Stalker - a job queueing DSL for Beanstalk

Beanstalkd is a fast, lightweight queueing backend inspired by mmemcached. The Ruby Beanstalk client is a bit raw, however, so Stalker provides a thin wrapper to make job queueing from your Ruby app easy and fun.

Queueing jobs

From anywhere in your app:

require 'stalker'

Stalker.enqueue('email.send', :to => '[email protected]')
Stalker.enqueue('post.cleanup.all')
Stalker.enqueue('post.cleanup', :id => post.id)

Working jobs

In a standalone file, typically jobs.rb or worker.rb:

require 'stalker'
include Stalker

job 'email.send' do |args|
  Pony.send(:to => args['to'], :subject => "Hello there")
end

job 'post.cleanup.all' do |args|
  Post.all.each do |post|
    enqueue('post.cleanup', :id => post.all)
  end
end

job 'post.cleanup' do |args|
  Post.find(args['id']).cleanup
end

Running

First, make sure you have Beanstalkd installed and running:

$ sudo port install beanstalkd
$ beanstalkd

Stalker:

$ sudo gem install stalker

Now run a worker using the stalk binary:

$ stalk jobs.rb
[Thu May 13 01:08:19 -0700 2010] Working 3 jobs: [ email.send post.cleanup.all post.cleanup ]
[Thu May 13 01:08:21 -0700 2010] -> send.email ([email protected])
[Thu May 13 01:08:21 -0700 2010] -> send.email finished in 31ms

Stalker will log to stdout as it starts working each job, and then again when the job finishes including the ellapsed time in milliseconds.

Filter to a list of jobs you wish to run with an argument:

$ stalk jobs.rb post.cleanup.all,post.cleanup
[Sat Apr 17 14:13:40 -0700 2010] Working 2 jobs: [ post.cleanup.all post.cleanup ]

In a production environment you may run one or more high-priority workers (limited to short/urgent jobs) and any number of regular workers (working all jobs). For example, two workers working just the email.send job, and four running all jobs:

$ for i in 1 2; do stalk jobs.rb email.send > log/urgent-worker.log 2>&1; end
$ for i in 1 2 3 4; do stalk jobs.rb > log/worker.log 2>&1; end

Error Handling

If you include an error block in your jobs definition, that block will be invoked when a worker encounters an error. You might use this to report errors to an external monitoring service:

error do |e|
   Exceptional.handle(e)
end

Tidbits

  • Jobs are serialized as JSON, so you should stick to strings, integers, arrays, and hashes as arguments to jobs. e.g. don't pass full Ruby objects - use something like an ActiveRecord/MongoMapper/CouchRest id instead.
  • Because there are no class definitions associated with jobs, you can queue jobs from anywhere without needing to include your full app's environment.
  • If you need to change the location of your Beanstalk from the default (localhost:11300), set BEANSTALK_URL in your environment, e.g. export BEANSTALK_URL=beanstalk://example.com:11300/
  • The stalk binary is just for convenience, you can also run a worker with a straight Ruby command: $ ruby -r jobs -e Stalker.work

Running the tests

If you wish to hack on Stalker, install these extra gems:

$ gem install contest mocha turn

Run the tests:

$ turn

Meta

Created by Adam Wiggins

Patches from Jamie Cobbett, Scott Water, Keith Rarick, Mark McGranaghan

Heavily inspired by Minion by Orion Henry

Released under the MIT License: http://www.opensource.org/licenses/mit-license.php

http://github.com/adamwiggins/stalker