celery/docs/userguide/workers.rst

.. _guide-worker:

===============
 Workers Guide
===============

.. contents::
    :local:

.. _worker-starting:

Starting the worker
===================

You can start celeryd to run in the foreground by executing the command::

    $ celeryd --loglevel=INFO

You probably want to use a daemonization tool to start
`celeryd` in the background.  See :ref:`daemonizing` for help
using `celeryd` with popular daemonization tools.

For a full list of available command line options see
:mod:`~celery.bin.celeryd`, or simply do::

    $ celeryd --help

You can also start multiple workers on the same machine. If you do so
be sure to give a unique name to each individual worker by specifying a
host name with the :option:`--hostname|-n` argument::

    $ celeryd --loglevel=INFO --concurrency=10 -n worker1.example.com
    $ celeryd --loglevel=INFO --concurrency=10 -n worker2.example.com
    $ celeryd --loglevel=INFO --concurrency=10 -n worker3.example.com

.. _worker-stopping:

Stopping the worker
===================

Shutdown should be accomplished using the :sig:`TERM` signal.

When shutdown is initiated the worker will finish all currently executing
tasks before it actually terminates, so if these tasks are important you should
wait for it to finish before doing anything drastic (like sending the :sig:`KILL`
signal).

If the worker won't shutdown after considerate time, for example because
of tasks stuck in an infinite-loop, you can use the :sig:`KILL` signal to
force terminate the worker, but be aware that currently executing tasks will
be lost (unless the tasks have the :attr:`~@Task.acks_late`
option set).

Also as processes can't override the :sig:`KILL` signal, the worker will
not be able to reap its children, so make sure to do so manually.  This
command usually does the trick::

    $ ps auxww | grep celeryd | awk '{print $2}' | xargs kill -9

.. _worker-restarting:

Restarting the worker
=====================

Other than stopping then starting the worker to restart, you can also
restart the worker using the :sig:`HUP` signal::

    $ kill -HUP $pid

The worker will then replace itself with a new instance using the same
arguments as it was started with.

.. _worker-concurrency:

Concurrency
===========

By default multiprocessing is used to perform concurrent execution of tasks,
but you can also use :ref:`Eventlet <concurrency-eventlet>`.  The number
of worker processes/threads can be changed using the :option:`--concurrency`
argument and defaults to the number of CPUs available on the machine.

.. admonition:: Number of processes (multiprocessing)

    More worker processes are usually better, but there's a cut-off point where
    adding more processes affects performance in negative ways.
    There is even some evidence to support that having multiple celeryd's running,
    may perform better than having a single worker.  For example 3 celeryd's with
    10 worker processes each.  You need to experiment to find the numbers that
    works best for you, as this varies based on application, work load, task
    run times and other factors.

.. _worker-persistent-revokes:

Persistent revokes
==================

Revoking tasks works by sending a broadcast message to all the workers,
the workers then keep a list of revoked tasks in memory.

If you want tasks to remain revoked after worker restart you need to
specify a file for these to be stored in, either by using the `--statedb`
argument to :mod:`~celery.bin.celeryd` or the :setting:`CELERYD_STATE_DB`
setting.  See :setting:`CELERYD_STATE_DB` for more information.

Note that remote control commands must be working for revokes to work.
Remote control commands are only supported by the amqp, redis and mongodb
transports at this point.

.. _worker-time-limits:

Time limits
===========

.. versionadded:: 2.0

:supported pools: processes

A single task can potentially run forever, if you have lots of tasks
waiting for some event that will never happen you will block the worker
from processing new tasks indefinitely.  The best way to defend against
this scenario happening is enabling time limits.

The time limit (`--time-limit`) is the maximum number of seconds a task
may run before the process executing it is terminated and replaced by a
new process.  You can also enable a soft time limit (`--soft-time-limit`),
this raises an exception the task can catch to clean up before the hard
time limit kills it:

.. code-block:: python

    from myapp import celery
    from celery.exceptions import SoftTimeLimitExceeded

    @celery.task
    def mytask():
        try:
            do_work()
        except SoftTimeLimitExceeded:
            clean_up_in_a_hurry()

Time limits can also be set using the :setting:`CELERYD_TASK_TIME_LIMIT` /
:setting:`CELERYD_SOFT_TASK_TIME_LIMIT` settings.

.. note::

    Time limits do not currently work on Windows and other
    platforms that do not support the ``SIGUSR1`` signal.


Changing time limits at runtime
-------------------------------
.. versionadded:: 2.3

You can change the soft and hard time limits for a task by using the
``time_limit`` remote control command.

Example changing the time limit for the ``tasks.crawl_the_web`` task
to have a soft time limit of one minute, and a hard time limit of
two minutes::

    >>> celery.control.time_limit("tasks.crawl_the_web",
                                  soft=60, hard=120, reply=True)
    [{'worker1.example.com': {'ok': 'time limits set successfully'}}]

Only tasks that starts executing after the time limit change will be affected.

.. _worker-maxtasksperchild:

Max tasks per child setting
===========================

.. versionadded:: 2.0

:supported pools: processes

With this option you can configure the maximum number of tasks
a worker can execute before it's replaced by a new process.

This is useful if you have memory leaks you have no control over
for example from closed source C extensions.

The option can be set using the `--maxtasksperchild` argument
to `celeryd` or using the :setting:`CELERYD_MAX_TASKS_PER_CHILD` setting.

.. _worker-autoreload:

Autoreloading
=============

.. versionadded:: 2.5

:supported pools: processes, eventlet, gevent, threads, solo

Starting :program:`celeryd` with the :option:`--autoreload` option will
enable the worker to watch for file system changes to all imported task
modules imported (and also any non-task modules added to the
:setting:`CELERY_IMPORTS` setting or the :option:`-I|--include` option).

This is an experimental feature intended for use in development only,
using auto-reload in production is discouraged as the behavior of reloading
a module in Python is undefined, and may cause hard to diagnose bugs and
crashes.  Celery uses the same approach as the auto-reloader found in e.g.
the Django ``runserver`` command.

When auto-reload is enabled the worker starts an additional thread
that watches for changes in the file system.  New modules are imported,
and already imported modules are reloaded whenever a change is detected,
and if the processes pool is used the child processes will finish the work
they are doing and exit, so that they can be replaced by fresh processes
effectively reloading the code.

File system notification backends are pluggable, and it comes with three
implementations:

* inotify (Linux)

    Used if the :mod:`pyinotify` library is installed.
    If you are running on Linux this is the recommended implementation,
    to install the :mod:`pyinotify` library you have to run the following
    command::

        $ pip install pyinotify

* kqueue (OS X/BSD)

* stat

    The fallback implementation simply polls the files using ``stat`` and is very
    expensive.

You can force an implementation by setting the :envvar:`CELERYD_FSNOTIFY`
environment variable::

    $ env CELERYD_FSNOTIFY=stat celeryd -l info --autoreload

.. _worker-remote-control:

Remote control
==============

.. versionadded:: 2.0

:supported pools: processes, eventlet, gevent, blocking:threads/solo (see note)
:supported transports: amqp, redis, mongodb

Workers have the ability to be remote controlled using a high-priority
broadcast message queue.  The commands can be directed to all, or a specific
list of workers.

Commands can also have replies.  The client can then wait for and collect
those replies.  Since there's no central authority to know how many
workers are available in the cluster, there is also no way to estimate
how many workers may send a reply, so the client has a configurable
timeout — the deadline in seconds for replies to arrive in.  This timeout
defaults to one second.  If the worker doesn't reply within the deadline
it doesn't necessarily mean the worker didn't reply, or worse is dead, but
may simply be caused by network latency or the worker being slow at processing
commands, so adjust the timeout accordingly.

In addition to timeouts, the client can specify the maximum number
of replies to wait for.  If a destination is specified, this limit is set
to the number of destination hosts.

.. seealso::

    The :program:`celeryctl` program is used to execute remote control
    commands from the command line.  It supports all of the commands
    listed below.  See :ref:`monitoring-celeryctl` for more information.

.. note::

    The solo and threads pool supports remote control commands,
    but any task executing will block any waiting control command,
    so it is of limited use if the worker is very busy.  In that
    case you must increase the timeout waitin for replies in the client.

.. _worker-broadcast-fun:

The :meth:`~@control.broadcast` function.
----------------------------------------------------

This is the client function used to send commands to the workers.
Some remote control commands also have higher-level interfaces using
:meth:`~@control.broadcast` in the background, like
:meth:`~@control.rate_limit` and :meth:`~@control.ping`.

Sending the :control:`rate_limit` command and keyword arguments::

    >>> from celery.task.control import broadcast
    >>> celery.control.broadcast("rate_limit",
    ...                          arguments={"task_name": "myapp.mytask",
    ...                                     "rate_limit": "200/m"})

This will send the command asynchronously, without waiting for a reply.
To request a reply you have to use the `reply` argument::

    >>> celery.control.broadcast("rate_limit", {
    ...     "task_name": "myapp.mytask", "rate_limit": "200/m"}, reply=True)
    [{'worker1.example.com': 'New rate limit set successfully'},
     {'worker2.example.com': 'New rate limit set successfully'},
     {'worker3.example.com': 'New rate limit set successfully'}]

Using the `destination` argument you can specify a list of workers
to receive the command::

    >>> celery.control.broadcast("rate_limit", {
    ...     "task_name": "myapp.mytask",
    ...     "rate_limit": "200/m"}, reply=True,
    ...                             destination=["worker1.example.com"])
    [{'worker1.example.com': 'New rate limit set successfully'}]


Of course, using the higher-level interface to set rate limits is much
more convenient, but there are commands that can only be requested
using :meth:`~@control.broadcast`.

.. _worker-rate-limits:

.. control:: rate_limit

Rate limits
-----------

Example changing the rate limit for the `myapp.mytask` task to accept
200 tasks a minute on all servers::

    >>> celery.control.rate_limit("myapp.mytask", "200/m")

Example changing the rate limit on a single host by specifying the
destination host name::

    >>> celery.control.rate_limit("myapp.mytask", "200/m",
    ...            destination=["worker1.example.com"])

.. warning::

    This won't affect workers with the
    :setting:`CELERY_DISABLE_RATE_LIMITS` setting on. To re-enable rate limits
    then you have to restart the worker.

.. control:: revoke

Revoking tasks
--------------

All worker nodes keeps a memory of revoked task ids, either in-memory or
persistent on disk (see :ref:`worker-persistent-revokes`).

When a worker receives a revoke request it will skip executing
the task, but it won't terminate an already executing task unless
the `terminate` option is set.

If `terminate` is set the worker child process processing the task
will be terminated.  The default signal sent is `TERM`, but you can
specify this using the `signal` argument.  Signal can be the uppercase name
of any signal defined in the :mod:`signal` module in the Python Standard
Library.

Terminating a task also revokes it.

**Example**

::

    >>> celery.control.revoke("d9078da5-9915-40a0-bfa1-392c7bde42ed")

    >>> celery.control.revoke("d9078da5-9915-40a0-bfa1-392c7bde42ed",
    ...                       terminate=True)

    >>> celery.control.revoke("d9078da5-9915-40a0-bfa1-392c7bde42ed",
    ...                       terminate=True, signal="SIGKILL")

.. control:: shutdown

Remote shutdown
---------------

This command will gracefully shut down the worker remotely::

    >>> celery.control.broadcast("shutdown") # shutdown all workers
    >>> celery.control.broadcast("shutdown, destination="worker1.example.com")

.. control:: ping

Ping
----

This command requests a ping from alive workers.
The workers reply with the string 'pong', and that's just about it.
It will use the default one second timeout for replies unless you specify
a custom timeout::

    >>> celery.control.ping(timeout=0.5)
    [{'worker1.example.com': 'pong'},
     {'worker2.example.com': 'pong'},
     {'worker3.example.com': 'pong'}]

:meth:`~@control.ping` also supports the `destination` argument,
so you can specify which workers to ping::

    >>> ping(['worker2.example.com', 'worker3.example.com'])
    [{'worker2.example.com': 'pong'},
     {'worker3.example.com': 'pong'}]

.. _worker-enable-events:

.. control:: enable_events
.. control:: disable_events

Enable/disable events
---------------------

You can enable/disable events by using the `enable_events`,
`disable_events` commands.  This is useful to temporarily monitor
a worker using :program:`celeryev`/:program:`celerymon`.

.. code-block:: python

    >>> celery.control.broadcast("enable_events")
    >>> celery.control.broadcast("disable_events")

Adding/Reloading modules
------------------------

.. versionadded:: 2.5

The remote control command ``pool_restart`` sends restart requests to
the workers child processes.  It is particularly useful for forcing
the worker to import new modules, or for reloading already imported
modules.  This command does not interrupt executing tasks.

Example
~~~~~~~

Running the following command will result in the `foo` and `bar` modules
being imported by the worker processes:

.. code-block:: python

    >>> from celery.task.control import broadcast
    >>> celery.control.broadcast("pool_restart",
    ...                          arguments={"modules": ["foo", "bar"]})

Use the ``reload`` argument to reload modules it has already imported:

.. code-block:: python

    >>> celery.control.broadcast("pool_restart",
    ...                          arguments={"modules": ["foo"],
    ...                                     "reload": True})

If you don't specify any modules then all known tasks modules will
be imported/reloaded:

.. code-block:: python

    >>> celery.control.broadcast("pool_restart", arguments={"reload": True})

The ``modules`` argument is a list of modules to modify. ``reload``
specifies whether to reload modules if they have previously been imported.
By default ``reload`` is disabled. The `pool_restart` command uses the
Python :func:`reload` function to reload modules, or you can provide
your own custom reloader by passing the ``reloader`` argument.

.. note::

    Module reloading comes with caveats that are documented in :func:`reload`.
    Please read this documentation and make sure your modules are suitable
    for reloading.

.. seealso::

    - http://pyunit.sourceforge.net/notes/reloading.html
    - http://www.indelible.org/ink/python-reloading/
    - http://docs.python.org/library/functions.html#reload

.. _worker-custom-control-commands:

Writing your own remote control commands
----------------------------------------

Remote control commands are registered in the control panel and
they take a single argument: the current
:class:`~celery.worker.control.ControlDispatch` instance.
From there you have access to the active
:class:`~celery.worker.consumer.Consumer` if needed.

Here's an example control command that restarts the broker connection:

.. code-block:: python

    from celery.worker.control import Panel

    @Panel.register
    def reset_connection(panel):
        panel.logger.critical("Connection reset by remote control.")
        panel.consumer.reset_connection()
        return {"ok": "connection reset"}


These can be added to task modules, or you can keep them in their own module
then import them using the :setting:`CELERY_IMPORTS` setting::

    CELERY_IMPORTS = ("myapp.worker.control", )

.. _worker-inspect:

Inspecting workers
==================

:class:`@control.inspect` lets you inspect running workers.  It
uses remote control commands under the hood.

.. code-block:: python

    # Inspect all nodes.
    >>> i = celery.control.inspect()

    # Specify multiple nodes to inspect.
    >>> i = celery.control.inspect(["worker1.example.com",
                                    "worker2.example.com"])

    # Specify a single node to inspect.
    >>> i = celery.control.inspect("worker1.example.com")

.. _worker-inspect-registered-tasks:

Dump of registered tasks
------------------------

You can get a list of tasks registered in the worker using the
:meth:`~@control.inspect.registered`::

    >>> i.registered()
    [{'worker1.example.com': ['celery.delete_expired_task_meta',
                              'celery.execute_remote',
                              'celery.map_async',
                              'celery.ping',
                              'celery.task.http.HttpDispatchTask',
                              'tasks.add',
                              'tasks.sleeptask']}]

.. _worker-inspect-active-tasks:

Dump of currently executing tasks
---------------------------------

You can get a list of active tasks using
:meth:`~@control.inspect.active`::

    >>> i.active()
    [{'worker1.example.com':
        [{"name": "tasks.sleeptask",
          "id": "32666e9b-809c-41fa-8e93-5ae0c80afbbf",
          "args": "(8,)",
          "kwargs": "{}"}]}]

.. _worker-inspect-eta-schedule:

Dump of scheduled (ETA) tasks
-----------------------------

You can get a list of tasks waiting to be scheduled by using
:meth:`~@control.inspect.scheduled`::

    >>> i.scheduled()
    [{'worker1.example.com':
        [{"eta": "2010-06-07 09:07:52", "priority": 0,
          "request": {
            "name": "tasks.sleeptask",
            "id": "1a7980ea-8b19-413e-91d2-0b74f3844c4d",
            "args": "[1]",
            "kwargs": "{}"}},
         {"eta": "2010-06-07 09:07:53", "priority": 0,
          "request": {
            "name": "tasks.sleeptask",
            "id": "49661b9a-aa22-4120-94b7-9ee8031d219d",
            "args": "[2]",
            "kwargs": "{}"}}]}]

Note that these are tasks with an eta/countdown argument, not periodic tasks.

.. _worker-inspect-reserved:

Dump of reserved tasks
----------------------

Reserved tasks are tasks that has been received, but is still waiting to be
executed.

You can get a list of these using
:meth:`~@control.inspect.reserved`::

    >>> i.reserved()
    [{'worker1.example.com':
        [{"name": "tasks.sleeptask",
          "id": "32666e9b-809c-41fa-8e93-5ae0c80afbbf",
          "args": "(8,)",
          "kwargs": "{}"}]}]