Send code from your rails project on a long journey, monitor it's progress, and get results back when it's done. Built on Beanstalk, Memcached, AsyncObserver, and Elock.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Pack Mule

Pack Mule is a Ruby on Rails plugin that gives you the facilities you need to run your code for hours or seconds in parallel across your cluster of Beanstalk workers. You can monitor the progress, coordinate tasks, and get results back when they complete.

Usage Example

We name our runner so that we can query it later. The false argument tells the runner not to publish progress updates -- instead we'll be asking for it when we need it.

>> runner ="this_is_an_arbitrary_name", false)
=> #<SleepyRunner:0x6444c>

First we can run the tasks in the background and block until they are done:

=> 51

Or we can run it in the background:

>> task = runner.run_async
=> [1, 'localhost:11300']

The enqueue method returns a job reference which is just a tuple of a job id and a beanstalk server.

We can now ask for overall progress:

>> runner.update_progress
=> "53.0 / 101.0"

progress is returned as a string that can be eval'd to give a percentage or you can split on " / " to get the number of jobs completed and queued.

Any instance with the same name will work:

=> "59.0 / 101.0"

So we wait a while and ask for an updated progress

>> runner.update_progress
=> "101.0 / 101.0"

Now that it's finished, we can access our return value now:

>> runner.get_return_value!
=> 53

Building Your Mule

Your mule is any class that inherits from PackMule::Runner

require 'pack_mule'
class SleepyRunner < PackMule::Runner
  def initialize(*args)
    @record_return_values = true
  def run_async(nap_count = 100)
    enqueue :run
  def run(nap_count = 100)
    jobs = []
    nap_count.times do
      jobs << enqueue :snooze, rand(2)
    deferred_result :wakeup, jobs
  def wakeup(durations)
    durations.inject(0){|total, duration| total + duration}
  def snooze(duration)
    sleep duration

You really only need to know about two special methods to build your mule:

enqueue(method_name, *args)

The enqueue method has the same syntax as send but instead of calling the method immediately, it queues the method call through beanstalk and it will run as a separate job. Enqueue returns a reference to the job instead of a return value. Hold on to that, because you will need it later. Enqueue may be called any number of times, but if you plan to call it a lot, you might consider using enqueue_each(method, list, *args) because it will run faster.

deferred_result(method_name, jobs, *args)

The deferred_result method will call the method you specify passing the return values for the jobs and any additional arguments you passed to it. If you call deferred_result from a synchronous context, it will block until all the jobs have completed and return the result of the deferred method. If you call it from within an asynchronous context, it will not block -- instead it returns a JobReference instance to the job that will check for results later. PackMule knows about JobReferences and will take care of all the tracking for you. If you ask for the return value of the original task it will go find it for you.

Instantiation Options

  • record_return_values - set to true if you want to capture return values. Otherwise, they'll be mixed with the hay and feed to the mules. All return values must respond to rrepr.
  • push_progress_updates - If you are reading your progress more often than it changes set this to true. You can then use current_progress to read it instead of update_progress and the current_progress will be updated for you whenever it changes.
  • time_to_run - Defaults to 2.minutes. If any method takes longer there's a pretty good chance it will get ran again. Set this to the twice the maximum amount of time you expect any method to take.


  • MemCache - Memcache stores the temporary information related to a Mule run like return values, related jobs, error counts, etc.
  • Beanstalk - Beanstalk is a priority queue server.
  • AsyncObserver - AsyncObserver is a rails plugin client for Beanstalk. It provides a worker that can run in a Rails context and easy access to queueing methods for asynchronous execution. It does not allow for task coordination. You can and should use it directly without pack mule for your simple tasks that don't take very long.
  • Elock - Elock is a mutex server that allows for resource access coordination across many processes and machines. Copy the included elock initializer to your initializer directories to setup access to your elock servers.