Sucker Punch

Build Status Code Climate

Sucker Punch is a single-process Ruby asynchronous processing library. It's girl_friday and DSL sugar on top of Celluloid. With Celluloid's actor pattern, we can do asynchronous processing within a single process. This reduces costs of hosting on a service like Heroku along with the memory footprint of having to maintain additional jobs if hosting on a dedicated server. All queues can run within a single Rails/Sinatra process.

Sucker Punch is perfect for asynchronous processes like emailing, data crunching, or social platform manipulation. No reason to hold up a user when you can do these things in the background within the same process as your web application...

Sucker Punch is built on top of Celluloid Pools. Each job is setup as a pool, which equates to its own queue with individual workers working against the jobs. Unlike most other background processing libraries, Sucker Punch's jobs are stored in memory. The benefit to this is there is no additional infrastructure requirement (ie. database, redis, etc.). The downside is that if the web processes is restarted and there are jobs that haven't yet been processed, they will be lost. For this reason, Sucker Punch is generally recommended for jobs that are fast and non-mission critical (ie. logs, emails, etc.).

Installation

Add this line to your application's Gemfile:

gem 'sucker_punch', '~> 1.0'

And then execute:

$ bundle

Or install it yourself as:

$ gem install sucker_punch

Usage

Each job acts as its own queue and should be a separate Ruby class that:

  • includes SuckerPunch::Job
  • defines the instance method perform that includes the code the job will run when enqueued
# app/jobs/log_job.rb

class LogJob
  include SuckerPunch::Job

  def perform(event)
    Log.new(event).track
  end
end

Synchronous:

LogJob.new.perform("login")

Asynchronous:

LogJob.new.async.perform("login") # => nil

Jobs interacting with ActiveRecord should take special precaution not to exhaust connections in the pool. This can be done with ActiveRecord::Base.connection_pool.with_connection, which ensures the connection is returned back to the pool when completed.

# app/jobs/awesome_job.rb

class AwesomeJob
  include SuckerPunch::Job

  def perform(user_id)
    ActiveRecord::Base.connection_pool.with_connection do
      user = User.find(user_id)
      user.update_attributes(is_awesome: true)
    end
  end
end

We can create a job from within another job:

class AwesomeJob
  include SuckerPunch::Job

  def perform(user_id)
    ActiveRecord::Base.connection_pool.with_connection do
      user = User.find(user_id)
      user.update_attributes(is_awesome: true)
      LogJob.new.async.perform("User #{user.id} became awesome!")
    end
  end
end

The number of workers can be set from the Job using the workers method:

class LogJob
  include SuckerPunch::Job
  workers 4

  def perform(event)
    Log.new(event).track
  end
end

If the workers method is not set, the default is 2.

Perform In

Many background processing libraries have methods to perform operations after a certain amount of time. Fortunately, timers are built-in to Celluloid, so you can take advantage of them with the later method:

class Job
  include SuckerPunch::Job

  def perform(data)
    puts data
  end

  def later(sec, data)
    after(sec) { perform(data) }
  end
end

Job.new.async.perform("asdf")
Job.new.async.later(60, "asdf") # `perform` will be excuted 60 sec. later

Logger

SuckerPunch.logger = Logger.new('sucker_punch.log')
SuckerPunch.logger # => #<Logger:0x007fa1f28b83f0>

Note: If Sucker Punch is being used within a Rails application, Sucker Punch's logger is set to Rails.logger by default.

Exceptions

You can customize how to handle uncaught exceptions that are raised by your jobs.

For example, using Rails and the ExceptionNotification gem, add a new initializer config/initializers/sucker_punch.rb:

SuckerPunch.exception_handler { |ex| ExceptionNotifier.notify_exception(ex) }

Or, using Airbrake:

SuckerPunch.exception_handler { |ex| Airbrake.notify(ex) }

Full job data can be reported like this:

def perform(all, my, arguments)
  ... your code ...
rescue Exception
  Airbrake.error($!, [self.class.name, all, my, arguments].inspect)
  raise
end

Timeouts

Using Timeout causes persistent connections to randomly get corrupted. Do not use timeouts as control flow, use builtin connection timeouts. If you decide to use Timeout, only use it as last resort to know something went very wrong and ideally restart the worker process after every timeout.

Testing

Requiring this library causes your jobs to run everything inline. So a call to the following will actually be SYNCHRONOUS:

# spec/spec_helper.rb
require 'sucker_punch/testing/inline'
Log.new.async.perform("login") # => Will be synchronous and block until job is finished

Rails

If you're using Sucker Punch with Rails, there's a built-in generator task:

$ rails g sucker_punch:job logger

would create the file app/jobs/logger_job.rb with a unimplemented #perform method.

Active Job

Sucker Punch has been added as an Active Job adapter in Rails 4.2. See the guide for configuration and implementation.

Add Sucker Punch to your Gemfile:

gem 'sucker_punch'

And then configure the backend to use Sucker Punch:

# config/initializers/sucker_punch.rb
Rails.application.configure do
  config.active_job.queue_adapter = :sucker_punch
end

Troubleshooting

Initializers for forking servers (Unicorn, Passenger, etc.)

Previously, Sucker Punch required an initializer and that posed problems for Unicorn and Passenger and other servers that fork. Version 1 was rewritten to not require any special code to be executed after forking occurs. Please remove if you're using version >= 1.0.0

Class naming

Job classes are ultimately Celluloid Actor classes. As a result, class names are susceptible to being clobbered by Celluloid's internal classes. To ensure the intended application class is loaded, preface classes with ::, or use names like NotificationsMailer or UserMailer. Example:

class EmailJob
  include SuckerPunch::Job

  def perform(contact)
    @contact = contact
    ::Notifications.contact_form(@contact).deliver # => If you don't use :: in this case, the notifications class from Celluloid will be loaded
  end
end

Cleaning test data transactions

If you're running tests in transactions (using Database Cleaner or a native solution), Sucker Punch jobs may have trouble finding database records that were created during test setup because the job class is running in a separate thread and the Transaction operates on a different thread so it clears out the data before the job can do its business. The best thing to do is cleanup data created for tests jobs through a truncation strategy by tagging the rspec tests as jobs and then specifying the strategy in spec_helper like below. And do not forget to turn off transactional fixtures (delete, comment or set it to false).

# spec/spec_helper.rb
RSpec.configure do |config|

  # Turn off transactional fixtures (delete, comment or set it to `false`)
  # config.use_transactional_fixtures = true

  config.before(:each) do
    DatabaseCleaner.strategy = :transaction
  end

  # Clean up all jobs specs with truncation
  config.before(:each, job: true) do
    DatabaseCleaner.strategy = :truncation
  end

  config.before(:each) do
    DatabaseCleaner.start
  end

  config.after(:each) do
    DatabaseCleaner.clean
  end
end

# spec/jobs/email_job_spec.rb
require 'spec_helper'

# Tag the spec as a job spec so data is persisted long enough for the test
describe EmailJob, job: true do
  describe "#perform" do
    let(:user) { FactoryGirl.create(:user) }

    it "delivers an email" do
      expect {
        EmailJob.new.perform(user.id)
      }.to change{ ActionMailer::Base.deliveries.size }.by(1)
    end
  end
end

Gem Name

...is awesome. But I can't take credit for it. Thanks to @jmazzi for his superior naming skills. If you're looking for a name for something, he is the one to go to.

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 new Pull Request