Skip to content
This repository has been archived by the owner on Sep 19, 2022. It is now read-only.

chartmogul/sidekiq-priority_queue

Repository files navigation

Sidekiq Priority Queue

Extends Sidekiq with support for queuing jobs with a fine grained priority and emulating multiple queues using a single Redis sorted set, ideal for multi-tenant applications.

The standard Sidekiq setup performs really well using Redis lists, but lists can only be strict FIFO queues, which can be hugely problematic when they process slowly and one user may need to wait hours behind a backlog of jobs.

Sidekiq Priority Queue offers a plug-in solution retaining the simplicity and performance of Sidekiq. The priority queue is a building block for emulating sub-queues (per tenant or user) by de-prioritising jobs according to how many jobs are already in this sub-queue.

Sources of inspiration are naturally Sidekiq itself, the fantastic Redis documentation, and https://github.com/gocraft/work

Installation

gem install sidekiq-priority_queue

Configuration

Sidekiq.configure_server do |config|
  config.options[:fetch] = Sidekiq::PriorityQueue::Fetch
end

Sidekiq.configure_client do |config|
  config.client_middleware do |chain|
    chain.add Sidekiq::PriorityQueue::Client
  end
end

Usage

class Worker
  include Sidekiq::Worker
  sidekiq_options priority: 1000
end

Alternatively, you can split jobs into subqueues (via a proc) which are deprioritised based on the subqueue size:

class Worker
  include Sidekiq::Worker

  # args[0] will take the `user_id` argument below, and assign priority dynamically.
  sidekiq_options subqueue: ->(args){ args[0] }

  def perform(user_id, other_args)
    # do jobs
  end
end

Testing

For example in your spec/rails_helper.rb you can include this line:

require 'sidekiq/priority_queue/testing'

next to the call to Sidekiq::Testing.inline!. It disables the feature for testing and falls back to inline/fake modes. If you accidentally require this in production code, it will likewise fall back to normal Sidekiq scheduling.

Development

  • Run docker-compose up -d to start up a temporary redis instance.
  • Run bundle install to install dependencies.
  • Run the tests with bundle exec rake