/django-rq

A simple app that provides django integration for RQ (Redis Queue)

Primary LanguagePythonMIT LicenseMIT

Django-RQ

Build Status

Django integration with RQ, a Redis based Python queuing library. Django-RQ is a simple app that allows you to configure your queues in django's settings.py and easily use them in your project.

Support Django-RQ

If you find django-rq useful, please consider supporting its development via Tidelift.

Requirements

Installation

pip install django-rq
  • Add django_rq to INSTALLED_APPS in settings.py:
INSTALLED_APPS = (
    # other apps
    "django_rq",
)
  • Configure your queues in django's settings.py:
RQ_QUEUES = {
    'default': {
        'HOST': 'localhost',
        'PORT': 6379,
        'DB': 0,
        'USERNAME': 'some-user',
        'PASSWORD': 'some-password',
        'DEFAULT_TIMEOUT': 360,
        'REDIS_CLIENT_KWARGS': {    # Eventual additional Redis connection arguments
            'ssl_cert_reqs': None,
        },
    },
    'with-sentinel': {
        'SENTINELS': [('localhost', 26736), ('localhost', 26737)],
        'MASTER_NAME': 'redismaster',
        'DB': 0,
        # Redis username/password
        'USERNAME': 'redis-user',
        'PASSWORD': 'secret',
        'SOCKET_TIMEOUT': 0.3,
        'CONNECTION_KWARGS': {  # Eventual additional Redis connection arguments
            'ssl': True
        },
        'SENTINEL_KWARGS': {    # Eventual Sentinel connection arguments
            # If Sentinel also has auth, username/password can be passed here
            'username': 'sentinel-user',
            'password': 'secret',
        },
    },
    'high': {
        'URL': os.getenv('REDISTOGO_URL', 'redis://localhost:6379/0'), # If you're on Heroku
        'DEFAULT_TIMEOUT': 500,
    },
    'low': {
        'HOST': 'localhost',
        'PORT': 6379,
        'DB': 0,
    }
}

RQ_EXCEPTION_HANDLERS = ['path.to.my.handler'] # If you need custom exception handlers
  • Include django_rq.urls in your urls.py:
urlpatterns += [
    path('django-rq/', include('django_rq.urls'))
]

Usage

Putting jobs in the queue

Django-RQ allows you to easily put jobs into any of the queues defined in settings.py. It comes with a few utility functions:

  • enqueue - push a job to the default queue:
import django_rq
django_rq.enqueue(func, foo, bar=baz)
  • get_queue - returns an Queue instance.
import django_rq
queue = django_rq.get_queue('high')
queue.enqueue(func, foo, bar=baz)

In addition to name argument, get_queue also accepts default_timeout, is_async, autocommit, connection and queue_class arguments. For example:

queue = django_rq.get_queue('default', autocommit=True, is_async=True, default_timeout=360)
queue.enqueue(func, foo, bar=baz)

You can provide your own singleton Redis connection object to this function so that it will not create a new connection object for each queue definition. This will help you limit number of connections to Redis server. For example:

import django_rq
import redis
redis_cursor = redis.StrictRedis(host='', port='', db='', password='')
high_queue = django_rq.get_queue('high', connection=redis_cursor)
low_queue = django_rq.get_queue('low', connection=redis_cursor)
  • get_connection - accepts a single queue name argument (defaults to "default") and returns a connection to the queue's Redis server:
import django_rq
redis_conn = django_rq.get_connection('high')
  • get_worker - accepts optional queue names and returns a new RQ Worker instance for specified queues (or default queue):
import django_rq
worker = django_rq.get_worker() # Returns a worker for "default" queue
worker.work()
worker = django_rq.get_worker('low', 'high') # Returns a worker for "low" and "high"

@job decorator

To easily turn a callable into an RQ task, you can also use the @job decorator that comes with django_rq:

from django_rq import job

@job
def long_running_func():
    pass
long_running_func.delay() # Enqueue function in "default" queue

@job('high')
def long_running_func():
    pass
long_running_func.delay() # Enqueue function in "high" queue

You can pass in any arguments that RQ's job decorator accepts:

@job('default', timeout=3600)
def long_running_func():
    pass
long_running_func.delay() # Enqueue function with a timeout of 3600 seconds.

It's possible to specify default for result_ttl decorator keyword argument via DEFAULT_RESULT_TTL setting:

RQ = {
    'DEFAULT_RESULT_TTL': 5000,
}

With this setting, job decorator will set result_ttl to 5000 unless it's specified explicitly.

Running workers

django_rq provides a management command that starts a worker for every queue specified as arguments:

python manage.py rqworker high default low

If you want to run rqworker in burst mode, you can pass in the --burst flag:

python manage.py rqworker high default low --burst

If you need to use custom worker, job or queue classes, it is best to use global settings (see Custom queue classes and Custom job and worker classes). However, it is also possible to override such settings with command line options as follows.

To use a custom worker class, you can pass in the --worker-class flag with the path to your worker:

python manage.py rqworker high default low --worker-class 'path.to.GeventWorker'

To use a custom queue class, you can pass in the --queue-class flag with the path to your queue class:

python manage.py rqworker high default low --queue-class 'path.to.CustomQueue'

To use a custom job class, provide --job-class flag.

Starting from version 2.10, running RQ's worker-pool is also supported:

python manage.py rqworker-pool default low medium --num-workers 4

Support for Scheduled Jobs

With RQ 1.2.0. you can use built-in scheduler for your jobs. For example:

from django_rq.queues import get_queue
queue = get_queue('default')
job = queue.enqueue_at(datetime(2020, 10, 10), func)

If you are using built-in scheduler you have to start workers with scheduler support:

python manage.py rqworker --with-scheduler

Alternatively you can use RQ Scheduler. After install you can also use the get_scheduler function to return a Scheduler instance for queues defined in settings.py's RQ_QUEUES. For example:

import django_rq
scheduler = django_rq.get_scheduler('default')
job = scheduler.enqueue_at(datetime(2020, 10, 10), func)

You can also use the management command rqscheduler to start the scheduler:

python manage.py rqscheduler

Support for django-redis and django-redis-cache

If you have django-redis or django-redis-cache installed, you can instruct django_rq to use the same connection information from your Redis cache. This has two advantages: it's DRY and it takes advantage of any optimization that may be going on in your cache setup (like using connection pooling or Hiredis.)

To use configure it, use a dict with the key USE_REDIS_CACHE pointing to the name of the desired cache in your RQ_QUEUES dict. It goes without saying that the chosen cache must exist and use the Redis backend. See your respective Redis cache package docs for configuration instructions. It's also important to point out that since the django-redis-cache ShardedClient splits the cache over multiple Redis connections, it does not work.

Here is an example settings fragment for django-redis:

CACHES = {
    'redis-cache': {
        'BACKEND': 'redis_cache.cache.RedisCache',
        'LOCATION': 'localhost:6379:1',
        'OPTIONS': {
            'CLIENT_CLASS': 'django_redis.client.DefaultClient',
            'MAX_ENTRIES': 5000,
        },
    },
}

RQ_QUEUES = {
    'high': {
        'USE_REDIS_CACHE': 'redis-cache',
    },
    'low': {
        'USE_REDIS_CACHE': 'redis-cache',
    },
}

Suspending and Resuming Workers

Sometimes you may want to suspend RQ to prevent it from processing new jobs. A classic example is during the initial phase of a deployment script or in advance of putting your site into maintenance mode. This is particularly helpful when you have jobs that are relatively long-running and might otherwise be forcibly killed during the deploy.

The suspend command stops workers on _all_ queues (in a single Redis database) from picking up new jobs. However currently running jobs will continue until completion.

# Suspend indefinitely
python manage.py rqsuspend

# Suspend for a specific duration (in seconds) then automatically
# resume work again.
python manage.py rqsuspend -d 600

# Resume work again.
python manage.py rqresume

Queue Statistics

django_rq also provides a dashboard to monitor the status of your queues at /django-rq/ (or whatever URL you set in your urls.py during installation.

You can also add a link to this dashboard link in /admin by adding RQ_SHOW_ADMIN_LINK = True in settings.py. Be careful though, this will override the default admin template so it may interfere with other apps that modifies the default admin template.

These statistics are also available in JSON format via /django-rq/stats.json, which is accessible to staff members. If you need to access this view via other HTTP clients (for monitoring purposes), you can define RQ_API_TOKEN and access it via /django-rq/stats.json/<API_TOKEN>.

demo-django-rq-json-dashboard.png

Note: Statistics of scheduled jobs display jobs from RQ built-in scheduler, not optional RQ scheduler.

Additionally, these statistics are also accessible from the command line.

python manage.py rqstats
python manage.py rqstats --interval=1  # Refreshes every second
python manage.py rqstats --json  # Output as JSON
python manage.py rqstats --yaml  # Output as YAML

demo-django-rq-cli-dashboard.gif

Configuring Sentry

Sentry should be configured within the Django settings.py as described in the Sentry docs.

You can override the default Django Sentry configuration when running the rqworker command by passing the sentry-dsn option:

./manage.py rqworker --sentry-dsn=https://*****@sentry.io/222222

This will override any existing Django configuration and reinitialise Sentry, setting the following Sentry options:

{
    'debug': options.get('sentry_debug'),
    'ca_certs': options.get('sentry_ca_certs'),
    'integrations': [RedisIntegration(), RqIntegration(), DjangoIntegration()]
}

Configuring Logging

RQ uses Python's logging, this means you can easily configure rqworker's logging mechanism in django's settings.py. For example:

LOGGING = {
    "version": 1,
    "disable_existing_loggers": False,
    "formatters": {
        "rq_console": {
            "format": "%(asctime)s %(message)s",
            "datefmt": "%H:%M:%S",
        },
    },
    "handlers": {
        "rq_console": {
            "level": "DEBUG",
            "class": "rq.logutils.ColorizingStreamHandler",
            "formatter": "rq_console",
            "exclude": ["%(asctime)s"],
        },
    },
    'loggers': {
        "rq.worker": {
            "handlers": ["rq_console", "sentry"],
            "level": "DEBUG"
        },
    }
}

Custom Queue Classes

By default, every queue will use DjangoRQ class. If you want to use a custom queue class, you can do so by adding a QUEUE_CLASS option on a per queue basis in RQ_QUEUES:

RQ_QUEUES = {
    'default': {
        'HOST': 'localhost',
        'PORT': 6379,
        'DB': 0,
        'QUEUE_CLASS': 'module.path.CustomClass',
    }
}

or you can specify DjangoRQ to use a custom class for all your queues in RQ settings:

RQ = {
    'QUEUE_CLASS': 'module.path.CustomClass',
}

Custom queue classes should inherit from django_rq.queues.DjangoRQ.

If you are using more than one queue class (not recommended), be sure to only run workers on queues with same queue class. For example if you have two queues defined in RQ_QUEUES and one has custom class specified, you would have to run at least two separate workers for each queue.

Custom Job and Worker Classes

Similarly to custom queue classes, global custom job and worker classes can be configured using JOB_CLASS and WORKER_CLASS settings:

RQ = {
    'JOB_CLASS': 'module.path.CustomJobClass',
    'WORKER_CLASS': 'module.path.CustomWorkerClass',
}

Custom job class should inherit from rq.job.Job. It will be used for all jobs if configured.

Custom worker class should inherit from rq.worker.Worker. It will be used for running all workers unless overridden by rqworker management command worker-class option.

Testing Tip

For an easier testing process, you can run a worker synchronously this way:

from django.test import TestCase
from django_rq import get_worker

class MyTest(TestCase):
    def test_something_that_creates_jobs(self):
        ...                      # Stuff that init jobs.
        get_worker().work(burst=True)  # Processes all jobs then stop.
        ...                      # Asserts that the job stuff is done.

Synchronous Mode

You can set the option ASYNC to False to make synchronous operation the default for a given queue. This will cause jobs to execute immediately and on the same thread as they are dispatched, which is useful for testing and debugging. For example, you might add the following after you queue configuration in your settings file:

# ... Logic to set DEBUG and TESTING settings to True or False ...

# ... Regular RQ_QUEUES setup code ...

if DEBUG or TESTING:
    for queueConfig in RQ_QUEUES.values():
        queueConfig['ASYNC'] = False

Note that setting the is_async parameter explicitly when calling get_queue will override this setting.

Running Tests

To run django_rq's test suite:

`which django-admin` test django_rq --settings=django_rq.tests.settings --pythonpath=.

Deploying on Ubuntu

Create an rqworker service that runs the high, default, and low queues.

sudo vi /etc/systemd/system/rqworker.service

[Unit]
Description=Django-RQ Worker
After=network.target

[Service]
WorkingDirectory=<<path_to_your_project_folder>>
ExecStart=/home/ubuntu/.virtualenv/<<your_virtualenv>>/bin/python \
    <<path_to_your_project_folder>>/manage.py \
    rqworker high default low

[Install]
WantedBy=multi-user.target

Enable and start the service

sudo systemctl enable rqworker
sudo systemctl start rqworker

Deploying on Heroku

Add django-rq to your requirements.txt file with:

pip freeze > requirements.txt

Update your Procfile to:

web: gunicorn --pythonpath="$PWD/your_app_name" config.wsgi:application

worker: python your_app_name/manage.py rqworker high default low

Commit and re-deploy. Then add your new worker with:

heroku scale worker=1

Changelog

See CHANGELOG.md.