Redlock - A ruby distributed lock using redis.

Distributed locks are a very useful primitive in many environments where different processes require to operate with shared resources in a mutually exclusive way.

There are a number of libraries and blog posts describing how to implement a DLM (Distributed Lock Manager) with Redis, but every library uses a different approach, and many use a simple approach with lower guarantees compared to what can be achieved with slightly more complex designs.

This is an implementation of a proposed distributed lock algorithm with Redis. It started as a fork from antirez implementation.

Compatibility

Redlock works with Redis versions 2.6 or later.

Installation

Add this line to your application's Gemfile:

gem 'redlock'

And then execute:

$ bundle

Or install it yourself as:

$ gem install redlock

Documentation

RubyDoc

Usage example

Acquiring a lock

NOTE: All expiration durations are in milliseconds.

  # Locking
  lock_manager = Redlock::Client.new([ "redis://127.0.0.1:7777", "redis://127.0.0.1:7778", "redis://127.0.0.1:7779" ])
  first_try_lock_info = lock_manager.lock("resource_key", 2000)
  second_try_lock_info = lock_manager.lock("resource_key", 2000)

  p first_try_lock_info
  # => {validity: 1987, resource: "resource_key", value: "generated_uuid4"}

  p second_try_lock_info
  # => false

  # Unlocking
  lock_manager.unlock(first_try_lock_info)

  second_try_lock_info = lock_manager.lock("resource_key", 2000)

  p second_try_lock_info
  # => {validity: 1962, resource: "resource_key", value: "generated_uuid5"}

There's also a block version that automatically unlocks the lock:

lock_manager.lock("resource_key", 2000) do |locked|
  if locked
    # critical code
  else
    # error handling
  end
end

There's also a bang version that only executes the block if the lock is successfully acquired, returning the block's value as a result, or raising an exception otherwise. Passing a block is mandatory.

begin
  block_result = lock_manager.lock!("resource_key", 2000) do
    # critical code
  end
rescue Redlock::LockError
  # error handling
end

Extending a lock

To extend the life of the lock:

begin
  lock_info = lock_manager.lock("resource_key", 2000)
  while lock_info
    # Critical code

    # Time up and more work to do? Extend the lock.
    lock_info = lock_manager.lock("resource key", 3000, extend: lock_info)
  end
rescue Redlock::LockError
  # error handling
end

The above code will also acquire the lock if the previous lock has expired and the lock is currently free. Keep in mind that this means the lock could have been acquired and released by someone else in the meantime. To only extend the life of the lock if currently locked by yourself, use the extend_only_if_locked parameter:

lock_manager.lock("resource key", 3000, extend: lock_info, extend_only_if_locked: true)

Querying lock status

You can check if a resource is locked:

resource = "resource_key"
lock_info = lock_manager.lock(resource, 2000)
lock_manager.locked?(resource)
#=> true

lock_manager.unlock(lock_info)
lock_manager.locked?(resource)
#=> false

Any caller can call the above method to query the status. If you hold a lock and would like to check if it is valid, you can use the valid_lock? method:

lock_info = lock_manager.lock("resource_key", 2000)
lock_manager.valid_lock?(lock_info)
#=> true

lock_manager.unlock(lock_info)
lock_manager.valid_lock?(lock_info)
#=> false

The above methods are not safe if you are using this to time critical code, since they return true if the lock has not expired, even if there's only (for example) 1ms left on the lock. If you want to safely time the lock validity, you can use the get_remaining_ttl_for_lock and get_remaining_ttl_for_resource methods.

Use get_remaining_ttl_for_lock if you hold a lock and want to check the TTL specifically for your lock:

resource = "resource_key"
lock_info = lock_manager.lock(resource, 2000)
sleep 1

lock_manager.get_remaining_ttl_for_lock(lock_info)
#=> 986

lock_manager.unlock(lock_info)
lock_manager.get_remaining_ttl_for_lock(lock_info)
#=> nil 

Use get_remaining_ttl_for_resource if you do not hold a lock, but want to know the remaining TTL on a locked resource:

# Some part of the code
resource = "resource_key"
lock_info = lock_manager.lock(resource, 2000)

# Some other part of the code
lock_manager.locked?(resource)
#=> true
lock_manager.get_remaining_ttl_for_resource(resource)
#=> 1975 

# Sometime later
lock_manager.locked?(resource)
#=> false
lock_manager.get_remaining_ttl_for_resource(resource)
#=> nil 

Redis client configuration

Redlock::Client expects URLs or Redis objects on initialization. Redis objects should be used for configuring the connection in more detail, i.e. setting username and password.

servers = [ 'redis://localhost:6379', Redis.new(:url => 'redis://someotherhost:6379') ]
redlock = Redlock::Client.new(servers)

Redlock works seamlessly with redis sentinel, which is supported in redis 3.2+.

Redlock configuration

It's possible to customize the retry logic providing the following options:

  lock_manager = Redlock::Client.new(
                  servers, {
                  retry_count:   3,
                  retry_delay:   200, # milliseconds
                  retry_jitter:  50,  # milliseconds
                  redis_timeout: 0.1  # seconds
                 })

It is possible to associate :retry_delay option with Proc object. It will be called every time, with attempt number as argument, to get delay time value before next retry.

retry_delay = proc { |attempt_number| 200 * attempt_number ** 2 } # delay of 200ms for 1st retry, 800ms for 2nd retry, etc.
lock_manager = Redlock::Client.new(servers, retry_delay: retry_delay)

For more information you can check documentation.

Run tests

Make sure you have docker installed.

$ make

Disclaimer

This code implements an algorithm which is currently a proposal, it was not formally analyzed. Make sure to understand how it works before using it in your production environments. You can see discussion about this approach at reddit and also the Antirez answers for some critics.

Contributing

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request