Sidekiq::Clutch

Sidekiq::Clutch is the API I always wanted when working with Sidekiq Pro Batches. So I built it!

Features:

• Add jobs to run in series or parallel or mix-and-match however you wish.
• Pass results from one job onto the next job in series.
• If running jobs in parallel, pass all the results to the following job in series.

Add this line to your application's Gemfile:

gem 'sidekiq-clutch'

And then execute:

$ bundle

Or install it yourself as:

$ gem install sidekiq-clutch

Include the Sidekiq::Clutch::Worker mixin in your job class and write your #perform method as usual. For example:

class MyJob1
  include Sidekiq::Worker
  include Sidekiq::Clutch::Worker

  def perform
    # do stuff
  end
end

Now, enqueue your jobs using a new Sidekiq::Clutch instance, like this:

clutch = Sidekiq::Clutch.new
clutch.jobs << MyJob1
clutch.jobs << [MyJob2, 'arg1', 'arg2']
clutch.parallel do
  clutch.jobs << [MyJob3, 3]
  clutch.jobs << [MyJob3, 4]
  clutch.jobs << [MyJob3, 5]
end
clutch.jobs << MyFinalizerJob
clutch.engage

The jobs will run in this order:

• First, MyJob1 will run.
• Second, MyJob2 with args 'arg1' and 'arg2' will run.
• Third, in parallel, three jobs of MyJob3, each with args 3, 4, and 5 will run.
• Last, the MyFinalizerJob will run.

You can access the results from the previous job in the series using the previous_results method:

class MyJob1
  include Sidekiq::Worker
  include Sidekiq::Clutch::Worker

  def perform
    'This string will be passed to the next job in the series'
  end
end

class MyJob2
  include Sidekiq::Worker
  include Sidekiq::Clutch::Worker

  def perform(arg1, arg2)
    puts "results of the previous job (MyJob1):"
    p previous_results # => ["This string will be passed to the next job in the series"]
  end
end

If the previous step in a series was a batch of parallel jobs, then previous_results will be an array containing all the results from all the parallel jobs.

Some notes:

• Results are always in an array, even if there was only a single prior job run in series.
• Results are serialized as JSON and temporarily stored in Redis, so only return small-ish values.

When jobs fail, Sidekiq will retry them as usual. See here for details on how Sidekiq handles job failure and retries.

Jobs next in a series will not be enqueued by Clutch as long as a prior job is failing.

If you wish to do something when a job in your batch fails, set on_failure like this:

clutch = Sidekiq::Clutch.new
clutch.on_failure = MyFailureHandlerJob
# add jobs here...
clutch.engage

MyFailureHandlerJob.new.perform(status) will be called if one of the following scenarios occur:

1. A job in series fails.
2. One or more jobs in parallel fail and the rest complete.

Note: retries will continue to occur even after your failure handler job is called.

You can nest Clutch instances too!

class MyNestedJob
  include Sidekiq::Worker
  include Sidekiq::Clutch::Worker

  def perform
    clutch = Sidekiq::Clutch.new(batch)
    clutch.jobs << [MyJob2, 'arg1', 'arg2']
    clutch.jobs << [MyJob3, 3]
    clutch.engage
    'result from MyNestedJob'
  end
end

clutch = Sidekiq::Clutch.new
clutch.jobs << MyJob1
clutch.jobs << MyNestedJob
clutch.jobs << MyFinalizerJob
clutch.engage

You can set the queue for jobs (overriding any queue specified in the job class itself) and callbacks by setting it on Clutch, like this:

clutch = Sidekiq::Clutch.new
clutch.queue = 'critical'
# add jobs here...
clutch.engage

I tested this with the third-party gem sidekiq-batch, which purports to be a drop-in replacement for Sidekiq Pro batches. However, that gem seems to not support nested callbacks, which my gem relies on. I'm sorry, this gem will not work with the sidekiq-batch gem -- it only works with sidekiq-pro.

The gem is available as open source under the terms of the MIT License.