This document describes an older version of Celery (2.5). For the latest stable version please go here.

Executing Tasks


Executing a task is done with apply_async(), and the shortcut: delay().

delay is simple and convenient, as it looks like calling a regular function:

Task.delay(arg1, arg2, kwarg1="x", kwarg2="y")

The same using apply_async is written like this:

Task.apply_async(args=[arg1, arg2], kwargs={"kwarg1": "x", "kwarg2": "y"})

While delay is convenient, it doesn’t give you as much control as using apply_async. With apply_async you can override the execution options available as attributes on the Task class (see Task options). In addition you can set countdown/eta, task expiry, provide a custom broker connection and more.

Let’s go over these in more detail. All the examples uses a simple task called add, returning the sum of two positional arguments:

def add(x, y):
    return x + y


You can also execute a task by name using send_task(), if you don’t have access to the task class:

>>> from celery.execute import send_task
>>> result = send_task("tasks.add", [2, 2])
>>> result.get()

ETA and countdown

The ETA (estimated time of arrival) lets you set a specific date and time that is the earliest time at which your task will be executed. countdown is a shortcut to set eta by seconds into the future.

>>> result = add.apply_async(args=[10, 10], countdown=3)
>>> result.get()    # this takes at least 3 seconds to return

The task is guaranteed to be executed at some time after the specified date and time, but not necessarily at that exact time. Possible reasons for broken deadlines may include many items waiting in the queue, or heavy network latency. To make sure your tasks are executed in a timely manner you should monitor queue lengths. Use Munin, or similar tools, to receive alerts, so appropriate action can be taken to ease the workload. See Munin.

While countdown is an integer, eta must be a datetime object, specifying an exact date and time (including millisecond precision, and timezone information):

>>> from datetime import datetime, timedelta

>>> tomorrow = + timedelta(days=1)
>>> add.apply_async(args=[10, 10], eta=tomorrow)


The expires argument defines an optional expiry time, either as seconds after task publish, or a specific date and time using datetime:

>>> # Task expires after one minute from now.
>>> add.apply_async(args=[10, 10], expires=60)

>>> # Also supports datetime
>>> from datetime import datetime, timedelta
>>> add.apply_async(args=[10, 10], kwargs,
...        + timedelta(days=1)

When a worker receives an expired task it will mark the task as REVOKED (TaskRevokedError).


Data transferred between clients and workers needs to be serialized. The default serializer is pickle, but you can change this globally or for each individual task. There is built-in support for pickle, JSON, YAML and msgpack, and you can also add your own custom serializers by registering them into the Kombu serializer registry (see Kombu: Serialization of Data).

Each option has its advantages and disadvantages.

json – JSON is supported in many programming languages, is now

a standard part of Python (since 2.6), and is fairly fast to decode using the modern Python libraries such as cjson or simplejson.

The primary disadvantage to JSON is that it limits you to the following data types: strings, Unicode, floats, boolean, dictionaries, and lists. Decimals and dates are notably missing.

Also, binary data will be transferred using Base64 encoding, which will cause the transferred data to be around 34% larger than an encoding which supports native binary types.

However, if your data fits inside the above constraints and you need cross-language support, the default setting of JSON is probably your best choice.

See for more information.

pickle – If you have no desire to support any language other than

Python, then using the pickle encoding will gain you the support of all built-in Python data types (except class instances), smaller messages when sending binary files, and a slight speedup over JSON processing.

See for more information.

yaml – YAML has many of the same characteristics as json,

except that it natively supports more data types (including dates, recursive references, etc.)

However, the Python libraries for YAML are a good bit slower than the libraries for JSON.

If you need a more expressive set of data types and need to maintain cross-language compatibility, then YAML may be a better fit than the above.

See for more information.

msgpack – msgpack is a binary serialization format that is closer to JSON

in features. It is very young however, and support should be considered experimental at this point.

See for more information.

The encoding used is available as a message header, so the worker knows how to deserialize any task. If you use a custom serializer, this serializer must be available for the worker.

The client uses the following order to decide which serializer to use when sending a task:

  1. The serializer argument to apply_async
  2. The tasks serializer attribute
  3. The default CELERY_TASK_SERIALIZER setting.
  • Using the serializer argument to apply_async:
>>> add.apply_async(args=[10, 10], serializer="json")

Connections and connection timeouts.

Automatic Pool Support

Since version 2.3 there is support for automatic connection pools, so you don’t have to manually handle connections and publishers to reuse connections.

The connection pool is enabled by default since version 2.5.

See the BROKER_POOL_LIMIT setting for more information.

You can handle the connection manually by creating a publisher:

numbers = [(2, 2), (4, 4), (8, 8), (16, 16)]

results = []
publisher = add.get_publisher()
    for args in numbers:
        res = add.apply_async(args=args, publisher=publisher)

print([res.get() for res in results])


This particular example is better expressed as a task set. See Task Sets. Tasksets already reuses connections.

The connection timeout is the number of seconds to wait before giving up on establishing the connection. You can set this by using the connect_timeout argument to apply_async:

add.apply_async([10, 10], connect_timeout=3)

Or if you handle the connection manually:

publisher = add.get_publisher(connect_timeout=3)

Routing options

Celery uses the AMQP routing mechanisms to route tasks to different workers.

Messages (tasks) are sent to exchanges, a queue binds to an exchange with a routing key. Let’s look at an example:

Let’s pretend we have an application with lot of different tasks: some process video, others process images, and some gather collective intelligence about its users. Some of these tasks are more important, so we want to make sure the high priority tasks get sent to dedicated nodes.

For the sake of this example we have a single exchange called tasks. There are different types of exchanges, each type interpreting the routing key in different ways, implementing different messaging scenarios.

The most common types used with Celery are direct and topic.

  • direct

    Matches the routing key exactly.

  • topic

    In the topic exchange the routing key is made up of words separated by dots (.). Words can be matched by the wild cards * and #, where * matches one exact word, and # matches one or many words.

    For example, *.stock.# matches the routing keys usd.stock and euro.stock.db but not stock.nasdaq.

We create three queues, video, image and lowpri that binds to the tasks exchange. For the queues we use the following binding keys:

video: video.#
image: image.#
lowpri: misc.#

Now we can send our tasks to different worker machines, by making the workers listen to different queues:

>>> add.apply_async(args=[filename],
...                               routing_key="video.compress")

>>> add.apply_async(args=[filename, 360],
...                             routing_key="image.rotate")

>>> add.apply_async(args=[filename, selection],
...                           routing_key="image.crop")
>>> add.apply_async(routing_key="misc.recommend")

Later, if the crop task is consuming a lot of resources, we can bind new workers to handle just the “image.crop” task, by creating a new queue that binds to “image.crop”.

See also

To find out more about routing, please see Routing Tasks.

AMQP options

  • mandatory

This sets the delivery to be mandatory. An exception will be raised if there are no running workers able to take on the task.

Not supported by amqplib.

  • immediate

Request immediate delivery. Will raise an exception if the task cannot be routed to a worker immediately.

Not supported by amqplib.

  • priority

A number between 0 and 9, where 0 is the highest priority.


RabbitMQ does not yet support AMQP priorities.

Previous topic


Next topic

Workers Guide

This Page