Resque/Cucumber Testing Integration for generic installations

cucumber_external_resque_worker_generic.rb

# This is adapted from this gist: https://gist.github.com/532100 by Square
# The main difference is that it doesn't require Bluth for WorkerBase
# It also calls prune_dead_workers on start so it doesn't hang on every other run
# It does not do anything special to avoid connecting to your main redis instance; you should be
# doing that elsewhere

class CucumberExternalResqueWorker
  DEFAULT_STARTUP_TIMEOUT = 1.minute
  COUNTER_KEY = "cucumber:counter"

  class << self
    attr_accessor :pid, :startup_timeout

    def start
      # Call from a Cucumber support file so it is run on startup
      return unless Rails.env.test?
      if self.pid = fork
        start_parent
        wait_for_worker_to_start
      else
        start_child
      end
    end

    def install_hooks_on_startup
      # Call from a Rails initializer
      return unless Rails.env.test?
      # Because otherwise crashed workers cause a fork and we pause the actual worker forever
      Resque::Worker.all.each { |worker| worker.prune_dead_workers }
      install_pause_on_start_hook
      install_worker_base_counter_patch
    end

    def process_all
      # Call from a Cucumber step
      unpause
      sleep 1 until done?
      pause
    end

    def incr
      Resque.redis.incr(COUNTER_KEY)
    end

    def decr
      Resque.redis.decr(COUNTER_KEY)
    end

    def reset_counter
      Resque.redis.set(COUNTER_KEY, 0)
    end

    private

    def done?
      Resque.redis.get(CucumberExternalResqueWorker::COUNTER_KEY).to_i.zero?
    end

    def pause(pid = self.pid)
      return unless Rails.env.test?
      Process.kill("USR2", pid)
    end

    def unpause
      return unless Rails.env.test?
      Process.kill("CONT", pid)
    end

    def start_parent
      at_exit do
        #reset_counter
        Process.kill("KILL", pid) if pid
      end
    end

    def start_child
      # Array form of exec() is required here, otherwise the worker is not a direct child process of cucumber.
      # If it's not the direct child process then the PID returned from fork() is wrong, which means we can't
      # communicate with the worker.
      exec('rake', 'resque:work', "QUEUE=*", "RAILS_ENV=test", "VVERBOSE=1")
    end

    def wait_for_worker_to_start
      self.startup_timeout ||= DEFAULT_STARTUP_TIMEOUT
      start = Time.now.to_i
      while (Time.now.to_i - start) < startup_timeout
        return if worker_started?
        sleep 1
      end

      raise "Timeout while waiting for the worker to start. Waited #{startup_timeout} seconds."
    end

    def worker_started?
      Resque.info[:workers].to_i > 0
    end

    def install_pause_on_start_hook
      Resque.before_first_fork do
        #reset_counter
        pause(Process.pid)
      end
    end

    def install_worker_base_counter_patch
      Resque.class_eval do
        class << self
          def enqueue_with_counters(*args, &block)
            CucumberExternalResqueWorker.incr
            enqueue_without_counters(*args, &block)
          end
          alias_method_chain :enqueue, :counters
        end
      end
      Resque::Job.class_eval do
        def perform_with_counters(*args, &block)
          perform_without_counters(*args, &block)
        ensure
          CucumberExternalResqueWorker.decr
        end
        alias_method_chain :perform, :counters
      end
    end
  end
end

other_changes.rb

# In the Cucumber environment file features/support/env.rb
# After
  require 'cucumber/rails'
# Add:
  require 'lib/cucumber_external_resque_worker'
  CucumberExternalResqueWorker.start

# Change:
  DatabaseCleaner.strategy = :transaction
# to
  DatabaseCleaner.strategy = :truncation

# Also, create a file: config/initializers/cucumber_external_resque.rb
require 'lib/cucumber_external_resque_worker'
CucumberExternalResqueWorker.install_hooks_on_startup

# In my controller, I have:
  def start
    if params[:job] then
      Resque.enqueue(DemoJob, j.id)
    end
    redirect_to :action => :index
  end
# DemoJob:
class DemoJob
  def self.queue
    :demojob
  end
  def perform(job_id)
    j = Job.find(job_id)
    j.value = "done"
    j.save
  end
    

# In your actual Cucumber step file, for instance:
When /I click the start button for "([^"]*)"/ do |jobname|
  CucumberExternalResqueWorker.reset_counter
  Resque.remove_queue(DemoJob.queue)
  click_button(jobname)
end
When /all open jobs are processed/ do
  CucumberExternalResqueWorker.process_all
end

# And you cucumber feature file looks like:
# Scenario: Starting a job
#   When I click the start button for "Test Job 1"
#     And all open jobs are processed
#   Then I am on the job index page
#     And I should see an entry for "Test Job 1" with a value of "done".

I tried using something like this and ran into a few problems:

• I might want to start one resque worker listening on one queue, and another resque worker listening on a different queue. This code has no concept of queues -- so if the Rails application starts five jobs on five different queues and the resque worker started by the test only listens on one queue, then it will never end.
• I think you want to override Resque.push instead of Resque.enqueue. If I need to enqueue a class name that's not available to Rails, I would be using Resque.push. (Resque.enqueue calls Resque.push)

I think something like this would be super useful to be put into a Gem, maybe even merged into resque itself.