Skip to content

Sucker Punch is a Ruby asynchronous processing library using concurrent-ruby, heavily influenced by Sidekiq and girl_friday.

License

Notifications You must be signed in to change notification settings

brandonhilkert/sucker_punch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Sucker Punch

Build Code Climate

Sucker Punch is a single-process Ruby asynchronous processing library. 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 application (eg. Rails, Sinatra, etc.) 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 concurrent-ruby. 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.). However, if the web processes are restarted with jobs remaining in the queue, 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', '~> 3.0'

And then execute:

$ bundle

You can also run (same as two steps above):

bundle add sucker_punch

Or install it yourself as:

$ gem install sucker_punch

Backwards Compatibility

No breaking changes were introduced in version 3.x.

In version ~> 2.0.0, the syntax to enqueue an asynchronous background job was changed from:

LogJob.new.async.perform(...)

to:

LogJob.perform_async(...)

If you're upgrading from a pre-2.0.0 release and want to retain the old syntax LogJob.new.async.perform(...), you can include sucker_punch/async_syntax in your application.

For Rails, you could add an initializer:

# config/initializers/sucker_punch.rb

require 'sucker_punch/async_syntax'

Usage

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

  • includes SuckerPunch::Job
  • defines the perform instance method 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.perform_async("login")

Configure the # of the Workers

The default number of workers (threads) running against your job is 2. If you'd like to configure this manually, the number of workers can be set on the job using the workers class method:

class LogJob
  include SuckerPunch::Job
  workers 4

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

Configure the Queue Size

The default number of jobs that can be queued is unlimited. If you wish to restrict this you can set max_jobs as follows:

class LogJob
  include SuckerPunch::Job
  max_jobs 10

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

Executing Jobs in the Future

Many background processing libraries have methods to perform operations after a certain amount of time and Sucker Punch is no different. Use the perform_in with an argument of the number of seconds in the future you would like the job to job to run.

class DataJob
  include SuckerPunch::Job

  def perform(data)
    puts data
  end
end

DataJob.perform_async("asdf") # immediately perform asynchronously
DataJob.perform_in(60, "asdf") # `perform` will be executed 60 sec. later

ActiveRecord Connection Pool Connections

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(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.perform_async("User #{user.id} became awesome!")
    end
  end
end

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:

# ex    => The caught exception object
# klass => The job class
# args  => An array of the args passed to the job

SuckerPunch.exception_handler = -> (ex, klass, args) { ExceptionNotifier.notify_exception(ex) }

Or, using Airbrake:

SuckerPunch.exception_handler = -> (ex, klass, args) { Airbrake.notify(ex) }

Shutdown Timeout

Sucker Punch goes through a series of checks to attempt to shut down the queues and their threads. A "shutdown" command is issued to the queues, which gives them notice but allows them to attempt to finish all remaining jobs. Subsequently enqueued jobs are discarded at this time.

The default shutdown_timeout (the # of seconds to wait before forcefully killing the threads) is 8 sec. This is to allow applications hosted on Heroku to attempt to shutdown prior to the 10 sec. they give an application to shutdown with some buffer.

To configure something other than the default 8 sec.:

  SuckerPunch.shutdown_timeout = 15 # # of sec. to wait before killing threads

Queue Class Methods

Sucker Punch leverages concurrent-friendly queues per-process. These class methods operating on all queues might be helpful in some edge cases.

  • SuckerPunch::Queue.stats - Returns a hash keyed by each queue's name with total, busy, and timeout statistics for each queue's workers and jobs.
  • SuckerPunch::Queue.clear - Calls clear for each queue. Susceptible to race conditions. Only use in testing.
  • SuckerPunch::Queue.shutdown_all - Used with SuckerPunch's at_exit hook. Waits for all queues to be idle using the shutdown timeout configuration above.
  • SuckerPunch::Queue.wait - Waits for all queues to become idle with no timeout.

Timeouts

Using Timeout causes persistent connections to randomly get corrupted. Do not use timeouts as control flow, use built-in 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'
LogJob.perform_async("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.

Sinatra

If you're using Sucker Punch with Sinatra, you must require Sucker Punch before Sinatra:

# app.rb

require 'sucker_punch'
require 'sinatra'

This will ensure Sucker Punch's at_exit() handler to clean up and shutdown queues does not happen before Sinatra starts up via its own at_exit() handler.

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/application.rb
class Application < Rails::Application
  # ...
  config.active_job.queue_adapter = :sucker_punch
end

If you want to use Sucker Punch version 2.0.0+ with Rails < 5.0.0, be sure to include the backwards compatibility module in an initializer:

# config/initializers/sucker_punch.rb

require 'sucker_punch/async_syntax'

FAQ

What is the difference between sucker_punch and ActiveJob::QueueAdapters::AsyncAdapter?

Not much at this point. SuckerPunch existed before ActiveJob, but ultimately uses similar code under the covers. I'd recommend using AsyncAdapter if you're using Rails.

Troubleshooting

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

Previously, Sucker Punch required an initializer and that posed problems for servers that fork (ie. Unicorn and Passenger). 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

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

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

About

Sucker Punch is a Ruby asynchronous processing library using concurrent-ruby, heavily influenced by Sidekiq and girl_friday.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages