Database backed asynchronous priority queue -- Extracted from Shopify
Switch branches/tags
Nothing to show
Pull request Compare This branch is 1 commit ahead of getaroom:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



Delayed_job (or DJ) encapsulates the common pattern of asynchronously executing longer tasks in the background.

It is a direct extraction from Shopify where the job table is responsible for a multitude of core tasks. Amongst those tasks are:

  • sending massive newsletters
  • image resizing
  • http downloads
  • updating smart collections
  • updating solr, our search server, after product changes
  • batch imports
  • spam checks

What is this fork for?

My purpose with this fork is make delayed_job for flexible. That's means you can customize how your workers behave.

The common use will be to have several workers running concurrently (in one process) and each one with differents constraints so they'll run different kind of jobs.


The library evolves around a delayed_jobs table which looks as follows:

create_table :delayed_jobs, :force => true do |table|
    table.integer  :priority, :default => 0      # Allows some jobs to jump to the front of the queue
    table.integer  :attempts, :default => 0      # Provides for retries, but still fail eventually.
    table.text     :handler                      # YAML-encoded string of the object that will do work
    table.string   :job_type                     # Class name of the job object, for type-specific workers
    table.string   :name                         # The display name, an informative field or can be use to filter jobs
    table.string   :last_error                   # reason for last failure (See Note below)
    table.datetime :run_at                       # When to run. Could be for immediately, or sometime in the future.
    table.datetime :locked_at                    # Set when a client is working on this object
    table.datetime :failed_at                    # Set when all retries have failed (actually, by default, the record is deleted instead)
    table.string   :locked_by                    # Who is working on this object (if locked)
    table.datetime :finished_at			 # Used for statiscics / monitoring

You can generate the migration executing:

$ script/generate delayed_job
  exists  db/migrate
  create  db/migrate/20090807090217_create_delayed_jobs.rb

On failure, the job is scheduled again in 5 seconds + N ** 4, where N is the number of retries.

The default MAX_ATTEMPTS is 25 (jobs can override this value by responding to :max_attempts). After this, the job either deleted (default), or left in the database with "failed_at" set. With the default of 25 attempts, the last retry will be 20 days later, with the last interval being almost 100 hours.

The default MAX_RUN_TIME is 4.hours. If your job takes longer than that, another computer could pick it up. It's up to you to make sure your job doesn't exceed this time. You should set this to the longest time you think the job could take.

By default, it will delete failed jobs. If you want to keep failed jobs, set Delayed::Job.destroy_failed_jobs = false. The failed jobs will be marked with non-null failed_at.

Same thing for successful jobs. They're deleted by default and, to keep them, set Delayed::Job.destroy_successful_jobs = false. They will be marked with finished_at. This is useful for gathering statistics like how long jobs took between entering the queue (created_at) and being finished (finished_at).

You have several named scopes for searching unfinished/finsihed/failed jobs, very useful when destroy_successful_jobs is false Delayed::Job.unfinished, Delayed::Job.finsihed, Delayed::Job.failed.

Here is an example of changing job parameters in Rails:

# config/initializers/delayed_job_config.rb
Delayed::Job.destroy_failed_jobs     = false
Delayed::Job.destroy_successful_jobs = false
silence_warnings do
  Delayed::Job.const_set("MAX_ATTEMPTS", 3)
  Delayed::Job.const_set("MAX_RUN_TIME", 5.hours)

  Delayed::Worker.const_set("SLEEP", 5.minutes.to_i)

Note: If your error messages are long, consider changing last_error field to a :text instead of a :string (255 character limit).


Jobs are simple ruby objects with a method called perform. Any object which responds to perform can be stuffed into the jobs table. Job objects are serialized to yaml so that they can later be resurrected by the job runner.

class NewsletterJob <, :emails)
  def perform
    emails.each { |e| NewsletterMailer.deliver_text_to_email(text, e) }

Delayed::Job.enqueue'lorem ipsum...', Customers.find(:all).collect(&:email))

There is also a second way to get jobs in the queue: send_later., massive_csv)

And also you can specified priority as second parameter and the time the job should execute as thrird one

class FooJob
  def perform

important_job =
normal_job    =

# Delayed::Job.enqueue( job, priority, start_at )
Delayed::Job.enqueue important_job, 100
Delayed::Job.enqueue normal_job, 1, 2.hours.from_now

This will simply create a Delayed::PerformableMethod job in the jobs table which serializes all the parameters you pass to it. There are some special smarts for active record objects which are stored as their text representation and loaded from the database fresh when the job is actually run later.

Running the jobs

You can invoke rake jobs:work which will start working off jobs. You can cancel the rake task with CTRL-C.

You can also run by writing a simple script/job_runner, and invoking it externally:

require File.dirname(__FILE__) + '/../config/environment'

Workers can be running on any computer, as long as they have access to the database and their clock is in sync. You can even run multiple workers on per computer, but you must give each one a unique name.

require File.dirname(__FILE__) + '/../config/environment'
N = 10
workers = []
N.times do |n|
workers << do :name => "Worker #{n}" ).start

workers.first.join # wait until finish (signal catched)

Keep in mind that each worker will check the database at least every 5 seconds.

Note: The rake task will exit if the database has any network connectivity problems.

If you only want to run specific types of jobs in a given worker, include them when initializing the worker: => "SimpleJob").start => ["SimpleJob", "NewsletterJob"]).start

Also for a more specific restriction you can define in your job's classes a display_name method, and create workers to specific kind of jobs

# 1 - The job class that does the real work
class MyJob
  def initialize( data )
    @some_data = data

  def perform
    # do the real work

  def display_name
    "foobar #{@some_data}"

# 2 - Enqueue jobs

# 3 - Create workers, one for each type of "data" {
  # This worker will only perform jobs which display name is like "%foobar%" :name => "Worker for foobar", :only_for => "foobar"
} { :name => "Worker for arrr", :only_for => "arrr"

Cleaning up

You can invoke rake jobs:clear to delete all jobs in the queue.


  • 2.0.2: Only update run_at when it's gonna be executed another time for sure (attempts < max_attempts)

  • 2.0.1: named_scope Delayed::Job.failed/finished/unfinished (jobs that have failed, have finished ok or still haven't been done)

  • 2.0.0: Contains the changes made in this fork, the ability to create workers with individual constraints without interfere to other workers

  • 1.7.0: Added failed_at column which can optionally be set after a certain amount of failed job attempts. By default failed job attempts are destroyed after about a month.

  • 1.6.0: Renamed locked_until to locked_at. We now store when we start a given job instead of how long it will be locked by the worker. This allows us to get a reading on how long a job took to execute.

  • 1.5.0: Job runners can now be run in parallel. Two new database columns are needed: locked_until and locked_by. This allows us to use pessimistic locking instead of relying on row level locks. This enables us to run as many worker processes as we need to speed up queue processing.

  • 1.2.0: Added #send_later to Object for simpler job creation

  • 1.0.0: Initial release