Skip to content

famoseagle/sweatshop

Repository files navigation

Sweatshop

Sweatshop provides an api to background resource intensive tasks. Much of the api design was copied from Workling, with a few tweaks. Currently, it runs rabbitmq and kestrel, but it can support any number of queues.

Installing

gem install sweatshop
freeze in your gems directory (add config.gem 'sweatshop' to your environment)
cd vendor/gems/sweatshop
rake setup

Writing workers

Put email_worker.rb into app/workers and sublcass Sweatshop::Worker:

class EmailWorker < Sweatshop::Worker
  def send_mail(to)
    user = User.find_by_id(to)
    Mailer.deliver_welcome(to)
  end
end

Then, anywhere in your app you can execute:

EmailWorker.async_send_mail(1)

The async signifies that this task will be placed on a queue to be serviced by the EmailWorker possibly on another machine. You can also call:

EmailWorker.send_mail(1)

That will do the work immediately, without placing the task on the queue. You can also define a queue_group at the top of the file which will allow you to split workers out into logical groups. This is important if you have various machines serving different queues.

Running the queue

Sweatshop has been tested with Rabbit and Kestrel, but it will also work with Starling. Please use the following resources to install the server:

Kestrel: http://github.com/robey/kestrel/tree/master

Rabbit: http://github.com/ezmobius/nanite/tree/master

config/sweatshop.yml specifies the machine address of the queue (default localhost:5672). You can also specify the queue type with the queue param.

Rabbit cluster support

The following example configuration shows support for Rabbit clusters within a queue group:

    default:
      queue: rabbit
      cluster:
        - hostA:5672
        - hostB:5672
      user: 'guest'
      pass: 'guest'
      vhost: '/'
    enable: true

Sweatshop will attempt to connect to each server listed under "cluster" in order, until it either manages to establish a connection or until it runs out of servers.

If you only have a single Rabbit server, you can omit the "cluster" option and just add "host" and "port" (or host: localhost:5672) options, as shown below:

   default:
     queue: rabbit
     host: localhost
     port: 5672
     user: 'guest'
     pass: 'guest'
     vhost: '/'
   enable: true

Running the workers

Assuming you ran rake setup in Rails, you can type:

script/sweatshop

By default, the script will run all workers defined in the app/workers dir. Every task will be processed on each queue using a round-robin algorithm. You can also add the -d flag which will put the worker in daemon mode. The daemon also takes other params. Add a -h for more details.

script/sweatshop -d
script/sweatshop -d stop

If you would like to run Sweatshop as a daemon on a linux machine, use the initd.sh script provided in the sweatshop/script dir.

REQUIREMENTS

memcache (for kestrel)
carrot (for rabbit)

LICENSE

Copyright (c) 2009 Amos Elliston, Geni.com; Published under The MIT License, see License