yarnl

Yet Another Redis NodeJS Lock

A NodeJS library for single instance Redis v2.6.12+ locking.

For distributed locking, see http://redis.io/topics/distlock.

About

Implements the locking algorithm specified in http://redis.io/commands/set. Locking is implemented with a set-if-not-exists plus a TTL to cull stale locks:

SET resource-name random-token NX PX max-lock-time

The random token is important to prevent clients from erroneously believing they hold a lock. yarnl uses UUIDs (v1), via node-uuid, for its random tokens. node-uuid uses an internal counter to emulate UUIDv1's 100ns precision, so there is little risk of lock collisions even at high use. The UUID generator is also seeded with 128 bytes of cryptographically strong randomness so that separate processes will not collide.

Unlocking is implemented with a Lua script:

if redis.call("get", KEYS[1]) == ARGV[1] then
    return redis.call("del", KEYS[1])
else
    return 0
end

Requirements

• Redis v2.6.12+ for SET NX PX, but only tested on v2.8
• node-redis v0.10+

Install

npm install yarnl

Usage

Minimum example

var Lock = require('yarnl');

new Lock('lockName').lock(function (err, unlock) {
  if (err) { /* handle error... */ }

  if (!unlock) {
    // we didn't get the lock
  } else {
    // do stuff with lock...

    unlock(function (err) {
      if (err) {
        // error releasing lock, not a big deal since we set TTLs
      }
    });
  }
});

Custom backoff strategies

You can create custom backoff strategies by passing a function to retryDelay. This example shows how you might create an exponential backoff with delays [10, 20, 40, ...].

new Lock('expontentialBackoffLock').lock({
  retryDelay: (function () {
    var delay = 5;
    return function () {
      return delay * 2;
    };
  })()
}, function (err, unlock) {
  // etc.
});

Comparison to other libraries

There are a lot of existing Redis locking libraries for NodeJS. Unfortunately, a lot of the existing libraries are missing useful features, like automatic retry support. yarnl also tries to implement the current best practice Redis locking algorithm which has changed over the years as Redis has matured.

API

new Lock(name [, options])

Instantiates a new lock with the given name and optional options. options is an object with the following possible properties:

• ttl: How long until the lock is automatically deleted by Redis. Defaults to 5 seconds. Set this higher if you plan on holding the lock for longer.
• maxAttempts: Number of times to attempt to acquire the lock. For example, maxAttempts: 1 will only attempt to acquire the lock once, failing if it is unavailable. Defaults to null, which will keep trying forever.
• retryDelay: Milliseconds to wait between attempts. Has no effect if maxAttempts equals 1. Defaults to 50ms. Can also be a function (see custom backoff strategies).
• client: node-redis client instance to use. Defaults to require('redis').createClient().

.lock([options, ] callback)

Acquire the lock. If options.maxAttempts isn't set (which is the default), the callback will not be called until the lock is acquired, or an error occurs.

callback is called with a (potentially null) error and an unlock function. If we failed to acquire the lock after maxAttempts attempts, unlock will be null.

This method also takes an optional options argument which is an object. Options passed to .lock() will override options passed to the Lock constructor. Besides the constructor options (ttl, maxAttempts, etc.), .lock() accepts the following additional options:

• value: Value to save as the contents of the lock's Redis key. This should be a random token to prevent accidental deletion by clients that erroneously believe they hold the lock. Defaults to a UUIDv1.

.tryLock(callback)

Equivalent to lock({maxAttempts: 1})

.unlock(callback)

The same unlock function that's passed to lock(). Useful if you want to unlock outside the scope of the lock function. Since code passed to the lock function might still be using the lock, this method should be used with caution.