Active Job Basics

This guide provides you with all you need to get started in creating, enqueuing and executing background jobs.

After reading this guide, you will know:

How to create jobs.
How to enqueue jobs.
How to run jobs in the background.
How to send emails from your application asynchronously.

1 What is Active Job?

Active Job is a framework for declaring jobs and making them run on a variety of queuing backends. These jobs can be everything from regularly scheduled clean-ups, to billing charges, to mailings. Anything that can be chopped up into small units of work and run in parallel.

2 The Purpose of Active Job

The main point is to ensure that all Rails apps will have a job infrastructure in place. We can then have framework features and other gems build on top of that, without having to worry about API differences between various job runners such as Delayed Job and Resque. Picking your queuing backend becomes more of an operational concern, then. And you'll be able to switch between them without having to rewrite your jobs.

Rails by default comes with an asynchronous queuing implementation that runs jobs with an in-process thread pool. Jobs will run asynchronously, but any jobs in the queue will be dropped upon restart.

3 Create and Enqueue Jobs

This section will provide a step-by-step guide to creating a job and enqueuing it.

3.1 Create the Job

Active Job provides a Rails generator to create jobs. The following will create a job in app/jobs (with an attached test case under test/jobs):

$ bin/rails generate job guests_cleanup
invoke  test_unit
create    test/jobs/guests_cleanup_job_test.rb
create  app/jobs/guests_cleanup_job.rb

You can also create a job that will run on a specific queue:

$ bin/rails generate job guests_cleanup --queue urgent

If you don't want to use a generator, you could create your own file inside of app/jobs, just make sure that it inherits from ApplicationJob.

Here's what a job looks like:

class GuestsCleanupJob < ApplicationJob
  queue_as :default

  def perform(*guests)
    # Do something later
  end
end

Note that you can define perform with as many arguments as you want.

If you already have an abstract class and its name differs from ApplicationJob, you can pass the --parent option to indicate you want a different abstract class:

$ bin/rails generate job process_payment --parent=payment_job

class ProcessPaymentJob < PaymentJob
  queue_as :default

  def perform(*args)
    # Do something later
  end
end

3.2 Enqueue the Job

Enqueue a job using perform_later and, optionally, set. Like so:

# Enqueue a job to be performed as soon as the queuing system is
# free.
GuestsCleanupJob.perform_later guest

# Enqueue a job to be performed tomorrow at noon.
GuestsCleanupJob.set(wait_until: Date.tomorrow.noon).perform_later(guest)

# Enqueue a job to be performed 1 week from now.
GuestsCleanupJob.set(wait: 1.week).perform_later(guest)

# `perform_now` and `perform_later` will call `perform` under the hood so
# you can pass as many arguments as defined in the latter.
GuestsCleanupJob.perform_later(guest1, guest2, filter: "some_filter")

That's it!

3.3 Enqueue Jobs in Bulk

You can enqueue multiple jobs at once using perform_all_later. For more details see Bulk Enqueuing.

4 Job Execution

For enqueuing and executing jobs in production you need to set up a queuing backend, that is to say, you need to decide on a 3rd-party queuing library that Rails should use. Rails itself only provides an in-process queuing system, which only keeps the jobs in RAM. If the process crashes or the machine is reset, then all outstanding jobs are lost with the default async backend. This may be fine for smaller apps or non-critical jobs, but most production apps will need to pick a persistent backend.

4.1 Backends

Active Job has built-in adapters for multiple queuing backends (Sidekiq, Resque, Delayed Job, and others). To get an up-to-date list of the adapters see the API Documentation for ActiveJob::QueueAdapters.

4.2 Setting the Backend

You can easily set your queuing backend with config.active_job.queue_adapter:

# config/application.rb
module YourApp
  class Application < Rails::Application
    # Be sure to have the adapter's gem in your Gemfile
    # and follow the adapter's specific installation
    # and deployment instructions.
    config.active_job.queue_adapter = :sidekiq
  end
end

You can also configure your backend on a per job basis:

class GuestsCleanupJob < ApplicationJob
  self.queue_adapter = :resque
  # ...
end

# Now your job will use `resque` as its backend queue adapter, overriding what
# was configured in `config.active_job.queue_adapter`.

4.3 Starting the Backend

Since jobs run in parallel to your Rails application, most queuing libraries require that you start a library-specific queuing service (in addition to starting your Rails app) for the job processing to work. Refer to library documentation for instructions on starting your queue backend.

Here is a noncomprehensive list of documentation:

5 Queues

Most adapters support multiple queues. With Active Job you can schedule the job to run on a specific queue using queue_as:

class GuestsCleanupJob < ApplicationJob
  queue_as :low_priority
  # ...
end

You can prefix the queue name for all your jobs using config.active_job.queue_name_prefix in application.rb:

# config/application.rb
module YourApp
  class Application < Rails::Application
    config.active_job.queue_name_prefix = Rails.env
  end
end

# app/jobs/guests_cleanup_job.rb
class GuestsCleanupJob < ApplicationJob
  queue_as :low_priority
  # ...
end

# Now your job will run on queue production_low_priority on your
# production environment and on staging_low_priority
# on your staging environment

You can also configure the prefix on a per job basis.

class GuestsCleanupJob < ApplicationJob
  queue_as :low_priority
  self.queue_name_prefix = nil
  # ...
end

# Now your job's queue won't be prefixed, overriding what
# was configured in `config.active_job.queue_name_prefix`.

The default queue name prefix delimiter is '_'. This can be changed by setting config.active_job.queue_name_delimiter in application.rb:

# config/application.rb
module YourApp
  class Application < Rails::Application
    config.active_job.queue_name_prefix = Rails.env
    config.active_job.queue_name_delimiter = "."
  end
end

# app/jobs/guests_cleanup_job.rb
class GuestsCleanupJob < ApplicationJob
  queue_as :low_priority
  # ...
end

# Now your job will run on queue production.low_priority on your
# production environment and on staging.low_priority
# on your staging environment

To control the queue from the job level you can pass a block to queue_as. The block will be executed in the job context (so it can access self.arguments), and it must return the queue name:

class ProcessVideoJob < ApplicationJob
  queue_as do
    video = self.arguments.first
    if video.owner.premium?
      :premium_videojobs
    else
      :videojobs
    end
  end

  def perform(video)
    # Do process video
  end
end

ProcessVideoJob.perform_later(Video.last)

If you want more control on what queue a job will be run you can pass a :queue option to set:

MyJob.set(queue: :another_queue).perform_later(record)

Make sure your queuing backend "listens" on your queue name. For some backends you need to specify the queues to listen to.

6 Priority

Some adapters support priorities at the job level, where jobs can be prioritized relative to others in the queue or across all queues.

You can schedule a job to run with a specific priority using queue_with_priority:

class GuestsCleanupJob < ApplicationJob
  queue_with_priority 10
  # ...
end

Note that this will not have any effect with adapters that do not support priorities.

Similar to queue_as, you can also pass a block to queue_with_priority to be evaluated in the job context:

class ProcessVideoJob < ApplicationJob
  queue_with_priority do
    video = self.arguments.first
    if video.owner.premium?
      0
    else
      10
    end
  end

  def perform(video)
    # Process video
  end
end

ProcessVideoJob.perform_later(Video.last)

You can also pass a :priority option to set:

MyJob.set(priority: 50).perform_later(record)

If a lower priority number performs before or after a higher priority number depends on the adapter implementation. Refer to documentation of your backend for more information. Adapter authors are encouraged to treat a lower number as more important.

7 Callbacks

Active Job provides hooks to trigger logic during the life cycle of a job. Like other callbacks in Rails, you can implement the callbacks as ordinary methods and use a macro-style class method to register them as callbacks:

class GuestsCleanupJob < ApplicationJob
  queue_as :default

  around_perform :around_cleanup

  def perform
    # Do something later
  end

  private
    def around_cleanup
      # Do something before perform
      yield
      # Do something after perform
    end
end

The macro-style class methods can also receive a block. Consider using this style if the code inside your block is so short that it fits in a single line. For example, you could send metrics for every job enqueued:

class ApplicationJob < ActiveJob::Base
  before_enqueue { |job| $statsd.increment "#{job.class.name.underscore}.enqueue" }
end

7.1 Available Callbacks

Please note that when enqueuing jobs in bulk using perform_all_later, callbacks such as around_enqueue will not be triggered on the individual jobs. See Bulk Enqueuing Callbacks.

8 Bulk Enqueuing

You can enqueue multiple jobs at once using perform_all_later. Bulk enqueuing reduces the number of round trips to the queue data store (like Redis or a database), making it a more performant operation than enqueuing the same jobs individually.

perform_all_later is a top-level API on Active Job. It accepts instantiated jobs as arguments (note that this is different from perform_later). perform_all_later does call perform under the hood. The arguments passed to new will be passed on to perform when it's eventually called.

Here is an example calling perform_all_later with GuestCleanupJob instances:

# Create jobs to pass to `perform_all_later`.
# The arguments to `new` are passed on to `perform`
guest_cleanup_jobs = Guest.all.map { |guest| GuestsCleanupJob.new(guest) }

# Will enqueue a separate job for each instance of `GuestCleanupJob`
ActiveJob.perform_all_later(guest_cleanup_jobs)

# Can also use `set` method to configure options before bulk enqueuing jobs.
guest_cleanup_jobs = Guest.all.map { |guest| GuestsCleanupJob.new(guest).set(wait: 1.day) }

ActiveJob.perform_all_later(guest_cleanup_jobs)

perform_all_later logs the number of jobs successfully enqueued, for example if Guest.all.map above resulted in 3 guest_cleanup_jobs, it would log Enqueued 3 jobs to Async (3 GuestsCleanupJob) (assuming all were enqueued).

The return value of perform_all_later is nil. Note that this is different from perform_later, which returns the instance of the queued job class.

8.1 Enqueue Multiple Active Job Classes

With perform_all_later, it's also possible to enqueue different Active Job class instances in the same call. For example:

class ExportDataJob < ApplicationJob
  def perform(*args)
    # Export data
  end
end

class NotifyGuestsJob < ApplicationJob
  def perform(*guests)
    # Email guests
  end
end

# Instantiate job instances
cleanup_job = GuestsCleanupJob.new(guest)
export_job = ExportDataJob.new(data)
notify_job = NotifyGuestsJob.new(guest)

# Enqueues job instances from multiple classes at once
ActiveJob.perform_all_later(cleanup_job, export_job, notify_job)

8.2 Bulk Enqueue Callbacks

When enqueuing jobs in bulk using perform_all_later, callbacks such as around_enqueue will not be triggered on the individual jobs. This behavior is in line with other Active Record bulk methods. Since callbacks run on individual jobs, they can't take advantage of the bulk nature of this method.

However, the perform_all_later method does fire an enqueue_all.active_job event which you can subscribe to using ActiveSupport::Notifications.

The method successfully_enqueued? can be used to find out if a given job was successfully enqueued.

8.3 Queue Backend Support

For perform_all_later, bulk enqueuing needs to be backed by the queue backend.

For example, Sidekiq has a push_bulk method, which can push a large number of jobs to Redis and prevent the round trip network latency. GoodJob also supports bulk enqueuing with the GoodJob::Bulk.enqueue method. The new queue backend Solid Queue has added support for bulk enqueuing as well.

If the queue backend does not support bulk enqueuing, perform_all_later will enqueue jobs one by one.

9 Action Mailer

One of the most common jobs in a modern web application is sending emails outside of the request-response cycle, so the user doesn't have to wait on it. Active Job is integrated with Action Mailer so you can easily send emails asynchronously:

# If you want to send the email now use #deliver_now
UserMailer.welcome(@user).deliver_now

# If you want to send the email through Active Job use #deliver_later
UserMailer.welcome(@user).deliver_later

Using the asynchronous queue from a Rake task (for example, to send an email using .deliver_later) will generally not work because Rake will likely end, causing the in-process thread pool to be deleted, before any/all of the .deliver_later emails are processed. To avoid this problem, use .deliver_now or run a persistent queue in development.

10 Internationalization

Each job uses the I18n.locale set when the job was created. This is useful if you send emails asynchronously:

I18n.locale = :eo

UserMailer.welcome(@user).deliver_later # Email will be localized to Esperanto.

11 Supported Types for Arguments

ActiveJob supports the following types of arguments by default:

Basic types (NilClass, String, Integer, Float, BigDecimal, TrueClass, FalseClass)
Symbol
Date
Time
DateTime
ActiveSupport::TimeWithZone
ActiveSupport::Duration
Hash (Keys should be of String or Symbol type)
ActiveSupport::HashWithIndifferentAccess
Array
Range
Module
Class

11.1 GlobalID

Active Job supports GlobalID for parameters. This makes it possible to pass live Active Record objects to your job instead of class/id pairs, which you then have to manually deserialize. Before, jobs would look like this:

class TrashableCleanupJob < ApplicationJob
  def perform(trashable_class, trashable_id, depth)
    trashable = trashable_class.constantize.find(trashable_id)
    trashable.cleanup(depth)
  end
end

Now you can simply do:

class TrashableCleanupJob < ApplicationJob
  def perform(trashable, depth)
    trashable.cleanup(depth)
  end
end

This works with any class that mixes in GlobalID::Identification, which by default has been mixed into Active Record classes.

11.2 Serializers

You can extend the list of supported argument types. You just need to define your own serializer:

# app/serializers/money_serializer.rb
class MoneySerializer < ActiveJob::Serializers::ObjectSerializer
  # Converts an object to a simpler representative using supported object types.
  # The recommended representative is a Hash with a specific key. Keys can be of basic types only.
  # You should call `super` to add the custom serializer type to the hash.
  def serialize(money)
    super(
      "amount" => money.amount,
      "currency" => money.currency
    )
  end

  # Converts serialized value into a proper object.
  def deserialize(hash)
    Money.new(hash["amount"], hash["currency"])
  end

  private
    # Checks if an argument should be serialized by this serializer.
    def klass
      Money
    end
end

and add this serializer to the list:

# config/initializers/custom_serializers.rb
Rails.application.config.active_job.custom_serializers << MoneySerializer

Note that autoloading reloadable code during initialization is not supported. Thus it is recommended to set-up serializers to be loaded only once, e.g. by amending config/application.rb like this:

# config/application.rb
module YourApp
  class Application < Rails::Application
    config.autoload_once_paths << "#{root}/app/serializers"
  end
end

12 Exceptions

Exceptions raised during the execution of the job can be handled with rescue_from:

class GuestsCleanupJob < ApplicationJob
  queue_as :default

  rescue_from(ActiveRecord::RecordNotFound) do |exception|
    # Do something with the exception
  end

  def perform
    # Do something later
  end
end

If an exception from a job is not rescued, then the job is referred to as "failed".

12.1 Retrying or Discarding Failed Jobs

A failed job will not be retried, unless configured otherwise.

It's possible to retry or discard a failed job by using retry_on or discard_on, respectively. For example:

class RemoteServiceJob < ApplicationJob
  retry_on CustomAppException # defaults to 3s wait, 5 attempts

  discard_on ActiveJob::DeserializationError

  def perform(*args)
    # Might raise CustomAppException or ActiveJob::DeserializationError
  end
end