redis-shard

Redis Shard ###########

.. image:: https://img.shields.io/travis/zhihu/redis-shard.svg?style=flat :target: https://travis-ci.org/zhihu/redis-shard :alt: Build Status

.. image:: https://img.shields.io/pypi/v/redis-shard.svg?style=flat :target: https://pypi.python.org/pypi/redis-shard :alt: Latest Version

.. image:: https://img.shields.io/pypi/pyversions/redis-shard.svg?style=flat :target: https://pypi.python.org/pypi/redis-shard :alt: Supported Python versions

.. image:: https://img.shields.io/pypi/l/redis-shard.svg?style=flat :target: https://pypi.python.org/pypi/redis-shard :alt: License

A Redis sharding implementation.

Redis is great. It's fast, lightweight and easy to use. But when we want to store a mass of data into one single instance, we may encounter some problems such as performance degradation and slow recovery and we need to scale it.

Usage

First, Create an RedisShardAPI instance with multiple nodes, node name must be unique::

from redis_shard.shard import RedisShardAPI

servers = [
    {'name': 'server1', 'host': '127.0.0.1', 'port': 10000, 'db': 0},
    {'name': 'server2', 'host': '127.0.0.1', 'port': 11000, 'db': 0},
    {'name': 'server3', 'host': '127.0.0.1', 'port': 12000, 'db': 0},
]

client = RedisShardAPI(servers, hash_method='md5')

Then, you can access the Redis cluster as you use redis-py <https://github.com/andymccurdy/redis-py>_::

client.set('test', 1)
client.get('test')  # 1
client.zadd('testset', 'first', 1)
client.zadd('testset', 'second', 2)
client.zrange('testset', 0, -1)  # [first, second]

Hash Tags

If you want to store specific keys on one node for some reason (such as you prefer single instance pipeline, or you want to use multi-keys command such as sinter), you should use Hash Tags::

client.set('foo', 2)
client.set('a{foo}', 5)
client.set('b{foo}', 5)
client.set('{foo}d', 5)
client.set('d{foo}e', 5)

client.get_server_name('foo') == client.get_server_name('a{foo}') == client.get_server_name('{foo}d') \
    == client.get_server_name('d{foo}e')  # True

The string in a braces of a key is the Hash Tag of the key. The hash of a Hash Tag will be treated the hash of the key. So, keys foo, bar{foo} and b{foo}ar will be sotred in the same node.

.. note:: Hash Tags are not supported with bytes keys in Python 3.

Tag method

Just add ``tag_`` prefix, you can use any of the normal redis method on the same hash tag::

    client.tag_mget("{user:1}question1", "{user:1}question2")
    client.tag_delete("{user:1}question1", "{user:1}question2")


Multi-keys method

Only support mget, mset and flushdb.

Config Details

There are three parameters servers, hash_method and sentinel in the class redis_shard.shard.RedisShardAPI.

servers is a list. Each element in it should be a dict or a URL schema.

• dict::

  [ {'name': 'server1', 'host': '127.0.0.1', 'port': 10000, 'db': 0}, {'name': 'server2', 'host': '127.0.0.1', 'port': 11000, 'db': 0, 'max_connections': 50}, {'name': 'server3', 'host': '127.0.0.1', 'port': 12000, 'db': 0}, ]

• URL schema::

  [ 'redis://127.0.0.1:10000/0?name=node1', 'redis://127.0.0.1:11000/0?name=node2&max_connections=50', 'redis://127.0.0.1:12000/0?name=node3' ]

If the following parameter sentinel is enabled, only name is needed for the servers config.

hash_method is a string which indicate the method of generating the hash key of the consistent hash ring. The default value is crc32. It also supports md5 and sha1.

sentinel is the config for Redis Sentinel <http://redis.io/topics/sentinel>_. With the sentinel support, redis-shard will do read/write splitting. Config is like this::

{"hosts": [('localhost', 26379)], "socket_timeout": 0.1}

Limitations

• Redis Shard dose not support all Redis commands.
• As mentioned above, Redis Shard does not support all multi-keys commands crossing different nodes, you have to use Hash Tag to work with those commands.
• Redis Shard does not have any replication mechanism.

How it Works

Redis Shard is basically inspired by this article <http://oldblog.antirez.com/post/redis-presharding.html>_.

Change History

0.3.7 (2021-02-14)

• Fix hash tags

0.3.6 (2020-10-15)

• Made Redis URL parsing smarter (@huntrax11 )
• Compact with python 3.8
• Compact with redis-py 3.0.0+

0.3.5 (2017-10-25)

• Allow for bytes based keys in Python3 (@gmr)
• Fix crc-32 calculation per Python

0.3.4 (2016-07-19)

• fix md5 hash_method occus NameError on python3

0.3.3 (2016-03-18)

• max_connections can now be set when instantiating client instance, and has a default value 100.

0.3.2 (2015-06-02)

• add strict redis support

0.3.1 (2015-02-02)

• transaction support

0.3.0 (2014-11-20)

• shard key generate function now support md5 and sha1 .
• add Redis Sentinel support

0.2.4 (2014-06-26)

• remove gevent dependency

0.2.3 (2014-06-02)

• better pipeline support

0.2.1 (2013-08-07)

• add evel method

0.2.0 (2013-06-02)

• add python3 support

0.1.11 (2013-05-06)

• add mset support

0.1.10 (2013-04-13)

• add mget support ,thks to @Yuekui

0.1.9 (2013-03-04)

• add an reshard example
• tidy the pipeline code
• add more shard methods

0.1.8 (2013-01-21)

• add append and getrange method, thks @simon-liu

0.1.7 (2012-11-19)

• use new redis url config instead of dict

0.1.5 (2012-07-16)

• Add many new methods, support socket_timeout and password

0.1.4 (2011-07-20)

• modify hash key algor, support suffix match, thks to dkong
• support more redis methods, include keys.

0.1.3 (2011-06-20)

• support 2.4.X version of redis-py

0.1.2 (2011-06-01)

• add MANIFEST.in file

0.1.1 (2011-05-29)

• create hashring use server's name config.

0.1 (2011-05-28)

• first version