rediscluster-py

a Python interface to a Cluster(*) of Redis key-value stores.

NOTE: This client is not (yet) compatible with the Redis Cluster Spec found here: http://redis.io/topics/cluster-spec

Project Goals

The goal of rediscluster-py, together with rediscluster-php, is to have a consistent, compatible client libraries accross programming languages when sharding among different Redis instances in a transparent, fast, and fault tolerant way. rediscluster-py is based on the awesome redis-py StrictRedis Api, thus the original api commands would work without problems within the context of a cluster of redis servers

Continuous Integration

Currently, rediscluster-py is being tested via travis/drone.io ci for python version 2.6, 2.7 and 3.2:

Installation

$ sudo pip install rediscluster

or alternatively (you really should be using pip though):

$ sudo easy_install rediscluster

From source:

$ sudo python setup.py install

Running Tests

$ git clone https://github.com/salimane/rediscluster-py.git
$ cd rediscluster-py
$ vi tests/config.py
$ ./run_tests

Getting Started

>>> import rediscluster
>>> cluster = {
...          # node names
...          'nodes' : { # masters
...                      'node_1' : {'host' : '127.0.0.1', 'port' : 63791},
...                      'node_2' : {'host' : '127.0.0.1', 'port' : 63792},
...                    }
...     }
>>> r = rediscluster.StrictRedisCluster(cluster=cluster, db=0)
>>> r.set('foo', 'bar')
True
>>> r.get('foo')
'bar'

Cluster Configuration

The cluster configuration is a hash that is mostly based on the idea of a node, which is simply a host:port pair that points to a single redis-server instance. This is to make sure it doesn’t get tied it to a specific host (or port). The advantage of this is that it is easy to add or remove nodes from the system to adjust the capacity while the system is running.

Read Slaves & Write Masters

rediscluster uses the master servers stored in the cluster hash passed during instantiation to auto discover if any slave is attached to them. It then transparently relay read redis commands to slaves and writes commands to masters.

There is also support to only use masters even if read redis commands are issued, just specify it at client instantiation like :

>>> r = rediscluster.StrictRedisCluster(cluster=cluster, db=0) # read redis commands are routed to slaves
>>>
>>> r = rediscluster.StrictRedisCluster(cluster=cluster, db=0, mastersonly=True) # read redis commands are routed to masters

Partitioning Algorithm

rediscluster doesn't used a consistent hashing like some other libraries. In order to map every given key to the appropriate Redis node, the algorithm used, based on crc32 and modulo, is :

(abs(binascii.crc32(<key>) & 0xffffffff) % <number of masters>) + 1

this is used to ensure some compatibility with other languages, php in particular. A function getnodefor is provided to get the node a particular key will be/has been stored to.

>>> r.getnodefor('foo')
{'node_2': {'host': '127.0.0.1', 'port': 63792}}
>>>

Hash Tags

In order to specify your own hash key (so that related keys can all land on a given node), rediscluster allows you to pass a string in the form "a{b}" where you’d normally pass a scalar. The first element of the list is the key to use for the hash and the second is the real key that should be fetched/modify:

>>> r.get("bar{foo}")
>>>
>>> r.mset({"bar{foo}": "bar", "foo": "foo"})
>>>
>>> r.mget(["bar{foo}", "foo"])

In that case “foo” is the hash key but “bar” is still the name of the key that is fetched from the redis node that “foo” hashes to.

Multiple Keys Redis Commands

In the context of storing an application data accross many redis servers, commands taking multiple keys as arguments are harder to use since, if the two keys will hash to two different instances, the operation can not be performed. Fortunately, rediscluster is a little fault tolerant in that it still fetches the right result for those multi keys operations as far as the client is concerned. To do so it processes the related involved redis servers at interface level.

>>> r.sadd('foo', *['a1', 'a2', 'a3'])
3
>>> r.sadd('bar', *['b1', 'a2', 'b3'])
3
>>> r.sdiffstore('foobar', 'foo', 'bar')
2
>>> r.smembers('foobar')
set(['a1', 'a3'])
>>> r.getnodefor('foo')
{'node_2': {'host': '127.0.0.1', 'port': 63792}}
>>> r.getnodefor('bar')
{'node_1': {'host': '127.0.0.1', 'port': 63791}}
>>> r.getnodefor('foobar')
{'node_2': {'host': '127.0.0.1', 'port': 63792}}
>>>

Redis-Sharding & Redis-Copy

In order to help with moving an application with a single redis server to a cluster of redis servers that could take advantage of rediscluster, i wrote redis-sharding and redis-copy

Information

• Code: git clone git://github.com/salimane/rediscluster-py.git
• Home: http://github.com/salimane/rediscluster-py
• Bugs: http://github.com/salimane/rediscluster-py/issues

Author

rediscluster-py is developed and maintained by Salimane Adjao Moustapha (me@salimane.com). It can be found here: http://github.com/salimane/rediscluster-py