This document describes an older version of Celery (2.1). For the latest stable version please go here.

Tasks¶

This guide gives an overview of how tasks are defined. For a complete listing of task attributes and methods, please see the API reference.

Basics ¶

A task is a class that encapsulates a function and its execution options. Given a function create_user, that takes two arguments: username and password, you can create a task like this:

from celery.task import Task
from django.contrib.auth import User

class CreateUserTask(Task):
    def run(self, username, password):
        User.objects.create(username=username, password=password)

For convenience there is a shortcut decorator that turns any function into a task:

from celery.decorators import task
from django.contrib.auth import User

@task
def create_user(username, password):
    User.objects.create(username=username, password=password)

The task decorator takes the same execution options as the Task class does:

@task(serializer="json")
def create_user(username, password):
    User.objects.create(username=username, password=password)

Default keyword arguments ¶

Celery supports a set of default arguments that can be forwarded to any task. Tasks can choose not to take these, or list the ones they want. The worker will do the right thing.

The current default keyword arguments are:

task_id:	The unique id of the executing task.
task_name:	Name of the currently executing task.
task_retries:	How many times the current task has been retried. An integer starting at `0`.
is_eager:	Set to `True` if the task is executed locally in the client, and not by a worker.
logfile:	The log file, can be passed on to `get_logger()` to gain access to the workers log file. See Logging.
loglevel:	The current log level used.
delivery_info:	Additional message delivery information. This is a mapping containing the exchange and routing key used to deliver this task. It’s used by e.g. `retry()` to resend the task to the same destination queue. NOTE As some messaging backends doesn’t have advanced routing capabilities, you can’t trust the availability of keys in this mapping.

Logging ¶

You can use the workers logger to add diagnostic output to the worker log:

class AddTask(Task):

    def run(self, x, y, **kwargs):
        logger = self.get_logger(**kwargs)
        logger.info("Adding %s + %s" % (x, y))
        return x + y

or using the decorator syntax:

@task()
def add(x, y, **kwargs):
    logger = add.get_logger(**kwargs)
    logger.info("Adding %s + %s" % (x, y))
    return x + y

There are several logging levels available, and the workers loglevel setting decides whether or not they will be written to the log file.

Of course, you can also simply use print as anything written to standard out/-err will be written to the log file as well.

Retrying a task if something fails ¶

Simply use retry() to re-send the task. It will do the right thing, and respect the max_retries attribute:

@task()
def send_twitter_status(oauth, tweet, **kwargs):
    try:
        twitter = Twitter(oauth)
        twitter.update_status(tweet)
    except (Twitter.FailWhaleError, Twitter.LoginError), exc:
        send_twitter_status.retry(args=[oauth, tweet], kwargs=kwargs, exc=exc)

Here we used the exc argument to pass the current exception to retry(). At each step of the retry this exception is available as the tombstone (result) of the task. When max_retries has been exceeded this is the exception raised. However, if an exc argument is not provided the RetryTaskError exception is raised instead.

Important note: The task has to take the magic keyword arguments in order for max retries to work properly, this is because it keeps track of the current number of retries using the task_retries keyword argument passed on to the task. In addition, it also uses the task_id keyword argument to use the same task id, and delivery_info to route the retried task to the same destination.

Using a custom retry delay ¶

When a task is to be retried, it will wait for a given amount of time before doing so. The default delay is in the default_retry_delay attribute on the task. By default this is set to 3 minutes. Note that the unit for setting the delay is in seconds (int or float).

You can also provide the countdown argument to retry() to override this default.

class MyTask(Task):
    default_retry_delay = 30 * 60 # retry in 30 minutes

    def run(self, x, y, **kwargs):
        try:
            ...
        except Exception, exc:
            self.retry([x, y], kwargs, exc=exc,
                       countdown=60) # override the default and
                                     # - retry in 1 minute

Task options ¶

General ¶

Task.name¶

The name the task is registered as.

You can set this name manually, or just use the default which is automatically generated using the module and class name. See Task names.

Task.abstract¶: Abstract classes are not registered, but are used as the base class for new task types.

Task.max_retries¶: The maximum number of attempted retries before giving up. If this exceeds the MaxRetriesExceeded an exception will be raised. NOTE: You have to retry() manually, it’s not something that happens automatically.

Task.default_retry_delay¶: Default time in seconds before a retry of the task should be executed. Can be either int or float. Default is a 3 minute delay.

Task.rate_limit¶

Set the rate limit for this task type, i.e. how many times in a given period of time is the task allowed to run.

If this is None no rate limit is in effect. If it is an integer, it is interpreted as “tasks per second”.

The rate limits can be specified in seconds, minutes or hours by appending "/s", "/m" or "/h" to the value. Example: "100/m" (hundred tasks a minute). Default is the CELERY_DEFAULT_RATE_LIMIT setting, which if not specified means rate limiting for tasks is disabled by default.

Task.ignore_result¶: Don’t store task state. Note that this means you can’t use AsyncResult to check if the task is ready, or get its return value.

Task.store_errors_even_if_ignored¶: If True, errors will be stored even if the task is configured to ignore results.

Task.send_error_emails¶: Send an e-mail whenever a task of this type fails. Defaults to the CELERY_SEND_TASK_ERROR_EMAILS setting. See Error E-Mails for more information.

Task.error_whitelist¶: If the sending of error e-mails is enabled for this task, then this is a white list of exceptions to actually send e-mails about.

Task.serializer¶

A string identifying the default serialization method to use. Defaults to the CELERY_TASK_SERIALIZER setting. Can be pickle json, yaml, or any custom serialization methods that have been registered with carrot.serialization.registry.

Please see Serializers for more information.

Task.backend¶: The result store backend to use for this task. Defaults to the CELERY_RESULT_BACKEND setting.

Task.acks_late¶

If set to True messages for this task will be acknowledged after the task has been executed, not just before, which is the default behavior.

Note that this means the task may be executed twice if the worker crashes in the middle of execution, which may be acceptable for some applications.

The global default can be overridden by the CELERY_ACKS_LATE setting.

Task.track_started¶

If True the task will report its status as “started” when the task is executed by a worker. The default value is False 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 host name and process id of the worker executing the task will be available in the state metadata (e.g. result.info[“pid”])

The global default can be overridden by the CELERY_TRACK_STARTED setting.

See also

The API reference for Task.

Message and routing options ¶

Task.queue¶: Use the routing settings from a queue defined in CELERY_QUEUES. If defined the exchange and routing_key options will be ignored.

Task.exchange¶: Override the global default exchange for this task.

Task.routing_key¶: Override the global default routing_key for this task.

Task.mandatory¶

If set, the task message has mandatory routing. By default the task is silently dropped by the broker if it can’t be routed to a queue. However – If the task is mandatory, an exception will be raised instead.

Not supported by amqplib.

Task.immediate¶

Request immediate delivery. If the task cannot be routed to a task worker immediately, an exception will be raised. This is instead of the default behavior, where the broker will accept and queue the task, but with no guarantee that the task will ever be executed.

Not supported by amqplib.

Task.priority¶

The message priority. A number from 0 to 9, where 0 is the highest priority.

Not supported by RabbitMQ.

See also

Routing options for more information about message options, and Routing Tasks.

Task names ¶

The task type is identified by the task name.

If not provided a name will be automatically generated using the module and class name.

For example:

>>> @task(name="sum-of-two-numbers")
>>> def add(x, y):
...     return x + y

>>> add.name
'sum-of-two-numbers'

The best practice is to use the module name as a prefix to classify the tasks using namespaces. This way the name won’t collide with the name from another module:

>>> @task(name="tasks.add")
>>> def add(x, y):
...     return x + y

>>> add.name
'tasks.add'

Which is exactly the name that is automatically generated for this task if the module name is “tasks.py”:

>>> @task()
>>> def add(x, y):
...     return x + y

>>> add.name
'tasks.add'

Automatic naming and relative imports ¶

Relative imports and automatic name generation does not go well together, so if you’re using relative imports you should set the name explicitly.

For example if the client imports the module “myapp.tasks” as ”.tasks”, and the worker imports the module as “myapp.tasks”, the generated names won’t match and an NotRegistered error will be raised by the worker.

This is also the case if using Django and using project.myapp:

INSTALLED_APPS = ("project.myapp", )

The worker will have the tasks registered as “project.myapp.tasks.*”, while this is what happens in the client if the module is imported as “myapp.tasks”:

>>> from myapp.tasks import add
>>> add.name
'myapp.tasks.add'

For this reason you should never use “project.app”, but rather add the project directory to the Python path:

import os
import sys
sys.path.append(os.getcwd())

INSTALLED_APPS = ("myapp", )

This makes more sense from the reusable app perspective anyway.

Decorating tasks ¶

Using decorators with tasks requires extra steps because of the magic keyword arguments.

If you have the following task and decorator:

from celery.utils.functional import wraps

def decorator(task):

    @wraps(task)
    def _decorated(*args, **kwargs):
        print("inside decorator")
        return task(*args, **kwargs)


@decorator
@task
def add(x, y):
    return x + y

Then the worker will see that the task is accepting keyword arguments, while it really doesn’t, resulting in an error.

The workaround is to either have your task accept arbitrary keyword arguments:

@decorator
@task
def add(x, y, **kwargs):
    return x + y

or patch the decorator to preserve the original signature:

from inspect import getargspec
from celery.utils.functional import wraps

def decorator(task):

    @wraps(task)
    def _decorated(*args, **kwargs):
        print("in decorator")
        return task(*args, **kwargs)
    _decorated.argspec = inspect.getargspec(task)

Also note the use of wraps() here, this is necessary to keep the original function name and docstring.

Note

The magic keyword arguments will be deprecated in the future, replaced by the task.request attribute in 2.2, and the keyword arguments will be removed in 3.0.

Task States ¶

During its lifetime a task will transition through several possible states, and each state may have arbitrary metadata attached to it. When a task moves into a new state the previous state is forgotten about, but some transitions can be deducted, (e.g. a task now in the FAILED state, is implied to have been in the STARTED state at some point).

There are also sets of states, like the set of failure states, and the set of ready states.

The client uses the membership of these sets to decide whether the exception should be re-raised (PROPAGATE_STATES), or whether the result can be cached (it can if the task is ready).

You can also define custom-states.

Built-in States ¶

PENDING ¶

Task is waiting for execution or unknown. Any task id that is not know is implied to be in the pending state.

STARTED ¶

Task has been started. Not reported by default, to enable please see :attr`Task.track_started`.

metadata:	`pid` and `hostname` of the worker process executing the task.

SUCCESS ¶

Task has been successfully executed.

metadata:	`result` contains the return value of the task.
propagates:	Yes
ready:	Yes

FAILURE ¶

Task execution resulted in failure.

metadata:	result contains the exception occurred, and traceback contains the backtrace of the stack at the point when the exception was raised.
propagates:	Yes

RETRY ¶

Task is being retried.

metadata:	`result` contains the exception that caused the retry, and `traceback` contains the backtrace of the stack at the point when the exceptions was raised.
propagates:	No

REVOKED ¶

Task has been revoked.

propagates:	Yes

Custom states ¶

You can easily define your own states, all you need is a unique name. The name of the state is usually an uppercase string. As an example you could have a look at abortable tasks which defines its own custom ABORTED state.

Use Task.update_state to update a tasks state:

@task
def upload_files(filenames, **kwargs):

    for i, file in enumerate(filenames):
        upload_files.update_state(kwargs["task_id"], "PROGRESS",
            {"current": i, "total": len(filenames)})

Here we created the state "PROGRESS", which tells any application aware of this state that the task is currently in progress, and also where it is in the process by having current and total counts as part of the state metadata. This can then be used to create e.g. progress bars.

How it works ¶

Here comes the technical details, this part isn’t something you need to know, but you may be interested.

All defined tasks are listed in a registry. The registry contains a list of task names and their task classes. You can investigate this registry yourself:

>>> from celery import registry
>>> from celery import task
>>> registry.tasks
{'celery.delete_expired_task_meta':
    <PeriodicTask: celery.delete_expired_task_meta (periodic)>,
 'celery.task.http.HttpDispatchTask':
    <Task: celery.task.http.HttpDispatchTask (regular)>,
 'celery.execute_remote':
    <Task: celery.execute_remote (regular)>,
 'celery.map_async':
    <Task: celery.map_async (regular)>,
 'celery.ping':
    <Task: celery.ping (regular)>}

This is the list of tasks built-in to celery. Note that we had to import celery.task first for these to show up. This is because the tasks will only be registered when the module they are defined in is imported.

The default loader imports any modules listed in the CELERY_IMPORTS setting.

The entity responsible for registering your task in the registry is a meta class, TaskType. This is the default meta class for Task.

If you want to register your task manually you can set mark the task as abstract:

class MyTask(Task):
    abstract = True

This way the task won’t be registered, but any task inheriting from it will be.

When tasks are sent, we don’t send any actual function code, just the name of the task to execute. When the worker then receives the message it can look up the name in its task registry to find the execution code.

This means that your workers should always be updated with the same software as the client. This is a drawback, but the alternative is a technical challenge that has yet to be solved.

Tips and Best Practices ¶

Ignore results you don’t want ¶

If you don’t care about the results of a task, be sure to set the ignore_result option, as storing results wastes time and resources.

@task(ignore_result=True)
def mytask(...)
    something()

Results can even be disabled globally using the CELERY_IGNORE_RESULT setting.

Disable rate limits if they’re not used ¶

Disabling rate limits altogether is recommended if you don’t have any tasks using them. This is because the rate limit subsystem introduces quite a lot of complexity.

Set the CELERY_DISABLE_RATE_LIMITS setting to globally disable rate limits:

CELERY_DISABLE_RATE_LIMITS = True

Avoid launching synchronous subtasks ¶

Having a task wait for the result of another task is really inefficient, and may even cause a deadlock if the worker pool is exhausted.

Make your design asynchronous instead, for example by using callbacks.

Bad:

@task()
def update_page_info(url):
    page = fetch_page.delay(url).get()
    info = parse_page.delay(url, page).get()
    store_page_info.delay(url, info)

@task()
def fetch_page(url):
    return myhttplib.get(url)

@task()
def parse_page(url, page):
    return myparser.parse_document(page)

@task()
def store_page_info(url, info):
    return PageInfo.objects.create(url, info)

Good:

@task(ignore_result=True)
def update_page_info(url):
    # fetch_page -> parse_page -> store_page
    fetch_page.delay(url, callback=subtask(parse_page,
                                callback=subtask(store_page_info)))

@task(ignore_result=True)
def fetch_page(url, callback=None):
    page = myhttplib.get(url)
    if callback:
        # The callback may have been serialized with JSON,
        # so best practice is to convert the subtask dict back
        # into a subtask object.
        subtask(callback).delay(url, page)

@task(ignore_result=True)
def parse_page(url, page, callback=None):
    info = myparser.parse_document(page)
    if callback:
        subtask(callback).delay(url, info)

@task(ignore_result=True)
def store_page_info(url, info):
    PageInfo.objects.create(url, info)

We use subtask here to safely pass around the callback task. subtask is a subclass of dict used to wrap the arguments and execution options for a single task invocation.

See also

Subtasks for more information about subtasks.

Performance and Strategies ¶

Granularity ¶

The task granularity is the amount of computation needed by each subtask. In general it is better to split the problem up into many small tasks, than have a few long running tasks.

With smaller tasks you can process more tasks in parallel and the tasks won’t run long enough to block the worker from processing other waiting tasks.

However, executing a task does have overhead. A message needs to be sent, data may not be local, etc. So if the tasks are too fine-grained the additional overhead may not be worth it in the end.

See also

The book Art of Concurrency has a whole section dedicated to the topic of task granularity.

Data locality ¶

The worker processing the task should be as close to the data as possible. The best would be to have a copy in memory, the worst would be a full transfer from another continent.

If the data is far away, you could try to run another worker at location, or if that’s not possible - cache often used data, or preload data you know is going to be used.

The easiest way to share data between workers is to use a distributed cache system, like memcached.

See also

The paper Distributed Computing Economics by Jim Gray is an excellent introduction to the topic of data locality.

State ¶

Since celery is a distributed system, you can’t know in which process, or on what machine the task will be executed. You can’t even know if the task will run in a timely manner.

The ancient async sayings tells us that “asserting the world is the responsibility of the task”. What this means is that the world view may have changed since the task was requested, so the task is responsible for making sure the world is how it should be; If you have a task that re-indexes a search engine, and the search engine should only be re-indexed at maximum every 5 minutes, then it must be the tasks responsibility to assert that, not the callers.

Another gotcha is Django model objects. They shouldn’t be passed on as arguments to tasks. It’s almost always better to re-fetch the object from the database when the task is running instead, as using old data may lead to race conditions.

Imagine the following scenario where you have an article and a task that automatically expands some abbreviations in it:

class Article(models.Model):
    title = models.CharField()
    body = models.TextField()

@task
def expand_abbreviations(article):
    article.body.replace("MyCorp", "My Corporation")
    article.save()

First, an author creates an article and saves it, then the author clicks on a button that initiates the abbreviation task.

>>> article = Article.objects.get(id=102)
>>> expand_abbreviations.delay(model_object)

Now, the queue is very busy, so the task won’t be run for another 2 minutes. In the meantime another author makes changes to the article, so when the task is finally run, the body of the article is reverted to the old version because the task had the old body in its argument.

Fixing the race condition is easy, just use the article id instead, and re-fetch the article in the task body:

@task
def expand_abbreviations(article_id):
    article = Article.objects.get(id=article_id)
    article.body.replace("MyCorp", "My Corporation")
    article.save()

>>> expand_abbreviations(article_id)

There might even be performance benefits to this approach, as sending large messages may be expensive.

Database transactions ¶

Let’s have a look at another example:

from django.db import transaction

@transaction.commit_on_success
def create_article(request):
    article = Article.objects.create(....)
    expand_abbreviations.delay(article.pk)

This is a Django view creating an article object in the database, then passing the primary key to a task. It uses the commit_on_success decorator, which will commit the transaction when the view returns, or roll back if the view raises an exception.

There is a race condition if the task starts executing before the transaction has been committed; The database object does not exist yet!

The solution is to always commit transactions before sending tasks depending on state from the current transaction:

@transaction.commit_manually
def create_article(request):
    try:
        article = Article.objects.create(...)
    except:
        transaction.rollback()
        raise
    else:
        transaction.commit()
        expand_abbreviations.delay(article.pk)

Example ¶

Let’s take a real wold example; A blog where comments posted needs to be filtered for spam. When the comment is created, the spam filter runs in the background, so the user doesn’t have to wait for it to finish.

We have a Django blog application allowing comments on blog posts. We’ll describe parts of the models/views and tasks for this application.

blog/models.py ¶

The comment model looks like this:

from django.db import models
from django.utils.translation import ugettext_lazy as _


class Comment(models.Model):
    name = models.CharField(_("name"), max_length=64)
    email_address = models.EmailField(_("e-mail address"))
    homepage = models.URLField(_("home page"),
                               blank=True, verify_exists=False)
    comment = models.TextField(_("comment"))
    pub_date = models.DateTimeField(_("Published date"),
                                    editable=False, auto_add_now=True)
    is_spam = models.BooleanField(_("spam?"),
                                  default=False, editable=False)

    class Meta:
        verbose_name = _("comment")
        verbose_name_plural = _("comments")

In the view where the comment is posted, we first write the comment to the database, then we launch the spam filter task in the background.

blog/views.py ¶

from django import forms
from django.http import HttpResponseRedirect
from django.template.context import RequestContext
from django.shortcuts import get_object_or_404, render_to_response

from blog import tasks
from blog.models import Comment


class CommentForm(forms.ModelForm):

    class Meta:
        model = Comment


def add_comment(request, slug, template_name="comments/create.html"):
    post = get_object_or_404(Entry, slug=slug)
    remote_addr = request.META.get("REMOTE_ADDR")

    if request.method == "post":
        form = CommentForm(request.POST, request.FILES)
        if form.is_valid():
            comment = form.save()
            # Check spam asynchronously.
            tasks.spam_filter.delay(comment_id=comment.id,
                                    remote_addr=remote_addr)
            return HttpResponseRedirect(post.get_absolute_url())
    else:
        form = CommentForm()

    context = RequestContext(request, {"form": form})
    return render_to_response(template_name, context_instance=context)

To filter spam in comments we use Akismet, the service used to filter spam in comments posted to the free weblog platform Wordpress. Akismet is free for personal use, but for commercial use you need to pay. You have to sign up to their service to get an API key.

To make API calls to Akismet we use the akismet.py library written by Michael Foord.

blog/tasks.py ¶

from akismet import Akismet
from celery.decorators import task

from django.core.exceptions import ImproperlyConfigured
from django.contrib.sites.models import Site

from blog.models import Comment


@task
def spam_filter(comment_id, remote_addr=None, **kwargs):
        logger = spam_filter.get_logger(**kwargs)
        logger.info("Running spam filter for comment %s" % comment_id)

        comment = Comment.objects.get(pk=comment_id)
        current_domain = Site.objects.get_current().domain
        akismet = Akismet(settings.AKISMET_KEY, "http://%s" % domain)
        if not akismet.verify_key():
            raise ImproperlyConfigured("Invalid AKISMET_KEY")


        is_spam = akismet.comment_check(user_ip=remote_addr,
                            comment_content=comment.comment,
                            comment_author=comment.name,
                            comment_author_email=comment.email_address)
        if is_spam:
            comment.is_spam = True
            comment.save()

        return is_spam

Tasks¶

Basics ¶

Default keyword arguments ¶

Logging ¶

Retrying a task if something fails ¶

Using a custom retry delay ¶

Task options ¶

General ¶

Message and routing options ¶

Task names ¶

Automatic naming and relative imports ¶

Decorating tasks ¶

Task States ¶

Built-in States ¶

PENDING ¶

STARTED ¶

SUCCESS ¶

FAILURE ¶

RETRY ¶

REVOKED ¶

Custom states ¶

How it works ¶

Tips and Best Practices ¶

Ignore results you don’t want ¶

Disable rate limits if they’re not used ¶

Avoid launching synchronous subtasks ¶

Performance and Strategies ¶

Granularity ¶

Data locality ¶

State ¶

Database transactions ¶

Example ¶

blog/models.py ¶

blog/views.py ¶

blog/tasks.py ¶

Previous topic

Next topic

This Page

Navigation

Tasks¶

Previous topic

Next topic

This Page

Quick search

Navigation