of the task.
:param task_id: Unique id of the executed task.
:param args: Original arguments for the executed task.
:param kwargs: Original keyword arguments for the executed task.
The return value of this handler is ignored.
"""
pass
def on_retry(self, exc, task_id, args, kwargs, einfo):
"""Retry handler.
This is run by the worker when the task is to be retried.
:param exc: The exception sent to :meth:`retry`.
:param task_id: Unique id of the retried task.
:param args: Original arguments for the retried task.
:param kwargs: Original keyword arguments for the retried task.
:keyword einfo: :class:`~billiard.einfo.ExceptionInfo`
instance, containing the traceback.
The return value of this handler is ignored.
"""
pass
def on_failure(self, exc, task_id, args, kwargs, einfo):
"""Error handler.
This is run by the worker when the task fails.
:param exc: The exception raised by the task.
:param task_id: Unique id of the failed task.
:param args: Original arguments for the task that failed.
:param kwargs: Original keyword arguments for the task
that failed.
:keyword einfo: :class:`~billiard.einfo.ExceptionInfo`
instance, containing the traceback.
The return value of this handler is ignored.
"""
pass
def after_return(self, status, retval, task_id, args, kwargs, einfo):
"""Handler called after the task returns.
:param status: Current task state.
:param retval: Task return value/exception.
:param task_id: Unique id of the task.
:param args: Original arguments for the task.
:param kwargs: Original keyword arguments for the task.
:keyword einfo: :class:`~billiard.einfo.ExceptionInfo`
instance, containing the traceback (if any).
The return value of this handler is ignored.
"""
pass
def send_error_email(self, context, exc, **kwargs):
if self.send_error_emails and \
not getattr(self, 'disable_error_emails', None):
self.ErrorMail(self, **kwargs).send(context, exc)
def add_trail(self, result):
if self.trail:
self.request.children.append(result)
return result
def push_request(self, *args, **kwargs):
self.request_stack.push(Context(*args, **kwargs))
def pop_request(self):
self.request_stack.pop()
def __repr__(self):
"""`repr(task)`"""
return _reprtask(self, R_SELF_TASK if self.__self__ else R_INSTANCE)
def _get_request(self):
"""Get current request object."""
req = self.request_stack.top
if req is None:
# task was not called, but some may still expect a request
# to be there, perhaps that should be deprecated.
if self._default_request is None:
self._default_request = Context()
return self._default_request
return req
request = property(_get_request)
def _get_exec_options(self):
if self._exec_options is None:
self._exec_options = extract_exec_options(self)
return self._exec_options
@property
def backend(self):
backend = self._backend
if backend is None:
return self.app.backend
return backend
@backend.setter
def backend(self, value): # noqa
self._backend = value
@property
def __name__(self):
return self.__class__.__name__
class Task(object):
"""Task base class.
When called tasks apply the :meth:`run` method. This method must
be defined by all tasks (that is unless the :meth:`__call__` method
is overridden).
"""
__trace__ = None
__v2_compat__ = False # set by old base in celery.task.base
MaxRetriesExceededError = MaxRetriesExceededError
OperationalError = OperationalError
#: Execution strategy used, or the qualified name of one.
Strategy = 'celery.worker.strategy:default'
#: This is the instance bound to if the task is a method of a class.
__self__ = None
#: The application instance associated with this task class.
_app = None
#: Name of the task.
name = None
#: Maximum number of retries before giving up. If set to :const:`None`,
#: it will **never** stop retrying.
max_retries = 3
#: Default time in seconds before a retry of the task should be
#: executed. 3 minutes by default.
default_retry_delay = 3 * 60
#: Rate limit for this task type. Examples: :const:`None` (no rate
#: limit), `'100/s'` (hundred tasks a second), `'100/m'` (hundred tasks
#: a minute),`'100/h'` (hundred tasks an hour)
rate_limit = None
#: If enabled the worker won't store task state and return values
#: for this task. Defaults to the :setting:`task_ignore_result`
#: setting.
ignore_result = None
#: If enabled the request will keep track of subtasks started by
#: this task, and this information will be sent with the result
#: (``result.children``).
trail = True
#: If enabled the worker will send monitoring events related to
#: this task (but only if the worker is configured to send
#: task related events).
#: Note that this has no effect on the task-failure event case
#: where a task is not registered (as it will have no task class
#: to check this flag).
send_events = True
#: When enabled errors will be stored even if the task is otherwise
#: configured to ignore results.
store_errors_even_if_ignored = None
#: The name of a serializer that are registered with
#: :mod:`kombu.serialization.registry`. Default is `'pickle'`.
serializer = None
#: Hard time limit.
#: Defaults to the :setting:`task_time_limit` setting.
time_limit = None
#: Soft time limit.
#: Defaults to the :setting:`task_soft_time_limit` setting.
soft_time_limit = None
#: The result store backend used for this task.
backend = None
#: If disabled this task won't be registered automatically.
autoregister = True
#: If enabled the task will report its status as 'started' when the task
#: is executed by a worker. Disabled by default as the normal behavior
#: is to not report that level of granularity. Tasks are either pending,
#: finished, or waiting to be retried.
#:
#: Having a 'started' status can be useful for when there are long
#: running tasks and there's a need to report what task is currently
#: running.
#:
#: The application default can be overridden using the
#: :setting:`task_track_started` setting.
track_started = None
#: When enabled messages for this task will be acknowledged **after**
#: the task has been executed, and not *just before* (the
#: default behavior).
#:
#: Please note that this means the task may be executed twice if the
#: worker crashes mid execution.
#:
#: The application default can be overridden with the
#: :setting:`task_acks_late` setting.
acks_late = None
#: Even if :attr:`acks_late` is enabled, the worker will
#: acknowledge tasks when the worker process executing them abruptly
#: exits or is signaled (e.g., :sig:`KILL`/:sig:`INT`, etc).
#:
#: Setting this to true allows the message to be re-queued instead,
#: so that the task will execute again by the same worker, or another
#: worker.
#:
#: Warning: Enabling this can cause message loops; make sure you know
#: what you're doing.
reject_on_worker_lost = None
#: Tuple of expected exceptions.
#:
#: These are errors that are expected in normal operation
#: and that shouldn't be regarded as a real error by the worker.
#: Currently this means that the state will be updated to an error
#: state, but the worker won't log the event as an error.
throws = ()
#: Default task expiry time.
expires = None
#: Max length of result representation used in logs and events.
resultrepr_maxsize = 1024
#: Task request stack, the current request will be the topmost.
request_stack = None
#: Some may expect a request to exist even if the task hasn't been
#: called. This should probably be deprecated.
_default_request = None
#: Deprecated attribute ``abstract`` here for compatibility.
abstract = True
_exec_options = None
__bound__ = False
from_config = (
('serializer', 'task_serializer'),
('rate_limit', 'task_default_rate_limit'),
('track_started', 'task_track_started'),
('acks_late', 'task_acks_late'),
('reject_on_worker_lost', 'task_reject_on_worker_lost'),
('ignore_result', 'task_ignore_result'),
('store_errors_even_if_ignored', 'task_store_errors_even_if_ignored'),
)
_backend = None # set by backend property.
# - Tasks are lazily bound, so that configuration is not set
# - until the task is actually used
@classmethod
def bind(self, app):
was_bound, self.__bound__ = self.__bound__, True
self._app = app
conf = app.conf
self._exec_options = None # clear option cache
for attr_name, config_name in self.from_config:
if getattr(self, attr_name, None) is None:
setattr(self, attr_name, conf[config_name])
# decorate with annotations from config.
if not was_bound:
self.annotate()
from celery.utils.threads import LocalStack
self.request_stack = LocalStack()
# PeriodicTask uses this to add itself to the PeriodicTask schedule.
self.on_bound(app)
return app
@classmethod
def on_bound(self, app):
"""This method can be defined to do additional actions when the
task class is bound to an app."""
pass
@classmethod
def _get_app(self):
if self._app is None:
self._app = current_app
if not self.__bound__:
# The app property's __set__ method is not called
# if Task.app is set (on the class), so must bind on use.
self.bind(self._app)
return self._app
app = class_property(_get_app, bind)
@classmethod
def annotate(self):
for d in resolve_all_annotations(self.app.annotations, self):
for key, value in items(d):
if key.startswith('@'):
self.add_around(key[1:], value)
else:
setattr(self, key, value)
@classmethod
def add_around(self, attr, around):
orig = getattr(self, attr)
if getattr(orig, '__wrapped__', None):
orig = orig.__wrapped__
meth = around(orig)
meth.__wrapped__ = orig
setattr(self, attr, meth)
def __call__(self, *args, **kwargs):
_task_stack.push(self)
self.push_request(args=args, kwargs=kwargs)
try:
# add self if this is a bound task
if self.__self__ is not None:
return self.run(self.__self__, *args, **kwargs)
return self.run(*args, **kwargs)
finally:
self.pop_request()
_task_stack.pop()
def __reduce__(self):
# - tasks are pickled into the name of the task only, and the reciever
# - simply grabs it from the local registry.
# - in later versions the module of the task is also included,
# - and the receiving side tries to import that module so that
# - it will work even if the task hasn't been registered.
mod = type(self).__module__
mod = mod if mod and mod in sys.modules else None
return (_unpickle_task_v2, (self.name, mod), None)
def run(self, *args, **kwargs):
"""The body of the task executed by workers."""
raise NotImplementedError('Tasks must define the run method.')
def start_strategy(self, app, consumer, **kwargs):
return instantiate(self.Strategy, self, app, consumer, **kwargs)
def delay(self, *args, **kwargs):
"""Star argument version of :meth:`apply_async`.
Does not support the extra options enabled by :meth:`apply_async`.
Arguments:
*args (Any): Positional arguments passed on to the task.
**kwargs (Any): Keyword arguments passed on to the task.
Returns:
celery.result.AsyncResult: Future promise.
"""
return self.apply_async(args, kwargs)
def apply_async(self, args=None, kwargs=None, task_id=None, producer=None,
link=None, link_error=None, shadow=None, **options):
"""Apply tasks asynchronously by sending a message.
Arguments:
args (Tuple): The positional arguments to pass on to the task.
kwargs (Dict): The keyword arguments to pass on to the task.
countdown (float): Number of seconds into the future that the
task should execute. Defaults to immediate execution.
eta (~datetime.datetime): Absolute time and date of when the task
should be executed. May not be specified if `countdown`
is also supplied.
expires (float, ~datetime.datetime): Datetime or
seconds in the future for the task should expire.
The task won't be executed after the expiration time.
shadow (str): Override task name used in logs/monitoring.
Default is retrieved from :meth:`shadow_name`.
connection (kombu.Connection): Re-use existing broker connection
instead of acquiring one from the connection pool.
retry (bool): If enabled sending of the task message will be
retried in the event of connection loss or failure.
Default is taken from the :setting:`task_publish_retry`
setting. Note that you need to handle the
producer/connection manually for this to work.
retry_policy (Mapping): Override the retry policy used.
See the :setting:`task_publish_retry_policy` setting.
queue (str, kombu.Queue): The queue to route the task to.
This must be a key present in :setting:`task_queues`, or
:setting:`task_create_missing_queues` must be
enabled. See :ref:`guide-routing` for more
information.
exchange (str, kombu.Exchange): Named custom exchange to send the
task to. Usually not used in combination with the ``queue``
argument.
routing_key (str): Custom routing key used to route the task to a
worker server. If in combination with a ``queue`` argument
only used to specify custom routing keys to topic exchanges.
priority (int): The task priority, a number between 0 and 9.
Defaults to the :attr:`priority` attribute.
serializer (str): Serialization method to use.
Can be `pickle`, `json`, `yaml`, `msgpack` or any custom
serialization method that's been registered
with :mod:`kombu.serialization.registry`.
Defaults to the :attr:`serializer` attribute.
compression (str): Optional compression method
to use. Can be one of ``zlib``, ``bzip2``,
or any custom compression methods registered with
:func:`kombu.compression.register`.
Defaults to the :setting:`task_compression` setting.
link (~@Signature): A single, or a list of tasks signatures
to apply if the task returns successfully.
link_error (~@Signature): A single, or a list of task signatures
to apply if an error occurs while executing the task.
producer (kombu.Producer): custom producer to use when publishing
the task.
add_to_parent (bool): If set to True (default) and the task
is applied while executing another task, then the result
will be appended to the parent tasks ``request.children``
attribute. Trailing can also be disabled by default using the
:attr:`trail` attribute
publisher (kombu.Producer): Deprecated alias to ``producer``.
headers (Dict): Message headers to be included in the message.
Returns:
~@AsyncResult: Future promise.
Note:
Also supports all keyword arguments supported by
:meth:`kombu.Producer.publish`.
"""
try:
check_arguments = self.__header__
except AttributeError: # pragma: no cover
pass
else:
check_arguments(*(args or ()), **(kwargs or {}))
app = self._get_app()
if app.conf.task_always_eager:
return self.apply(args, kwargs, task_id=task_id or uuid(),
link=link, link_error=link_error, **options)
# add 'self' if this is a "task_method".
if self.__self__ is not None:
args = args if isinstance(args, tuple) else tuple(args or ())
args = (self.__self__,) + args
shadow = shadow or self.shadow_name(args, kwargs, options)
preopts = self._get_exec_options()
options = dict(preopts, **options) if options else preopts
return app.send_task(
self.name, args, kwargs, task_id=task_id, producer=producer,
link=link, link_error=link_error, result_cls=self.AsyncResult,
shadow=shadow, task_type=self,
**options
)
def shadow_name(self, args, kwargs, options):
"""Override for custom task name in worker logs/monitoring.
Example:
.. code-block:: python
from celery.utils.imports import qualname
def shadow_name(task, args, kwargs, options):
return qualname(args[0])
@app.task(shadow_name=shadow_name, serializer='pickle')
def apply_function_async(fun, *args, **kwargs):
return fun(*args, **kwargs)
Arguments:
args (Tuple): Task positional arguments.
kwargs (Dict): Task keyword arguments.
options (Dict): Task execution options.
"""
pass
def signature_from_request(self, request=None, args=None, kwargs=None,
queue=None, **extra_options):
request = self.request if request is None else request
args = request.args if args is None else args
kwargs = request.kwargs if kwargs is None else kwargs
options = request.as_execution_options()
options.update(
{'queue': queue} if queue else (request.delivery_info or {}),
)
return self.signature(
args, kwargs, options, type=self, **extra_options
)
subtask_from_request = signature_from_request # XXX compat
def retry(self, args=None, kwargs=None, exc=None, throw=True,
eta=None, countdown=None, max_retries=None, **options):
"""Retry the task.
Example:
>>> from imaginary_twitter_lib import Twitter
>>> from proj.celery import app
>>> @app.task(bind=True)
... def tweet(self, auth, message):
... twitter = Twitter(oauth=auth)
... try:
... twitter.post_status_update(message)
... except twitter.FailWhale as exc:
... # Retry in 5 minutes.
... raise self.retry(countdown=60 * 5, exc=exc)
Note:
Although the task will never return above as `retry` raises an
exception to notify the worker, we use `raise` in front of the
retry to convey that the rest of the block won't be executed.
Arguments:
args (Tuple): Positional arguments to retry with.
kwargs (Dict): Keyword arguments to retry with.
exc (Exception): Custom exception to report when the max restart
limit has been exceeded (default:
:exc:`~@MaxRetriesExceededError`).
If this argument is set and retry is called while
an exception was raised (``sys.exc_info()`` is set)
it will attempt to re-raise the current exception.
If no exception was raised it will raise the ``exc``
argument provided.
countdown (float): Time in seconds to delay the retry for.
eta (~datetime.dateime): Explicit time and date to run the
retry at.
max_retries (int): If set, overrides the default retry limit for
this execution. Changes to this parameter don't propagate to
subsequent task retry attempts. A value of :const:`None`,
means "use the default", so if you want infinite retries you'd
have to set the :attr:`max_retries` attribute of the task to
:const:`None` first.
time_limit (int): If set, overrides the default time limit.
soft_time_limit (int): If set, overrides the default soft
time limit.
throw (bool): If this is :const:`False`, don't raise the
:exc:`~@Retry` exception, that tells the worker to mark
the task as being retried. Note that this means the task
will be marked as failed if the task raises an exception,
or successful if it returns after the retry call.
**options (Any): Extra options to pass on to :meth:`apply_async`.
Raises:
celery.exceptions.Retry:
To tell the worker that the task has been re-sent for retry.
This always happens, unless the `throw` keyword argument
has been explicitly set to :const:`False`, and is considered
normal operation.
"""
request = self.request
retries = request.retries + 1
max_retries = self.max_retries if max_retries is None else max_retries
# Not in worker or emulated by (apply/always_eager),
# so just raise the original exception.
if request.called_directly:
# raises orig stack if PyErr_Occurred,
# and augments with exc' if that argument is defined.
raise_with_context(exc or Retry('Task can be retried', None))
if not eta and countdown is None:
countdown = self.default_retry_delay
is_eager = request.is_eager
S = self.signature_from_request(
request, args, kwargs,
countdown=countdown, eta=eta, retries=retries,
**options
)
if max_retries is not None and retries > max_retries:
if exc:
# On Py3: will augment any current exception with
# the exc' argument provided (raise exc from orig)
raise_with_context(exc)
raise self.MaxRetriesExceededError(
"Can't retry {0}[{1}] args:{2} kwargs:{3}".format(
self.name, request.id, S.args, S.kwargs))
ret = Retry(exc=exc, when=eta or countdown)
if is_eager:
# if task was executed eagerly using apply(),
# then the retry must also be executed eagerly.
S.apply().get()
if throw:
raise ret
return ret
try:
S.apply_async()
except Exception as exc:
raise Reject(exc, requeue=False)
if throw:
raise ret
return ret
def apply(self, args=None, kwargs=None,
link=None, link_error=None,
task_id=None, retries=None, throw=None,
logfile=None, loglevel=None, headers=None, **options):
"""Execute this task locally, by blocking until the task returns.
Arguments:
args (Tuple): positional arguments passed on to the task.
kwargs (Dict): keyword arguments passed on to the task.
throw (bool): Re-raise task exceptions.
Defaults to the :setting:`task_eager_propagates` setting.
Returns:
celery.result.EagerResult: pre-evaluated result.
"""
# trace imports Task, so need to import inline.
from celery.app.trace import build_tracer
app = self._get_app()
args = args or ()
# add 'self' if this is a bound method.
if self.__self__ is not None:
args = (self.__self__,) + tuple(args)
kwargs = kwargs or {}
task_id = task_id or uuid()
retries = retries or 0
if throw is None:
throw = app.conf.task_eager_propagates
# Make sure we get the task instance, not class.
task = app._tasks[self.name]
request = {
'id': task_id,
'retries': retries,
'is_eager': True,
'logfile': logfile,
'loglevel': loglevel or 0,
'callbacks': maybe_list(link),
'errbacks': maybe_list(link_error),
'headers': headers,
'delivery_info': {'is_eager': True},
}
tb = None
tracer = build_tracer(
task.name, task, eager=True,
propagate=throw, app=self._get_app(),
)
ret = tracer(task_id, args, kwargs, request)
retval = ret.retval
if isinstance(retval, ExceptionInfo):
retval, tb = retval.exception, retval.traceback
state = states.SUCCESS if ret.info is None else ret.info.state
return EagerResult(task_id, retval, state, traceback=tb)
def AsyncResult(self, task_id, **kwargs):
"""Get AsyncResult instance for this kind of task.
Arguments:
task_id (str): Task id to get result for.
"""
return self._get_app().AsyncResult(task_id, backend=self.backend,
task_name=self.name, **kwargs)
def signature(self, args=None, *starargs, **starkwargs):
"""Return :class:`~celery.signature` object for
this task, wrapping arguments and execution options
for a single task invocation."""
starkwargs.setdefault('app', self.app)
return signature(self, args, *starargs, **starkwargs)
subtask = signature
def s(self, *args, **kwargs):
"""``.s(*a, **k) -> .signature(a, k)``"""
return self.signature(args, kwargs)
def si(self, *args, **kwargs):
"""``.si(*a, **k) -> .signature(a, k, immutable=True)``"""
return self.signature(args, kwargs, immutable=True)
def chunks(self, it, n):
"""Creates a :class:`~celery.canvas.chunks` task for this task."""
from celery import chunks
return chunks(self.s(), it, n, app=self.app)
def map(self, it):
"""Creates a :class:`~celery.canvas.xmap` task from ``it``."""
from celery import xmap
return xmap(self.s(), it, app=self.app)
def starmap(self, it):
"""Creates a :class:`~celery.canvas.xstarmap` task from ``it``."""
from celery import xstarmap
return xstarmap(self.s(), it, app=self.app)
def send_event(self, type_, **fields):
req = self.request
with self.app.events.default_dispatcher(hostname=req.hostname) as d:
return d.send(type_, uuid=req.id, **fields)
def replace(self, sig):
"""Replace the current task, with a new task inheriting the
same task id.
.. versionadded:: 4.0
Arguments:
sig (~@Signature): signature to replace with.
Raises:
~@Ignore: This is always raised, so the best practice
is to always use ``raise self.replace(...)`` to convey
to the reader that the task won't continue after being replaced.
"""
chord = self.request.chord
if 'chord' in sig.options:
if chord:
chord = sig.options['chord'] | chord
else:
chord = sig.options['chord']
if isinstance(sig, group):
sig |= self.app.tasks['celery.accumulate'].s(index=0).set(
chord=chord,
link=self.request.callbacks,
link_error=self.request.errbacks,
)
chord = None
if self.request.chain:
for t in self.request.chain:
sig |= signature(t)
sig.freeze(self.request.id,
group_id=self.request.group,
chord=chord,
root_id=self.request.root_id)
sig.delay()
raise Ignore('Replaced by new task')
def add_to_chord(self, sig, lazy=False):
"""Add signature to the chord the current task is a member of.
.. versionadded:: 4.0
Currently only supported by the Redis result backend.
Arguments:
sig (~@Signature): Signature to extend chord with.
lazy (bool): If enabled the new task won't actually be called,
and ``sig.delay()`` must be called manually.
"""
if not self.request.chord:
raise ValueError('Current task is not member of any chord')
result = sig.freeze(group_id=self.request.group,
chord=self.request.chord,
root_id=self.request.root_id)
self.backend.add_to_chord(self.request.group, result)
return sig.delay() if not lazy else sig
def update_state(self, task_id=None, state=None, meta=None):
"""Update task state.
Arguments:
task_id (str): Id of the task to update.
Defaults to the id of the current task.
state (str): New state.
meta (Dict): State meta-data.
"""
if task_id is None:
task_id = self.request.id
self.backend.store_result(task_id, meta, state)
def on_success(self, retval, task_id, args, kwargs):
"""Success handler.
Run by the worker if the task executes successfully.
Arguments:
retval (Any): The return value of the task.
task_id (str): Unique id of the executed task.
args (Tuple): Original arguments for the executed task.
kwargs (Dict): Original keyword arguments for the executed task.
Returns:
None: The return value of this handler is ignored.
"""
pass
def on_retry(self, exc, task_id, args, kwargs, einfo):
"""Retry handler.
This is run by the worker when the task is to be retried.
Arguments:
exc (Exception): The exception sent to :meth:`retry`.
task_id (str): Unique id of the retried task.
args (Tuple): Original arguments for the retried task.
kwargs (Dict): Original keyword arguments for the retried task.
einfo (~billiard.einfo.ExceptionInfo): Exception information.
Returns:
None: The return value of this handler is ignored.
"""
pass
def on_failure(self, exc, task_id, args, kwargs, einfo):
"""Error handler.
This is run by the worker when the task fails.
Arguments:
exc (Exception): The exception raised by the task.
task_id (str): Unique id of the failed task.
args (Tuple): Original arguments for the task that failed.
kwargs (Dict): Original keyword arguments for the task that failed.
einfo (~billiard.einfo.ExceptionInfo): Exception information.
Returns:
None: The return value of this handler is ignored.
"""
pass
def after_return(self, status, retval, task_id, args, kwargs, einfo):
"""Handler called after the task returns.
Arguments:
status (str): Current task state.
retval (Any): Task return value/exception.
task_id (str): Unique id of the task.
args (Tuple): Original arguments for the task.
kwargs (Dict): Original keyword arguments for the task.
einfo (~billiard.einfo.ExceptionInfo): Exception information.
Returns:
None: The return value of this handler is ignored.
"""
pass
def add_trail(self, result):
if self.trail:
self.request.children.append(result)
return result
def push_request(self, *args, **kwargs):
self.request_stack.push(Context(*args, **kwargs))
def pop_request(self):
self.request_stack.pop()
def __repr__(self):
"""`repr(task)`"""
return _reprtask(self, R_SELF_TASK if self.__self__ else R_INSTANCE)
def _get_request(self):
"""Get current request object."""
req = self.request_stack.top
if req is None:
# task was not called, but some may still expect a request
# to be there, perhaps that should be deprecated.
if self._default_request is None:
self._default_request = Context()
return self._default_request
return req
request = property(_get_request)
def _get_exec_options(self):
if self._exec_options is None:
self._exec_options = extract_exec_options(self)
return self._exec_options
@property
def backend(self):
backend = self._backend
if backend is None:
return self.app.backend
return backend
@backend.setter
def backend(self, value): # noqa
self._backend = value
@property
def __name__(self):
return self.__class__.__name__
class JobtasticTask(Task):
"""
A base ``Celery.Task`` class that provides some common niceties for running
tasks that return some kind of result for which you need to wait.
To create a task that uses these helpers, use ``JobtasticTask`` as a
subclass and define a ``calculate_result`` method which returns a
dictionary to be turned in to JSON. You will also need to define the
following class variables:
* ``significant_kwargs`` The kwarg values that will be converted to strings
and hashed to determine if two versions of the same task are equivalent.
This is a list of 2-tuples with the first item being the kwarg string and
the second being a callable that converts the value to a hashable string.
If no second item is given, it's assumed that calling ``str()`` on the
value works just fine.
* ``herd_avoidance_timeout`` Number of seconds to hold a lock on this task
for other equivalent runs. Generally, this should be set to the longest
estimated amount of time the task could consume.
The following class members are optional:
* ``cache_prefix`` A unique string representing this task. Eg.
``foo.bar.tasks.BazzTask``
* ``cache_duration`` The number of seconds for which the result of this
task should be cached, meaning subsequent equivalent runs will skip
computation. The default is to do no result caching.
* ``memleak_threshold`` When a single run of a Task increase the resident
process memory usage by more than this number of MegaBytes, a warning is
logged to the logger. This is useful for finding tasks that are behaving
badly under certain conditions. By default, no logging is performed.
Set this value to 0 to log all RAM changes and -1 to disable logging.
Provided are helpers for:
1. Handling failures to connect the task broker by either directly
running the task (`delay_or_run`) or by returning a task that
contains the connection error (`delay_or_fail`). This minimizes
the user-facing impact of a dead task broker.
2. Defeating any thundering herd issues by ensuring only one of a task with
specific arguments can be running at a time by directing subsequent calls
to latch on to the appropriate result.
3. Caching the final result for a designated time period so that subsequent
equivalent calls return quickly.
4. Returning the results as JSON, so that they can be processed easily by
client-side javascript.
5. Returning time-based, continually updating progress estimates to
front-end code so that users know what to expect.
"""
abstract = True
#: The shared cache used for locking and thundering herd protection
_cache = None
@classmethod
def async_or_eager(self, **options):
"""
Attempt to call self.apply_async, or if that fails because of a problem
with the broker, run the task eagerly and return an EagerResult.
"""
args = options.pop("args", None)
kwargs = options.pop("kwargs", None)
possible_broker_errors = self._get_possible_broker_errors_tuple()
try:
return self.apply_async(args, kwargs, **options)
except possible_broker_errors:
return self.apply(args, kwargs, **options)
@classmethod
def async_or_fail(self, **options):
"""
Attempt to call self.apply_async, but if that fails with an exception,
we fake the task completion using the exception as the result. This
allows us to seamlessly handle errors on task creation the same way we
handle errors when a task runs, simplifying the user interface.
"""
args = options.pop("args", None)
kwargs = options.pop("kwargs", None)
possible_broker_errors = self._get_possible_broker_errors_tuple()
try:
return self.apply_async(args, kwargs, **options)
except possible_broker_errors as e:
return self.simulate_async_error(e)
@classmethod
def delay_or_eager(self, *args, **kwargs):
"""
Wrap async_or_eager with a convenience signiture like delay
"""
return self.async_or_eager(args=args, kwargs=kwargs)
@classmethod
def delay_or_run(self, *args, **kwargs):
"""
Attempt to call self.delay, or if that fails, call self.run.
Returns a tuple, (result, required_fallback). ``result`` is the result
of calling delay or run. ``required_fallback`` is True if the broker
failed we had to resort to `self.run`.
"""
warnings.warn(
"delay_or_run is deprecated. Please use delay_or_eager",
DeprecationWarning,
)
possible_broker_errors = self._get_possible_broker_errors_tuple()
try:
result = self.apply_async(args=args, kwargs=kwargs)
required_fallback = False
except possible_broker_errors:
result = self().run(*args, **kwargs)
required_fallback = True
return result, required_fallback
@classmethod
def delay_or_fail(self, *args, **kwargs):
"""
Wrap async_or_fail with a convenience signiture like delay
"""
return self.async_or_fail(args=args, kwargs=kwargs)
@classmethod
def _get_possible_broker_errors_tuple(self):
if hasattr(self.app, 'connection'):
dummy_conn = self.app.connection()
else:
# Celery 2.5 uses `broker_connection` instead
dummy_conn = self.app.broker_connection()
return dummy_conn.connection_errors + dummy_conn.channel_errors
@classmethod
def simulate_async_error(self, exception):
"""
Take this exception and store it as an error in the result backend.
This unifies the handling of broker-connection errors with any other
type of error that might occur when running the task. So the same
error-handling that might retry a task or display a useful message to
the user can also handle this error.
"""
task_id = gen_unique_id()
async_result = self.AsyncResult(task_id)
einfo = ExceptionInfo(sys.exc_info())
async_result.backend.mark_as_failure(
task_id,
exception,
traceback=einfo.traceback,
)
return async_result
@classmethod
def apply_async(self, args, kwargs, **options):
"""
Put this task on the Celery queue as a singleton. Only one of this type
of task with its distinguishing args/kwargs will be allowed on the
queue at a time. Subsequent duplicate tasks called while this task is
still running will just latch on to the results of the running task by
synchronizing the task uuid. Additionally, identical task calls will
return those results for the next ``cache_duration`` seconds.
"""
self._validate_required_class_vars()
cache_key = self._get_cache_key(**kwargs)
# Check for an already-computed and cached result
task_id = self.cache.get(cache_key) # Check for the cached result
if task_id:
# We've already built this result, just latch on to the task that
# did the work
logging.info('Found existing cached and completed task: %s',
task_id)
return self.AsyncResult(task_id)
# Check for an in-progress equivalent task to avoid duplicating work
task_id = self.cache.get('herd:%s' % cache_key)
if task_id:
logging.info('Found existing in-progress task: %s', task_id)
return self.AsyncResult(task_id)
# It's not cached and it's not already running. Use an atomic lock to
# start the task, ensuring there isn't a race condition that could
# result in multiple identical tasks being fired at once.
with self.cache.lock('lock:%s' % cache_key):
task_meta = super(JobtasticTask,
self).apply_async(args, kwargs, **options)
logging.info('Current status: %s', task_meta.status)
if task_meta.status in (PROGRESS, PENDING):
self.cache.set('herd:%s' % cache_key,
task_meta.task_id,
timeout=self.herd_avoidance_timeout)
logging.info('Setting herd-avoidance cache for task: %s',
cache_key)
return task_meta
def calc_progress(self, completed_count, total_count):
"""
Calculate the percentage progress and estimated remaining time based on
the current number of items completed of the total.
Returns a tuple of ``(percentage_complete, seconds_remaining)``.
"""
self.logger.debug(
"calc_progress(%s, %s)",
completed_count,
total_count,
)
current_time = time.time()
time_spent = current_time - self.start_time
self.logger.debug("Progress time spent: %s", time_spent)
if total_count == 0:
return 100, 1
completion_fraction = completed_count / total_count
if completion_fraction == 0:
completion_fraction = 1
total_time = 0
total_time = time_spent / completion_fraction
time_remaining = total_time - time_spent
completion_display = completion_fraction * 100
if completion_display == 100:
return 100, 1 # 1 second to finish up
return completion_display, time_remaining
def update_progress(
self,
completed_count,
total_count,
update_frequency=1,
):
"""
Update the task backend with both an estimated percentage complete and
number of seconds remaining until completion.
``completed_count`` Number of task "units" that have been completed out
of ``total_count`` total "units."
``update_frequency`` Only actually store the updated progress in the
background at most every ``N`` ``completed_count``.
"""
if completed_count - self._last_update_count < update_frequency:
# We've updated the progress too recently. Don't stress out the
# result backend
return
# Store progress for display
progress_percent, time_remaining = self.calc_progress(
completed_count, total_count)
self.logger.debug("Updating progress: %s percent, %s remaining",
progress_percent, time_remaining)
if self.request.id:
self._last_update_count = completed_count
self.update_state(
None, PROGRESS, {
"progress_percent": progress_percent,
"time_remaining": time_remaining,
})
def run(self, *args, **kwargs):
if get_task_logger:
self.logger = get_task_logger(self.__class__.__name__)
else:
# Celery 2.X fallback
self.logger = self.get_logger(**kwargs)
self.logger.info("Starting %s", self.__class__.__name__)
self.cache_key = self._get_cache_key(**kwargs)
# Record start time to give estimated time remaining estimates
self.start_time = time.time()
# Keep track of progress updates for update_frequency tracking
self._last_update_count = 0
# Report to the backend that work has been started.
if self.request.id:
self.update_state(None, PROGRESS, {
"progress_percent": 0,
"time_remaining": -1,
})
memleak_threshold = int(getattr(self, 'memleak_threshold', -1))
if memleak_threshold >= 0:
begining_memory_usage = self._get_memory_usage()
self.logger.info("Calculating result")
try:
task_result = self.calculate_result(*args, **kwargs)
except Exception:
# Don't want other tasks waiting for this task to finish, since it
# won't
self._break_thundering_herd_cache()
raise # We can use normal celery exception handling for this
if hasattr(self, 'cache_duration'):
cache_duration = self.cache_duration
else:
cache_duration = -1 # By default, don't cache
if cache_duration >= 0:
# If we're configured to cache this result, do so.
self.cache.set(self.cache_key, self.request.id, cache_duration)
# Now that the task is finished, we can stop all of the thundering herd
# avoidance
self._break_thundering_herd_cache()
if memleak_threshold >= 0:
self._warn_if_leaking_memory(
begining_memory_usage,
self._get_memory_usage(),
memleak_threshold,
task_kwargs=kwargs,
)
return task_result
def calculate_result(self, *args, **kwargs):
raise NotImplementedError(("Tasks using JobtasticTask must implement "
"their own calculate_result"))
@classmethod
def _validate_required_class_vars(self):
"""
Ensure that this subclass has defined all of the required class
variables.
"""
required_members = (
'significant_kwargs',
'herd_avoidance_timeout',
)
for required_member in required_members:
if not hasattr(self, required_member):
raise Exception("JobtasticTask's must define a %s" %
required_member)
def on_success(self, retval, task_id, args, kwargs):
"""
Store results in the backend even if we're always eager. This ensures
the `delay_or_run` calls always at least have results.
"""
if self.request.is_eager:
# Store the result because celery wouldn't otherwise
self.update_state(task_id, SUCCESS, retval)
def _break_thundering_herd_cache(self):
self.cache.delete('herd:%s' % self.cache_key)
@classmethod
def _get_cache(self):
"""
Return the cache to use for thundering herd protection, etc.
"""
if not self._cache:
self._cache = get_cache(self.app)
return self._cache
@classmethod
def _set_cache(self, cache):
"""
Set the Jobtastic Cache for the Task
The cache must support get/set (with timeout)/delete/lock (as a context
manager).
"""
self._cache = cache
cache = class_property(_get_cache, _set_cache)
@classmethod
def _get_cache_key(self, **kwargs):
"""
Take this task's configured ``significant_kwargs`` and build a hash
that all equivalent task calls will match.
Takes in kwargs and returns a string.
To change the way the cache key is generated or do more in-depth
processing, override this method.
"""
m = md5()
for significant_kwarg in self.significant_kwargs:
key, to_str = significant_kwarg
try:
m.update(to_str(kwargs[key]))
except TypeError:
# Python 3.x strings aren't accepted by hash.update().
# String should be byte-encoded first.
m.update(to_str(kwargs[key]).encode('utf-8'))
if hasattr(self, 'cache_prefix'):
cache_prefix = self.cache_prefix
else:
cache_prefix = '%s.%s' % (self.__module__, self.__name__)
return '%s:%s' % (cache_prefix, m.hexdigest())
def _get_memory_usage(self):
current_process = psutil.Process(os.getpid())
usage = current_process.memory_info()
return usage.rss
def _warn_if_leaking_memory(
self,
begining_usage,
ending_usage,
threshold,
task_kwargs,
):
growth = ending_usage - begining_usage
threshold_in_bytes = threshold * 1000000
if growth > threshold_in_bytes:
self.warn_of_memory_leak(
growth,
begining_usage,
ending_usage,
task_kwargs,
)
def warn_of_memory_leak(
self,
growth,
begining_usage,
ending_usage,
task_kwargs,
):
self.logger.warning(
"Jobtastic:memleak memleak_detected. memory_increase=%05d unit=MB",
growth / 1000000,
)
self.logger.info(
"Jobtastic:memleak memory_usage_start=%05d unit=MB",
begining_usage / 1000000,
)
self.logger.info(
"Jobtastic:memleak memory_usage_end=%05d unit=MB",
ending_usage / 1000000,
)
self.logger.info(
"Jobtastic:memleak task_kwargs=%s",
repr(task_kwargs),
)
