Resque

Resque (pronounced like "rescue") is a Redis-backed library for creating background jobs, placing those jobs on multiple queues, and processing them later.

For the backstory, philosophy, and history of Resque's beginnings, please see the blog post (2009).

Background jobs can be any Ruby class or module that responds to perform. Your existing classes can easily be converted to background jobs or you can create new classes specifically to do work. Or, you can do both.

Resque is heavily inspired by DelayedJob (which rocks) and comprises three parts:

1. A Ruby library for creating, querying, and processing jobs
2. A Rake task for starting a worker which processes jobs
3. A Sinatra app for monitoring queues, jobs, and workers.

Resque workers can be given multiple queues (a "queue list"), distributed between multiple machines, run anywhere with network access to the Redis server, support priorities, are resilient to memory bloat / "leaks," tell you what they're doing, and expect failure.

Resque queues are persistent; support constant time, atomic push and pop (thanks to Redis); provide visibility into their contents; and store jobs as simple JSON packages.

The Resque frontend tells you what workers are doing, what workers are not doing, what queues you're using, what's in those queues, provides general usage stats, and helps you track failures.

Resque now supports Ruby 2.3.0 and above. We will also only be supporting Redis 3.0 and above going forward.

Would you like to be involved in Resque? Do you have thoughts about what Resque should be and do going forward? There's currently an open discussion here on just that topic, so please feel free to join in. We'd love to hear your thoughts and/or have people volunteer to be a part of the project!

Resque jobs are Ruby classes (or modules) which respond to the perform method. Here's an example:

class Archive
  @queue = :file_serve

  def self.perform(repo_id, branch = 'master')
    repo = Repository.find(repo_id)
    repo.create_archive(branch)
  end
end

The @queue class instance variable determines which queue Archive jobs will be placed in. Queues are arbitrary and created on the fly - you can name them whatever you want and have as many as you want.

To place an Archive job on the file_serve queue, we might add this to our application's pre-existing Repository class:

class Repository
  def async_create_archive(branch)
    Resque.enqueue(Archive, self.id, branch)
  end
end

Now when we call repo.async_create_archive('masterbrew') in our application, a job will be created and placed on the file_serve queue.

Later, a worker will run something like this code to process the job:

klass, args = Resque.reserve(:file_serve)
klass.perform(*args) if klass.respond_to? :perform

Which translates to:

Archive.perform(44, 'masterbrew')

Let's start a worker to run file_serve jobs:

$ cd app_root
$ QUEUE=file_serve rake resque:work

This starts one Resque worker and tells it to work off the file_serve queue. As soon as it's ready it'll try to run the Resque.reserve code snippet above and process jobs until it can't find any more, at which point it will sleep for a small period and repeatedly poll the queue for more jobs.

Add the gem to your Gemfile:

gem 'resque'

Next, install it with Bundler:

$ bundle

In your Rakefile, or some other file in lib/tasks (ex: lib/tasks/resque.rake), load the resque rake tasks:

require 'resque'
require 'resque/tasks'
require 'your/app' # Include this line if you want your workers to have access to your application

If you're using Rails 7.1.x, use rack-session version 1.0.2. This is because resque-web's dependency, Sinatra, isn't compatible with rack 3.0, which is required by rack-session 2.0.0.

  gem 'rails', '~> 7.1'
  gem 'rack-session', '~> 1.0', '>= 1.0.2'

To make resque specific changes, you can override the resque:setup job in lib/tasks (ex: lib/tasks/resque.rake). GitHub's setup task looks like this:

task "resque:setup" => :environment do
  Grit::Git.git_timeout = 10.minutes
end

We don't want the git_timeout as high as 10 minutes in our web app, but in the Resque workers it's fine.

Resque workers are rake tasks that run forever. They basically do this:

start
loop do
  if job = reserve
    job.process
  else
    sleep 5 # Polling frequency = 5
  end
end
shutdown

Starting a worker is simple:

$ QUEUE=* rake resque:work

Or, you can start multiple workers:

$ COUNT=2 QUEUE=* rake resque:workers

This will spawn two Resque workers, each in its own process. Hitting ctrl-c should be sufficient to stop them all.

Resque doesn't support numeric priorities but instead uses the order of queues you give it. We call this list of queues the "queue list."

Let's say we add a warm_cache queue in addition to our file_serve queue. We'd now start a worker like so:

$ QUEUES=file_serve,warm_cache rake resque:work

When the worker looks for new jobs, it will first check file_serve. If it finds a job, it'll process it then check file_serve again. It will keep checking file_serve until no more jobs are available. At that point, it will check warm_cache. If it finds a job it'll process it then check file_serve (repeating the whole process).

In this way you can prioritize certain queues. At GitHub we start our workers with something like this:

$ QUEUES=critical,archive,high,low rake resque:work

Notice the archive queue - it is specialized and in our future architecture will only be run from a single machine.

At that point we'll start workers on our generalized background machines with this command:

$ QUEUES=critical,high,low rake resque:work

And workers on our specialized archive machine with this command:

$ QUEUE=archive rake resque:work

If you want your workers to work off of every queue, including new queues created on the fly, you can use a splat:

$ QUEUE=* rake resque:work

Queues will be processed in alphabetical order.

Or, prioritize some queues above *:

# QUEUE=critical,* rake resque:work

If you want your workers to work off of all queues except for some, you can use negation:

$ QUEUE=*,!low rake resque:work

Negated globs also work. The following will instruct workers to work off of all queues except those beginning with file_:

$ QUEUE=*,!file_* rake resque:work

Note that the order in which negated queues are specified does not matter, so QUEUE=*,!file_* and QUEUE=!file_*,* will have the same effect.

There are scenarios where it's helpful to record the PID of a resque worker process. Use the PIDFILE option for easy access to the PID:

$ PIDFILE=./resque.pid QUEUE=file_serve rake resque:work

There are scenarios where it's helpful for the resque worker to run itself in the background (usually in combination with PIDFILE). Use the BACKGROUND option so that rake will return as soon as the worker is started.

$ PIDFILE=./resque.pid BACKGROUND=yes QUEUE=file_serve rake resque:work

You can pass an INTERVAL option which is a float representing the polling frequency. The default is 5 seconds, but for a semi-active app you may want to use a smaller value.

$ INTERVAL=0.1 QUEUE=file_serve rake resque:work

When INTERVAL is set to 0 it will run until the queue is empty and then shutdown the worker, instead of waiting for new jobs.

Resque comes with a Sinatra-based front end for seeing what's up with your queue.

If you've installed Resque as a gem running the front end standalone is easy:

$ resque-web

It's a thin layer around rackup so it's configurable as well:

$ resque-web -p 8282

If you have a Resque config file you want evaluated just pass it to the script as the final argument:

$ resque-web -p 8282 rails_root/config/initializers/resque.rb

You can also set the namespace directly using resque-web:

$ resque-web -p 8282 -N myapp

or set the Redis connection string if you need to do something like select a different database:

$ resque-web -p 8282 -r localhost:6379:2

Using Passenger? Resque ships with a config.ru you can use. See Phusion's guide:

Apache: https://www.phusionpassenger.com/library/deploy/apache/deploy/ruby/ Nginx: https://www.phusionpassenger.com/library/deploy/nginx/deploy/ruby/

If you want to load Resque on a subpath, possibly alongside other apps, it's easy to do with Rack's URLMap:

require 'resque/server'

run Rack::URLMap.new \
  "/"       => Your::App.new,
  "/resque" => Resque::Server.new

Check examples/demo/config.ru for a functional example (including HTTP basic auth).

You can also mount Resque on a subpath in your existing Rails app by adding require 'resque/server' to the top of your routes file or in an initializer then adding this to routes.rb:

mount Resque::Server.new, :at => "/resque"

What should you run in the background? Anything that takes any time at all. Slow INSERT statements, disk manipulating, data processing, etc.

At GitHub we use Resque to process the following types of jobs:

• Warming caches
• Counting disk usage
• Building tarballs
• Building Rubygems
• Firing off web hooks
• Creating events in the db and pre-caching them
• Building graphs
• Deleting users
• Updating our search index

As of writing we have about 35 different types of background jobs.

Keep in mind that you don't need a web app to use Resque - we just mention "foreground" and "background" because they make conceptual sense. You could easily be spidering sites and sticking data which needs to be crunched later into a queue.

Jobs are persisted to queues as JSON objects. Let's take our Archive example from above. We'll run the following code to create a job:

repo = Repository.find(44)
repo.async_create_archive('masterbrew')

The following JSON will be stored in the file_serve queue:

{
    'class': 'Archive',
    'args': [ 44, 'masterbrew' ]
}

Because of this your jobs must only accept arguments that can be JSON encoded.

So instead of doing this:

Resque.enqueue(Archive, self, branch)

do this:

Resque.enqueue(Archive, self.id, branch)

This is why our above example (and all the examples in examples/) uses object IDs instead of passing around the objects.

While this is less convenient than just sticking a marshaled object in the database, it gives you a slight advantage: your jobs will be run against the most recent version of an object because they need to pull from the DB or cache.

If your jobs were run against marshaled objects, they could potentially be operating on a stale record with out-of-date information.

Want something like DelayedJob's send_later or the ability to use instance methods instead of just methods for jobs? See the examples/ directory for goodies.

We plan to provide first class async support in a future release.

If a job raises an exception, it is logged and handed off to the Resque::Failure module. Failures are logged either locally in Redis or using some different backend. To see exceptions while developing, see details below under Logging.

For example, Resque ships with Airbrake support. To configure it, put the following into an initialisation file or into your rake job:

# send errors which occur in background jobs to redis and airbrake
require 'resque/failure/multiple'
require 'resque/failure/redis'
require 'resque/failure/airbrake'

Resque::Failure::Multiple.classes = [Resque::Failure::Redis, Resque::Failure::Airbrake]
Resque::Failure.backend = Resque::Failure::Multiple

Keep this in mind when writing your jobs: you may want to throw exceptions you would not normally throw in order to assist debugging.

If you are using ActiveJob here's how your job definition will look:

class ArchiveJob < ApplicationJob
  queue_as :file_serve

  def perform(repo_id, branch = 'master')
    repo = Repository.find(repo_id)
    repo.create_archive(branch)
  end
end

class Repository
  def async_create_archive(branch)
    ArchiveJob.perform_later(self.id, branch)
  end
end

It is important to run ArchiveJob.perform_later(self.id, branch) rather than Resque.enqueue(Archive, self.id, branch). Otherwise Resque will process the job without actually doing anything. Even if you put an obviously buggy line like 0/0 in the perform method, the job will still succeed.

You may want to change the Redis host and port Resque connects to, or set various other options at startup.

Resque has a redis setter which can be given a string or a Redis object. This means if you're already using Redis in your app, Resque can re-use the existing connection.

String: Resque.redis = 'localhost:6379'

Redis: Resque.redis = $redis

For our rails app we have a config/initializers/resque.rb file where we load config/resque.yml by hand and set the Redis information appropriately.

Here's our config/resque.yml:

development: localhost:6379
test: localhost:6379
staging: redis1.se.github.com:6379
fi: localhost:6379
production: <%= ENV['REDIS_URL'] %>

And our initializer:

rails_root = ENV['RAILS_ROOT'] || File.dirname(__FILE__) + '/../..'
rails_env = ENV['RAILS_ENV'] || 'development'
config_file = rails_root + '/config/resque.yml'

resque_config = YAML::load(ERB.new(IO.read(config_file)).result)
Resque.redis = resque_config[rails_env]

Easy peasy! Why not just use RAILS_ROOT and RAILS_ENV? Because this way we can tell our Sinatra app about the config file:

$ RAILS_ENV=production resque-web rails_root/config/initializers/resque.rb

Now everyone is on the same page.

Also, you could disable jobs queueing by setting 'inline' attribute. For example, if you want to run all jobs in the same process for cucumber, try:

Resque.inline = ENV['RAILS_ENV'] == "cucumber"

Workers support basic logging to STDOUT.

You can control the logging threshold using Resque.logger.level:

# config/initializers/resque.rb
Resque.logger.level = Logger::DEBUG

If you want Resque to log to a file, in Rails do:

# config/initializers/resque.rb
Resque.logger = Logger.new(Rails.root.join('log', "#{Rails.env}_resque.log"))

If you're running multiple, separate instances of Resque you may want to namespace the keyspaces so they do not overlap. This is not unlike the approach taken by many memcached clients.

This feature is provided by the redis-namespace library, which Resque uses by default to separate the keys it manages from other keys in your Redis server.

Simply use the Resque.redis.namespace accessor:

Resque.redis.namespace = "resque:GitHub"

We recommend sticking this in your initializer somewhere after Redis is configured.

Resque allows to store count of processed and failed jobs.

By default it will store it in Redis using the keys stats:processed and stats:failed.

Some apps would want another stats store, or even a null store:

# config/initializers/resque.rb
class NullDataStore
 def stat(stat)
   0
 end

 def increment_stat(stat, by)
 end

 def decrement_stat(stat, by)
 end

 def clear_stat(stat)
 end
end

Resque.stat_data_store = NullDataStore.new

For a list of available plugins see https://github.com/resque/resque/wiki/plugins.

If you'd like to write your own plugin, or want to customize Resque using hooks (such as Resque.after_fork), see docs/HOOKS.md.

How does Resque compare to DelayedJob, and why would you choose one over the other?

• Resque supports multiple queues
• DelayedJob supports finer grained priorities
• Resque workers are resilient to memory leaks / bloat
• DelayedJob workers are extremely simple and easy to modify
• Resque requires Redis
• DelayedJob requires ActiveRecord
• Resque can only place JSONable Ruby objects on a queue as arguments
• DelayedJob can place any Ruby object on its queue as arguments
• Resque includes a Sinatra app for monitoring what's going on
• DelayedJob can be queried from within your Rails app if you want to add an interface

If you're doing Rails development, you already have a database and ActiveRecord. DelayedJob is super easy to setup and works great. GitHub used it for many months to process almost 200 million jobs.

Choose Resque if:

• You need multiple queues
• You don't care / dislike numeric priorities
• You don't need to persist every Ruby object ever
• You have potentially huge queues
• You want to see what's going on
• You expect a lot of failure / chaos
• You can setup Redis
• You're not running short on RAM

Choose DelayedJob if:

• You like numeric priorities
• You're not doing a gigantic amount of jobs each day
• Your queue stays small and nimble
• There is not a lot failure / chaos
• You want to easily throw anything on the queue
• You don't want to setup Redis

In no way is Resque a "better" DelayedJob, so make sure you pick the tool that's best for your app.

On certain platforms, when a Resque worker reserves a job it immediately forks a child process. The child processes the job then exits. When the child has exited successfully, the worker reserves another job and repeats the process.

Why?

Because Resque assumes chaos.

Resque assumes your background workers will lock up, run too long, or have unwanted memory growth.

If Resque workers processed jobs themselves, it'd be hard to whip them into shape. Let's say one is using too much memory: you send it a signal that says "shutdown after you finish processing the current job," and it does so. It then starts up again - loading your entire application environment. This adds useless CPU cycles and causes a delay in queue processing.

Plus, what if it's using too much memory and has stopped responding to signals?

Thanks to Resque's parent / child architecture, jobs that use too much memory release that memory upon completion. No unwanted growth.

And what if a job is running too long? You'd need to kill -9 it then start the worker again. With Resque's parent / child architecture you can tell the parent to forcefully kill the child then immediately start processing more jobs. No startup delay or wasted cycles.

The parent / child architecture helps us keep tabs on what workers are doing, too. By eliminating the need to kill -9 workers we can have parents remove themselves from the global listing of workers. If we just ruthlessly killed workers, we'd need a separate watchdog process to add and remove them to the global listing - which becomes complicated.

Workers instead handle their own state.

Resque uses Kernel#exit! for exiting its workers' child processes. So any at_exit callback defined in your application won't be executed when the job is finished and the child process exits.

You can alter this behavior by setting the RUN_AT_EXIT_HOOKS environment variable.

Here's a parent / child pair doing some work:

$ ps -e -o pid,command | grep [r]esque
92099 resque: Forked 92102 at 1253142769
92102 resque: Processing file_serve since 1253142769

You can clearly see that process 92099 forked 92102, which has been working since 1253142769.

(By advertising the time they began processing you can easily use monit or god to kill stale workers.)

When a parent process is idle, it lets you know what queues it is waiting for work on:

$ ps -e -o pid,command | grep [r]esque
92099 resque: Waiting for file_serve,warm_cache

Resque workers respond to a few different signals:

• QUIT - Wait for child to finish processing then exit
• TERM / INT - Immediately kill child then exit
• USR1 - Immediately kill child but don't exit
• USR2 - Don't start to process any new jobs
• CONT - Start to process new jobs again after a USR2

If you want to gracefully shutdown a Resque worker, use QUIT.

If you want to kill a stale or stuck child, use USR1. Processing will continue as normal unless the child was not found. In that case Resque assumes the parent process is in a bad state and shuts down.

If you want to kill a stale or stuck child and shutdown, use TERM

If you want to stop processing jobs, but want to leave the worker running (for example, to temporarily alleviate load), use USR2 to stop processing, then CONT to start it again. It's also possible to pause all workers.

When shutting down processes, Heroku sends every process a TERM signal at the same time. By default this causes an immediate shutdown of any running job leading to frequent Resque::TermException errors. For short running jobs, a simple solution is to give a small amount of time for the job to finish before killing it.

Resque doesn't handle this out of the box (for both cedar-14 and heroku-16), you need to install the resque-heroku-signals addon which adds the required signal handling to make the behavior described above work. Related issue: #1559

To accomplish this set the following environment variables:

• RESQUE_PRE_SHUTDOWN_TIMEOUT - The time between the parent receiving a shutdown signal (TERM by default) and it sending that signal on to the child process. Designed to give the child process time to complete before being forced to die.

• TERM_CHILD - Must be set for RESQUE_PRE_SHUTDOWN_TIMEOUT to be used. After the timeout, if the child is still running it will raise a Resque::TermException and exit.

• RESQUE_TERM_TIMEOUT - By default you have a few seconds to handle Resque::TermException in your job. RESQUE_TERM_TIMEOUT and RESQUE_PRE_SHUTDOWN_TIMEOUT must be lower than the heroku dyno timeout.

Workers will not process pending jobs if the Redis key pause-all-workers is set with the string value "true".

Resque.redis.set('pause-all-workers', 'true')

Nothing happens to jobs that are already being processed by workers.

Unpause by removing the Redis key pause-all-workers.

Resque.redis.del('pause-all-workers')

If you're using god to monitor Resque, we have provided example configs in examples/god/. One is for starting / stopping workers, the other is for killing workers that have been running too long.

If you're using monit, examples/monit/resque.monit is provided free of charge. This is not used by GitHub in production, so please send patches for any tweaks or improvements you can make to it.

If your workers remain idle for too long they may lose their MySQL connection. Depending on your version of Rails, we recommend the following:

In your perform method, add the following line:

class MyTask
  def self.perform
    ActiveRecord::Base.verify_active_connections!
    # rest of your code
  end
end

The Rails doc says the following about verify_active_connections!:

Verify active connections and remove and disconnect connections associated with stale threads.

In your perform method, instead of verify_active_connections!, use:

class MyTask
  def self.perform
    ActiveRecord::Base.clear_active_connections!
    # rest of your code
  end
end

From the Rails docs on clear_active_connections!:

Returns any connections in use by the current thread back to the pool, and also returns connections to the pool cached by threads that are no longer alive.

Want to hack on Resque?

First clone the repo and run the tests:

git clone git://github.com/resque/resque.git
cd resque
rake test

If the tests do not pass make sure you have Redis installed correctly (though we make an effort to tell you if we feel this is the case). The tests attempt to start an isolated instance of Redis to run against.

Also make sure you've installed all the dependencies correctly. For example, try loading the redis-namespace gem after you've installed it:

$ irb
>> require 'rubygems'
=> true
>> require 'redis/namespace'
=> true

If you get an error requiring any of the dependencies, you may have failed to install them or be seeing load path issues.

Resque ships with a demo Sinatra app for creating jobs that are later processed in the background.

Try it out by looking at the README, found at examples/demo/README.markdown.

Read CONTRIBUTING.md first.

Once you've made your great commits:

1. Fork Resque
2. Create a topic branch - git checkout -b my_branch
3. Push to your branch - git push origin my_branch
4. Create a Pull Request from your branch

Please add them to the FAQ or open an issue on this repo.

• Code: git clone git://github.com/resque/resque.git
• Home: http://github.com/resque/resque
• Docs: http://rubydoc.info/gems/resque
• Bugs: http://github.com/resque/resque/issues
• Gems: https://rubygems.org/gems/resque

This project uses Semantic Versioning

Chris Wanstrath :: chris@ozmm.org :: @defunkt