This document describes the current stable version of Celery (4.0). For development docs, go here.



This document is fairly extensive and you aren’t really expected to study this in detail for small contributions;

The most important rule is that contributing must be easy and that the community is friendly and not nitpicking on details, such as coding style.

If you’re reporting a bug you should read the Reporting bugs section below to ensure that your bug report contains enough information to successfully diagnose the issue, and if you’re contributing code you should try to mimic the conventions you see surrounding the code you’re working on, but in the end all patches will be cleaned up by the person merging the changes so don’t worry too much.

Community Code of Conduct

The goal is to maintain a diverse community that’s pleasant for everyone. That’s why we would greatly appreciate it if everyone contributing to and interacting with the community also followed this Code of Conduct.

The Code of Conduct covers our behavior as members of the community, in any forum, mailing list, wiki, website, Internet relay chat (IRC), public meeting or private correspondence.

The Code of Conduct is heavily based on the Ubuntu Code of Conduct, and the Pylons Code of Conduct.

Be considerate

Your work will be used by other people, and you in turn will depend on the work of others. Any decision you take will affect users and colleagues, and we expect you to take those consequences into account when making decisions. Even if it’s not obvious at the time, our contributions to Celery will impact the work of others. For example, changes to code, infrastructure, policy, documentation and translations during a release may negatively impact others work.

Be respectful

The Celery community and its members treat one another with respect. Everyone can make a valuable contribution to Celery. We may not always agree, but disagreement is no excuse for poor behavior and poor manners. We might all experience some frustration now and then, but we cannot allow that frustration to turn into a personal attack. It’s important to remember that a community where people feel uncomfortable or threatened isn’t a productive one. We expect members of the Celery community to be respectful when dealing with other contributors as well as with people outside the Celery project and with users of Celery.

Be collaborative

Collaboration is central to Celery and to the larger free software community. We should always be open to collaboration. Your work should be done transparently and patches from Celery should be given back to the community when they’re made, not just when the distribution releases. If you wish to work on new code for existing upstream projects, at least keep those projects informed of your ideas and progress. It many not be possible to get consensus from upstream, or even from your colleagues about the correct implementation for an idea, so don’t feel obliged to have that agreement before you begin, but at least keep the outside world informed of your work, and publish your work in a way that allows outsiders to test, discuss, and contribute to your efforts.

When you disagree, consult others

Disagreements, both political and technical, happen all the time and the Celery community is no exception. It’s important that we resolve disagreements and differing views constructively and with the help of the community and community process. If you really want to go a different way, then we encourage you to make a derivative distribution or alternate set of packages that still build on the work we’ve done to utilize as common of a core as possible.

When you’re unsure, ask for help

Nobody knows everything, and nobody is expected to be perfect. Asking questions avoids many problems down the road, and so questions are encouraged. Those who are asked questions should be responsive and helpful. However, when asking a question, care must be taken to do so in an appropriate forum.

Step down considerately

Developers on every project come and go and Celery is no different. When you leave or disengage from the project, in whole or in part, we ask that you do so in a way that minimizes disruption to the project. This means you should tell people you’re leaving and take the proper steps to ensure that others can pick up where you leave off.

Reporting Bugs


You must never report security related issues, vulnerabilities or bugs including sensitive information to the bug tracker, or elsewhere in public. Instead sensitive bugs must be sent by email to

If you’d like to submit the information encrypted our PGP key is:

Version: GnuPG v1.4.15 (Darwin)


Other bugs

Bugs can always be described to the Mailing list, but the best way to report an issue and to ensure a timely response is to use the issue tracker.

  1. Create a GitHub account.

You need to create a GitHub account to be able to create new issues and participate in the discussion.

  1. Determine if your bug is really a bug.

You shouldn’t file a bug if you’re requesting support. For that you can use the Mailing list, or IRC.

  1. Make sure your bug hasn’t already been reported.

Search through the appropriate Issue tracker. If a bug like yours was found, check if you have new information that could be reported to help the developers fix the bug.

  1. Check if you’re using the latest version.

A bug could be fixed by some other improvements and fixes - it might not have an existing report in the bug tracker. Make sure you’re using the latest releases of celery, billiard, kombu, amqp, and vine.

  1. Collect information about the bug.

To have the best chance of having a bug fixed, we need to be able to easily reproduce the conditions that caused it. Most of the time this information will be from a Python traceback message, though some bugs might be in design, spelling or other errors on the website/docs/code.

  1. If the error is from a Python traceback, include it in the bug report.

  2. We also need to know what platform you’re running (Windows, macOS, Linux, etc.), the version of your Python interpreter, and the version of Celery, and related packages that you were running when the bug occurred.

  3. If you’re reporting a race condition or a deadlock, tracebacks can be hard to get or might not be that useful. Try to inspect the process to get more diagnostic data. Some ideas:

    • Enable Celery’s breakpoint signal and use it to inspect the process’s state. This will allow you to open a pdb session.
    • Collect tracing data using strace`_(Linux), :command:`dtruss (macOS), and ktrace (BSD), ltrace, and lsof.
  4. Include the output from the celery report command:

    $ celery -A proj report

    This will also include your configuration settings and it try to remove values for keys known to be sensitive, but make sure you also verify the information before submitting so that it doesn’t contain confidential information like API tokens and authentication credentials.

  1. Submit the bug.

By default GitHub will email you to let you know when new comments have been made on your bug. In the event you’ve turned this feature off, you should check back on occasion to ensure you don’t miss any questions a developer trying to fix the bug might ask.

Contributors guide to the code base

There’s a separate section for internal details, including details about the code base and a style guide.

Read Contributors Guide to the Code for more!


Version numbers consists of a major version, minor version and a release number. Since version 2.1.0 we use the versioning semantics described by SemVer:

Stable releases are published at PyPI while development releases are only available in the GitHub git repository as tags. All version tags starts with “v”, so version 0.8.0 is the tag v0.8.0.


Current active version branches:

You can see the state of any branch by looking at the Changelog:

If the branch is in active development the topmost version info should contain meta-data like:

:release-date: TBA
:branch: dev (git calls this master)

The status field can be one of:


    The branch is currently experimental and in the planning stage.


    The branch is in active development, but the test suite should be passing and the product should be working and possible for users to test.


    The branch is frozen, and no more features will be accepted. When a branch is frozen the focus is on testing the version as much as possible before it is released.

dev branch

The dev branch (called “master” by git), is where development of the next version happens.

Maintenance branches

Maintenance branches are named after the version – for example, the maintenance branch for the 2.2.x series is named 2.2.

Previously these were named releaseXX-maint.

The versions we currently maintain is:

  • 3.1

    This is the current series.

  • 3.0

    This is the previous series, and the last version to support Python 2.5.

Archived branches

Archived branches are kept for preserving history only, and theoretically someone could provide patches for these if they depend on a series that’s no longer officially supported.

An archived version is named X.Y-archived.

Our currently archived branches are:

Feature branches

Major new features are worked on in dedicated branches. There’s no strict naming requirement for these branches.

Feature branches are removed once they’ve been merged into a release branch.


  • Tags are used exclusively for tagging releases. A release tag is named with the format vX.Y.Z – for example v2.3.1.
  • Experimental releases contain an additional identifier vX.Y.Z-id – for example v3.0.0-rc1.
  • Experimental tags may be removed after the official release.

Working on Features & Patches


Contributing to Celery should be as simple as possible, so none of these steps should be considered mandatory.

You can even send in patches by email if that’s your preferred work method. We won’t like you any less, any contribution you make is always appreciated!

However following these steps may make maintainers life easier, and may mean that your changes will be accepted sooner.

Forking and setting up the repository

First you need to fork the Celery repository, a good introduction to this is in the GitHub Guide: Fork a Repo.

After you have cloned the repository you should checkout your copy to a directory on your machine:

$ git clone

When the repository is cloned enter the directory to set up easy access to upstream changes:

$ cd celery
$ git remote add upstream git://
$ git fetch upstream

If you need to pull in new changes from upstream you should always use the --rebase option to git pull:

git pull --rebase upstream master

With this option you don’t clutter the history with merging commit notes. See Rebasing merge commits in git. If you want to learn more about rebasing see the Rebase section in the GitHub guides.

If you need to work on a different branch than the one git calls master, you can fetch and checkout a remote branch like this:

git checkout --track -b 3.0-devel origin/3.0-devel

Running the unit test suite

To run the Celery test suite you need to install a few dependencies. A complete list of the dependencies needed are located in requirements/test.txt.

If you’re working on the development version, then you need to install the development requirements first:

$ pip install -U -r requirements/dev.txt


Both the stable and the development version have testing related dependencies, so install these next:

$ pip install -U -r requirements/test.txt
$ pip install -U -r requirements/default.txt

After installing the dependencies required, you can now execute the test suite by calling py.test <pytest:

$ py.test

Some useful options to py.test are:

  • -x

    Stop running the tests at the first test that fails.

  • -s

    Don’t capture output

  • -v

    Run with verbose output.

If you want to run the tests for a single test file only you can do so like this:

$ py.test t/unit/worker/

Creating pull requests

When your feature/bugfix is complete you may want to submit a pull requests so that it can be reviewed by the maintainers.

Creating pull requests is easy, and also let you track the progress of your contribution. Read the Pull Requests section in the GitHub Guide to learn how this is done.

You can also attach pull requests to existing issues by following the steps outlined here:

Calculating test coverage

To calculate test coverage you must first install the pytest-cov module.

Installing the pytest-cov module:

$ pip install -U pytest-cov
Code coverage in HTML format
  1. Run py.test with the --cov-report=html argument enabled:

    $ py.test --cov=celery --cov-report=html
  2. The coverage output will then be located in the htmlcov/ directory:

    $ open htmlcov/index.html
Code coverage in XML (Cobertura-style)
  1. Run py.test with the --cov-report=xml argument enabled:
$ py.test --cov=celery --cov-report=xml
  1. The coverage XML output will then be located in the coverage.xml file.

Running the tests on all supported Python versions

There’s a tox configuration file in the top directory of the distribution.

To run the tests for all supported Python versions simply execute:

$ tox

Use the tox -e option if you only want to test specific Python versions:

$ tox -e 2.7

Building the documentation

To build the documentation you need to install the dependencies listed in requirements/docs.txt:

$ pip install -U -r requirements/docs.txt

After these dependencies are installed you should be able to build the docs by running:

$ cd docs
$ rm -rf _build
$ make html

Make sure there are no errors or warnings in the build output. After building succeeds the documentation is available at _build/html.

Verifying your contribution

To use these tools you need to install a few dependencies. These dependencies can be found in requirements/pkgutils.txt.

Installing the dependencies:

$ pip install -U -r requirements/pkgutils.txt

pyflakes & PEP-8

To ensure that your changes conform to PEP 8 and to run pyflakes execute:

$ make flakecheck

To not return a negative exit code when this command fails use the flakes target instead:

$ make flakes§

API reference

To make sure that all modules have a corresponding section in the API reference please execute:

$ make apicheck
$ make indexcheck

If files are missing you can add them by copying an existing reference file.

If the module is internal it should be part of the internal reference located in docs/internals/reference/. If the module is public it should be located in docs/reference/.

For example if reference is missing for the module celery.worker.awesome and this module is considered part of the public API, use the following steps:

Use an existing file as a template:

$ cd docs/reference/
$ cp celery.schedules.rst celery.worker.awesome.rst

Edit the file using your favorite editor:

$ vim celery.worker.awesome.rst

    # change every occurrence of ``celery.schedules`` to
    # ``celery.worker.awesome``

Edit the index using your favorite editor:

$ vim index.rst

    # Add ``celery.worker.awesome`` to the index.

Commit your changes:

# Add the file to git
$ git add celery.worker.awesome.rst
$ git add index.rst
$ git commit celery.worker.awesome.rst index.rst \
    -m "Adds reference for celery.worker.awesome"

Coding Style

You should probably be able to pick up the coding style from surrounding code, but it is a good idea to be aware of the following conventions.

  • All Python code must follow the PEP 8 guidelines.

pep8 is a utility you can use to verify that your code is following the conventions.

  • Docstrings must follow the PEP 257 conventions, and use the following style.

    Do this:

    def method(self, arg):
        """Short description.
        More details.


    def method(self, arg):
        """Short description."""

    but not this:

    def method(self, arg):
        Short description.
  • Lines shouldn’t exceed 78 columns.

    You can enforce this in vim by setting the textwidth option:

    set textwidth=78

    If adhering to this limit makes the code less readable, you have one more character to go on. This means 78 is a soft limit, and 79 is the hard limit :)

  • Import order

    • Python standard library (import xxx)
    • Python standard library (‘from xxx import`)
    • Third-party packages.
    • Other modules from the current package.

    or in case of code using Django:

    • Python standard library (import xxx)
    • Python standard library (‘from xxx import`)
    • Third-party packages.
    • Django packages.
    • Other modules from the current package.

    Within these sections the imports should be sorted by module name.


    import threading
    import time
    from collections import deque
    from Queue import Queue, Empty
    from .platforms import Pidfile
    from .five import zip_longest, items, range
    from .utils.time import maybe_timedelta
  • Wild-card imports must not be used (from xxx import *).

  • For distributions where Python 2.5 is the oldest support version additional rules apply:

    • Absolute imports must be enabled at the top of every module:

      from __future__ import absolute_import
    • If the module uses the with statement and must be compatible with Python 2.5 (celery isn’t) then it must also enable that:

      from __future__ import with_statement
    • Every future import must be on its own line, as older Python 2.5 releases didn’t support importing multiple features on the same future import line:

      # Good
      from __future__ import absolute_import
      from __future__ import with_statement
      # Bad
      from __future__ import absolute_import, with_statement

    (Note that this rule doesn’t apply if the package doesn’t include support for Python 2.5)

  • Note that we use “new-style` relative imports when the distribution doesn’t support Python versions below 2.5

    This requires Python 2.5 or later:

    from . import submodule

Contributing features requiring additional libraries

Some features like a new result backend may require additional libraries that the user must install.

We use setuptools extra_requires for this, and all new optional features that require third-party libraries must be added.

  1. Add a new requirements file in requirements/extras

    For the Cassandra backend this is requirements/extras/cassandra.txt, and the file looks like this:


    These are pip requirement files so you can have version specifiers and multiple packages are separated by newline. A more complex example could be:

    # pycassa 2.0 breaks Foo
  2. Modify

    After the requirements file is added you need to add it as an option to in the extras_require section:

    extra['extras_require'] = {
        # ...
        'cassandra': extras('cassandra.txt'),
  3. Document the new feature in docs/includes/installation.txt

    You must add your feature to the list in the Bundles section of docs/includes/installation.txt.

    After you’ve made changes to this file you need to render the distro README file:

    $ pip install -U requirements/pkgutils.txt
    $ make readme

That’s all that needs to be done, but remember that if your feature adds additional configuration options then these needs to be documented in docs/configuration.rst. Also all settings need to be added to the celery/app/ module.

Result backends require a separate section in the docs/configuration.rst file.



Fork of multiprocessing containing improvements that’ll eventually be merged into the Python stdlib.



Very fast Python AMQP client written in C.



Actor library.



Distributed Celery Instance manager.


Release Procedure

Updating the version number

The version number must be updated two places:

  • celery/
  • docs/include/introduction.txt

After you have changed these files you must render the README files. There’s a script to convert sphinx syntax to generic reStructured Text syntax, and the make target readme does this for you:

$ make readme

Now commit the changes:

$ git commit -a -m "Bumps version to X.Y.Z"

and make a new version tag:

$ git tag vX.Y.Z
$ git push --tags


Commands to make a new public stable release:

$ make distcheck  # checks pep8, autodoc index, runs tests and more
$ make dist  # NOTE: Runs git clean -xdf and removes files not in the repo.
$ python sdist upload --sign --identity='Celery Security Team'
$ python bdist_wheel upload --sign --identity='Celery Security Team'

If this is a new release series then you also need to do the following:

  • Go to the Read The Docs management interface at:

  • Enter “Edit project”

    Change default branch to the branch of this series, for example, use the 2.4 branch for the 2.4 series.

  • Also add the previous version under the “versions” tab.