Distributed locks with Redis and Python
Clone or download
glasslion Merge pull request #35 from jdufresne/md
PyPI now supports markdown long_description -- use it
Latest commit 7f873cc Jan 21, 2019

README.md

RedLock logo

RedLock - Distributed locks with Redis and Python

Build Status

This library implements the RedLock algorithm introduced by @antirez

Yet another ...

There are already a few redis based lock implementations in the Python world, e.g. retools, redis-lock.

However, these libraries can only work with single-master redis server. When the Redis master goes down, your application has to face a single point of failure. We can't rely on the master-slave replication, because Redis replication is asynchronous.

This is an obvious race condition with the master-slave replication model :

  1. Client A acquires the lock into the master.
  2. The master crashes before the write to the key is transmitted to the slave.
  3. The slave gets promoted to master.
  4. Client B acquires the lock to the same resource A already holds a lock for. SAFETY VIOLATION!

A quick introduction to the RedLock algorithm

To resolve this problem, the Redlock algorithm assume we have N Redis masters. These nodes are totally independent (no replications). In order to acquire the lock, the client will try to acquire the lock in all the N instances sequentially. If and only if the client was able to acquire the lock in the majority ((N+1)/2)of the instances, the lock is considered to be acquired.

The detailed description of the RedLock algorithm can be found in the Redis documentation: Distributed locks with Redis.

APIs

The redlock.RedLock class shares a similar API with the threading.Lock class in the Python Standard Library.

Basic Usage

from redlock import RedLock
# By default, if no redis connection details are
# provided, RedLock uses redis://127.0.0.1:6379/0
lock =  RedLock("distributed_lock")
lock.acquire()
do_something()
lock.release()

With Statement / Context Manager

As with threading.Lock, redlock.RedLock objects are context managers thus support the With Statement. This way is more pythonic and recommended.

from redlock import RedLock
with RedLock("distributed_lock"):
    do_something()

Specify multiple Redis nodes

from redlock import RedLock
with RedLock("distributed_lock",
              connection_details=[
                {'host': 'xxx.xxx.xxx.xxx', 'port': 6379, 'db': 0},
                {'host': 'xxx.xxx.xxx.xxx', 'port': 6379, 'db': 0},
                {'host': 'xxx.xxx.xxx.xxx', 'port': 6379, 'db': 0},
                {'host': 'xxx.xxx.xxx.xxx', 'port': 6379, 'db': 0},
              ]
            ):
    do_something()

The connection_details parameter expects a list of keyword arguments for initializing Redis clients. Other acceptable Redis client arguments can be found on the redis-py doc.

Reuse Redis clients with the RedLockFactory

Usually the connection details of the Redis nodes are fixed. RedLockFactory can help reuse them, create multiple RedLocks but only initialize the clients once.

from redlock import RedLockFactory
factory = RedLockFactory(
    connection_details=[
        {'host': 'xxx.xxx.xxx.xxx'},
        {'host': 'xxx.xxx.xxx.xxx'},
        {'host': 'xxx.xxx.xxx.xxx'},
        {'host': 'xxx.xxx.xxx.xxx'},
    ])

with factory.create_lock("distributed_lock"):
    do_something()

with factory.create_lock("another_lock"):
    do_something()