1. Introduction

django-redis is a BSD Licensed, full featured Redis cache/session backend for Django.

1.1. Why use django-redis?

Because:

  • In active development.

  • Uses native redis-py url notation connection strings.

  • Modular client system (pluggable clients).

  • Master-Slave support in the default client.

  • Complete battery of tests.

  • Used in production in several projects as cache and session storage.

  • Supports infinite timeouts.

  • Python 3 support in same code base.

  • Facilities for raw access to Redis client/connection pool.

  • Highly configurable (can emulate memcached exception behavior, for example).

  • Unix sockets supported by default.

  • Pluggable parsers.

1.2. Supported versions

  • Latest stable version: 3.8.0

  • Previous stable version: 3.7.2

  • Previous stable version: 3.6.2

Note

Versions like 3.6, 3.7, …​ are considered major releases and can contain some backward incompatibilities. For more information is very recommended see the changelog before update.

Versions like 3.7.0, 3.7.1, …​ are considered minor or bug fix releases and are should contain only bug fixes. No new features.

1.3. Requirements

  • Python 2.7, 3.3 or 3.4

  • Django 1.4, 1.5, 1.6 and 1.7

  • redis-py >= 2.10.0

2. User guide

2.1. Installation

The simplest way to use django-redis in your project is to install it with pip:

pip install django-redis

2.2. Configure as cache backend

To start using django-redis, you should change your Django cache settings to something like this:

CACHES = {
    "default": {
        "BACKEND": "django_redis.cache.RedisCache",
        "LOCATION": "redis://127.0.0.1:6379/1",
        "OPTIONS": {
            "CLIENT_CLASS": "django_redis.client.DefaultClient",
        }
    }
}
Note
If you want use Unix sockets, change the LOCATION value to something like "unix:///path/to/socket?db=1".
Note
if you are coming fron django-redis < 3.8.x, you are probably using redis_cache. Since django-redis 3.8.x, redis_cache module is deprecated in favor to django_redis. The redis_cache module will be removed in 3.9.x versions.

2.3. Configure as session backend

Django can by default use any cache backend as session backend and you benefit from that by using django-redis as backend for session storage without installing any additional backends:

SESSION_ENGINE = "django.contrib.sessions.backends.cache"
SESSION_CACHE_ALIAS = "default"

3. Advanced usage

3.1. Pickle version

For almost all values, django-redis uses pickle to serialize objects.

The latest available version of pickle is used by default. If you want set a concrete version, you can do it, using PICKLE_VERSION option:

CACHES = {
    "default": {
        # ...
        "OPTIONS": {
            "PICKLE_VERSION": -1  # Use the latest protocol version
        }
    }
}

3.2. Socket timeout

Socket timeout can be set using SOCKET_TIMEOUT option:

CACHES = {
    "default": {
        # ...
        "OPTIONS": {
            "SOCKET_TIMEOUT": 5,  # in seconds
        }
    }
}

3.3. Compression support

django_redis comes with compression support out of the box, but is deactivated by default. You can activate it setting COMPRESS_MIN_LEN option to any value great than 0.

CACHES = {
    "default": {
        # ...
        "OPTIONS": {
            "COMPRESS_MIN_LEN": 10,
        }
    }
}

zlib is used as default commpression format. You can change it providing two callables, one for compress and an other for uncompress.

Let see an example, of how make it work with lzma compression format:

import lzma

CACHES = {
    "default": {
        # ...
        "OPTIONS": {
            "COMPRESS_MIN_LEN": 10,
            "COMPRESS_COMPRESSOR": lzma.compress,
            "COMPRESS_DECOMPRESSOR": lzma.decompress,
            "COMPRESS_DECOMPRESSOR_ERROR": lzma.LZMAError
        }
    }
}

3.4. Memcached exceptions behavior

In some situations, when Redis is only used for cache, you do not want exceptions when Redis is down. This is default behavior in the memcached backend and it can be emulated in django-redis.

For setup memcached like behaviour (ignore connection exceptions), you should set IGNORE_EXCEPTIONS settings on your cache configuration:

CACHES = {
    "default": {
        # ...
        "OPTIONS": {
            "IGNORE_EXCEPTIONS": True,
        }
    }
}

Also, you can apply the same settings to all configured caches, you can set the global flag in your settings:

DJANGO_REDIS_IGNORE_EXCEPTIONS = True

3.5. Infinite timeout

django-redis comes with infinite timeouts support, since before Django added it in version 1.6.

If used with Django before version 1.6, django-redis interprets a 0 timeout value as infinite. In Django >= 1.6 it also supports None as timeout value

These statements are then equivalent:
cache.set("key", "value", timeout=0)
cache.set("key", "value", timeout=None)

3.6. Get ttl (time-to-live) from key

With redis, you can access to ttl of any stored key, for it, django-redis exposes ttl function.

It returns:

  • ttl value for any volatile key (any key that has expiration)

  • 0 for expired and not existent keys

  • None for keys that does not have expiration

Simple search keys by pattern
>>> from django.core.cache import cache
>>> cache.set("foo", "value", timeout=25)
>>> cache.ttl("foo")
25
>>> cache.ttl("not-existent")
0

3.7. Scan & Delete keys in bulk

django-redis comes with some additional methods that help with searching or deleting keys using glob patterns.

Simple search keys by pattern
>>> from django.core.cache import cache
>>> cache.keys("foo_*")
["foo_1", "foo_2"]

A simple search like this will return all matched values. In databases with a large number of keys this isn’t suitable method. Instead, you can use the iter_keys function that works like the keys function but uses redis>=2.8 server side cursors. Calling iter_keys will return a generator that you can then iterate over efficiently.

Search using server side cursors
>>> from django.core.cache import cache
>>> cache.iter_keys("foo_*")
<generator object algo at 0x7ffa9c2713a8>
>>> next(cache.iter_keys("foo_*"))
"foo_1"

For deleting keys, you should use delete_pattern which has the same glob pattern syntax as the keys function and returns the number of deleted keys.

Example use of delete_pattern
>>> from django.core.cache import cache
>>> cache.delete_pattern("foo_*")

3.8. Redis native commands

django-redis has limited support for some Redis atomic operations, such as the commands SETNX and INCR.

You can use the SETNX command through the backend set() method with the nx parameter:

Example:
>>> from django.core.cache import cache
>>> cache.set("key", "value1", nx=True)
True
>>> cache.set("key", "value2", nx=True)
False
>>> cache.get("key")
"value1"

Also, incr and decr methods uses redis atomic operations when value that contains a key is suitable for it.

3.9. Raw client access

In some situations your application requires access to a raw Redis client to use some advanced features that aren’t exposed by the Django cache interface. To avoid storing another setting for creating a raw connection, django-redis exposes functions with which you can obtain a raw client reusing the cache connection string: get_redis_connection(alias).

>>> from django_redis import get_redis_connection
>>> con = get_redis_connection("default")
>>> con
<redis.client.StrictRedis object at 0x2dc4510>
Warning
Not all pluggable clients support this feature.

3.10. Connection pools

Behind the scenes, django-redis uses the underlying redis-py connection pool implementation, and exposes a simple way to configure it. Alternatively, you can directly customize a connection/connection pool creation for a backend.

The default redis-py behavior is to not close connections, recycling them when possible.

3.10.1. Configure default connection pool

The default connection pool is simple. You can only customize the maximum number of connections in the pool, by setting CONNECTION_POOL_KWARGS in the CACHES setting:

CACHES = {
    "default": {
        "BACKEND": "django_redis.cache.RedisCache",
        ...
        "OPTIONS": {
            "CONNECTION_POOL_KWARGS": {"max_connections": 100}
        }
    }
}

You can verify how many connections the pool has opened with the following snippet:

from django.core.cache import get_cache
django_redis = get_cache("default")  # Use the name you have defined for Redis in settings.CACHES
redis = django_redis.raw_client
connection_pool = redis.connection_pool
print("Created connections so far: %d" % connection_pool._created_connections)

3.10.2. Use your own connection pool subclass

Sometimes you want to use your own subclass of the connection pool. This is possible with django-redis using the CONNECTION_POOL_CLASS parameter in the backend options.

myproj/mypool.py
from redis.connection import ConnectionPool

class MyOwnPool(ConnectionPool):
    # Just doing nothing, only for example purpose
    pass
settings.py
# Omitting all backend declaration boilerplate code.

"OPTIONS": {
    "CONNECTION_POOL_CLASS": "myproj.mypool.MyOwnPool",
}

3.10.3. Customize connection factory

If none of the previous methods satisfies you, you can get in the middle of the django-redis connection factory process and customize or completely rewrite it.

By default, django-redis creates connections through the django_redis.pool.ConnectionFactory class that is specified in the global Django setting DJANGO_REDIS_CONNECTION_FACTORY.

Partial interface of ConnectionFactory class
# Note: Using Python 3 notation for code documentation ;)

class ConnectionFactory(object):
    def get_connection_pool(self, params:dict):
        # Given connection parameters in the `params` argument,
        # return new connection pool.
        # It should be overwritten if you want do something
        # before/after creating the connection pool, or return your
        # own connection pool.
        pass

    def get_connection(self, params:dict):
        # Given connection parameters in the `params` argument,
        # return a new connection.
        # It should be overwritten if you want to do something
        # before/after creating a new connection.
        # The default implementation uses `get_connection_pool`
        # to obtain a pool and create a new connection in the
        # newly obtained pool.
        pass

    def get_or_create_connection_pool(self, params:dict):
        # This is a high layer on top of `get_connection_pool` for
        # implementing a cache of created connection pools.
        # It should be overwritten if you want change the default
        # behavior.
        pass

    def make_connection_params(self, url:str) -> dict:
        # The responsibility of this method is to convert basic connection
        # parameters and other settings to fully connection pool ready
        # connection parameters.
        pass

    def connect(self, url:str):
        # This is really a public API and entry point for this
        # factory class. This encapsulates the main logic of creating
        # the previously mentioned `params` using `make_connection_params`
        # and creating a new connection using the `get_connection` method.
        pass

3.11. Pluggable parsers

redis-py (the Python Redis client used by django-redis) comes with a pure Python Redis parser that works very well for most common task, but if you want some performance boost, you can use hiredis.

hiredis is a Redis client written in C and it has its own parser that can be used with django-redis.

"OPTIONS": {
    "PARSER_CLASS": "redis.connection.HiredisParser",
}

3.12. Pluggable clients

django_redis is designed for to be very flexible and very configurable. For it, it exposes a pluggable backends that make easy extend the default behavior, and it comes with few ones out the box.

3.12.1. Default client

Almost all about the default client is explained, with one exception: the default client comes with master-slave support.

To connect to master-slave redis setup, you should change the LOCATION to something like this:

"LOCATION": [
    "redis://127.0.0.1:6379/1",
    "redis://127.0.0.1:6378/1",
]

The first connection string represents a master server and the rest to slave servers.

Warning
Master-Slave setup is not heavily tested in production environments.

3.12.2. Shard client

This pluggable client implements client-side sharding. It inherits almost all functionality from the default client. To use it, change your cache settings to something like this:

CACHES = {
    "default": {
        "BACKEND": "django_redis.cache.RedisCache",
        "LOCATION": [
            "redis://127.0.0.1:6379/1",
            "redis://127.0.0.1:6379/2",
        ],
        "OPTIONS": {
            "CLIENT_CLASS": "django_redis.client.ShardClient",
        }
    }
}
Warning
Shard client is still experimental, so be careful when using it in production environments.

3.12.3. Herd client

This pluggable client helps dealing with the thundering herd problem. You can read more about it on Wikipedia.

Like previus pluggable clients, it inherits all functionality from the default client, adding some additional methods for getting/setting keys.

Example setup
 CACHES = {
    "default": {
        "BACKEND": "django_redis.cache.RedisCache",
        "LOCATION": "redis://127.0.0.1:6379/1",
        "OPTIONS": {
            "CLIENT_CLASS": "django_redis.client.HerdClient",
        }
    }
}

This client exposes additional settings:

  • CACHE_HERD_TIMEOUT: Set default herd timeout. (Default value: 60s)

4. License

Copyright (c) 2011-2014 Andrey Antukh <niwi@niwi.be>
Copyright (c) 2011 Sean Bleier

All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
   notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
   notice, this list of conditions and the following disclaimer in the
   documentation and/or other materials provided with the distribution.
3. The name of the author may not be used to endorse or promote products
   derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.