Distrubuted mutex using Redis
Pull request Compare This branch is 45 commits behind kenn:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Redis Mutex

Build Status

Distrubuted mutex in Ruby using Redis. Supports both blocking and non-blocking semantics.

The idea was taken from the official SETNX doc.


In the following example, only one thread / process / server can enter the locked block at one time.

  # do something exclusively


mutex = Redis::Mutex.new(:your_lock_name)
if mutex.lock
  # do something exclusively
  puts "failed to obtain lock!"

By default, while one is holding a lock, others wait 1 second in total, polling every 100ms to see if the lock was released. When 1 second has passed, the lock method returns false and others give up. Note that if your job runs longer than 10 seconds, the lock will be automatically removed to avoid a deadlock situation in case your job is dead before releasing the lock. Also note that you can configure any of these timing values, as explained later.

Or if you want to immediately receive false on an unsuccessful locking attempt, you can change the mutex mode to non-blocking.


gem install redis-mutex


In Gemfile:

gem "redis-mutex"

Register the Redis server: (e.g. in config/initializers/redis_mutex.rb for Rails)

Redis::Classy.db = Redis.new(:host => 'localhost')

Note that Redis Mutex uses the redis-classy gem internally to organize keys in an isolated namespace.

There are four methods - new, lock, unlock and sweep:

mutex = Redis::Mutex.new(key, options)    # Configure a mutex lock
mutex.lock                                # Try to obtain the lock
mutex.unlock                              # Release the lock if it's not expired
Redis::Mutex.sweep                        # Forcibly remove all locks

Redis::Mutex.lock(key, options)           # Shortcut to new + lock

The key argument can be symbol, string, or any Ruby objects that respond to id method, where the key is automatically set as TheClass:id. For any given key, Redis::Mutex: prefix will be automatically prepended. For instance, if you pass a Room object with id of 123, the actual key in Redis will be Redis::Mutex:Room:123. The automatic prefixing and instance binding is the feature of Redis::Classy - for more internal details, refer to Redis Classy.

The initialize method takes several options.

:block  => 1    # Specify in seconds how long you want to wait for the lock to be released.
                # Speficy 0 if you need non-blocking sematics and return false immediately. (default: 1)
:sleep  => 0.1  # Specify in seconds how long the polling interval should be when :block is given.
                # It is recommended that you do NOT go below 0.01. (default: 0.1)
:expire => 10   # Specify in seconds when the lock should forcibly be removed when something went wrong
                # with the one who held the lock. (default: 10)

The lock method returns true when the lock has been successfully obtained, or returns false when the attempts failed after the seconds specified with :block. When 0 is given to :block, it is set to non-blocking mode and immediately returns false.

In the following Rails example, only one request can enter to a given room.

class RoomController < ApplicationController
  before_filter { @room = Room.find(params[:id]) }
  def enter
    success = Redis::Mutex.lock(@room) do    # key => "Room:123"
      # do something exclusively
    render :text => success ? 'success!' : 'failed to obtain lock!'

Note that you need to explicitly call the unlock method when you don't use the block syntax, and it is recommended to put the unlock method in the ensure clause unless you're sure your code won't raise any exception.

def enter
  mutex = Redis::Mutex.new('non-blocking', :block => 0, :expire => 10.minutes)
  if mutex.lock
      # do something exclusively
    render :text => 'success!'
    render :text => 'failed to obtain lock!'

Macro-style definition

If you want to wrap an entire method into a critical section, you can use the macro-style definition. The locking scope will be TheClass#method and only one method can run at any given time.

If you give a proc object to the after_failure option, it will get called after locking attempt failed.

class JobController < ApplicationController
  include Redis::Mutex::Macro
  auto_mutex :run, :block => 0, :after_failure => lambda { render :text => "failed to obtain lock!" }
  def run
    # do something exclusively
    render :text => "success!"