Stalker - a job queueing DSL for Beanstalk

Beanstalkd is a fast, lightweight queueing backend inspired by mmemcached. Stalker originally is a thin wrapper around the quite raw Ruby Beanstalk client, however, I updated it to use its successor Beaneater.

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)

Job Status

The current status of a job can be checked via

Stalker.status ‹job_id›

The ‹job_id› is a String or Integer. It's returned from Stalker.enqueue.

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.id)
  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 - with Bundler:

gem 'stalker', github: 'nkoehring/stalker'

Stalker - with Bundler: $ git clone git://github.com/nkoehring/stalker.git $ cd stalker $ gem build stalker.gemspec $ gem install stalker-.gem

Now run a worker using the stalk binary:

$ stalk jobs.rb
Working 3 jobs: [ email.send post.cleanup.all post.cleanup ]
Working send.email ([email protected])
Finished send.email 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
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; done
$ for i in 1 2 3 4; do stalk jobs.rb > log/worker.log 2>&1; done

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, job, args|
  Exceptional.handle(e)
end

Before filter

If you wish to run a block of code prior to any job:

before do |job|
  puts "About to work #{job}"
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/. You can specify multiple beanstalk servers, separated by whitespace or comma, e.g. export BEANSTALK_URL="beanstalk://b1.example.com:11300/, beanstalk://b2.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

Make sure you have a beanstalkd running, then run the tests:

$ ruby test/stalker_test.rb