Ent Unique Jobs

Mike Perham edited this page Jul 31, 2018 · 10 revisions

If your application code creates duplicate jobs, the unique jobs feature in Sidekiq Enterprise makes it easy to ensure only a single copy of a job is in Redis. For instance, perhaps you create a job to sync an address change with a 3rd party system every time a form is submitted. If the form is submitted twice, you don't need to create the second job if the first job is still pending.

See unique jobs in action here:

Unique Jobs

Enable

To activate the feature, enable it in your initializer:

# Disable uniqueness in testing, this has the potential to cause much head scratching...
Sidekiq::Enterprise.unique! unless Rails.env.test?

Use

Declare a time period during which your job should be considered unique:

class MyWorker
  include Sidekiq::Worker
  sidekiq_options unique_for: 10.minutes

  def perform(...)
  end
end

This means that a second job can be pushed to Redis after 10 minutes or after the first job has successfully processed. If your job retries for a while, 10 minutes can pass, thus allowing another copy of the same job to be pushed to Redis. Design your jobs so that uniqueness is considered best effort, not a 100% guarantee A time limit is mandatory so that if a process crashes, any locks it is holding won't last forever.

Jobs are considered unique based on (class, args, queue), meaning a job with the same args can be pushed to different queues.

Scheduled Jobs

The uniqueness period for a scheduled job includes the delay time. If you use MyWorker.perform_in(1.hour, ...), the uniqueness for this job will last 70 minutes (1.hour + the 10.minutes unique_for TTL). You won't be able to push the same job until it runs successfully or 70 minutes have passed.

Retries

A job that is pending retry will still hold the unique lock and prevent further jobs from being enqueued until the retry succeeds or the timeout passes. Manually removing the job from the retry queue will not remove the lock.

Unlock Policy

The unique_until option allows you to control when the unique lock is removed. The default value is success: the job will not unlock until it executes successfully, it will remain locked even if it raises an error and goes into the retry queue.

The alternative value is start: the job will unlock right before it starts executing. This fixes the possibility of a race condition for some unique jobs between the finish of the job and unlocking: read #3471 for detail.

sidekiq_options unique_for: 20.minutes, unique_until: :start

If the job raises an error, it will not try to retake the lock and may be duplicate enqueued. I recommend avoiding the start policy unless you know your job is affected by the race condition.

Bypassing Uniqueness

If you really want to push a job and bypass the uniqueness check, pass false as the unique_for option:

MyWorker.set(unique_for: false).perform_async(1,2,3)

Reliability

The uniqueness feature makes an extra call to Redis before pushing the job. This network call is not protected by reliable_push so uniqueness can raise network errors in your webapp and cause the push to Redis to fail. This is a trait of any client middleware that makes a network call and not specific to uniqueness.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.