deepak / resque-pool

quickly fork a pool resque workers, saving memory (w/REE) and monitoring their uptime.

Resque Pool

Resque pool is a simple library for managing a pool of resque workers. Given a a config file (resque-pool.yml or config/resque-pool.yml), it will manage your workers for you, starting up the appropriate number of workers for each.

Benefits

• Less memory consumption - If you are using Ruby Enterprise Edition, or any ruby with copy-on-write safe garbage collection, this could save you a lot of memory when you are managing many workers.
• Simpler (less) config - If you are using monit or an init script to start up your workers, you can start up one pool, and it will manage your workers for you.
• Faster startup - if you are starting many workers at once, you would normally have them competing for CPU as they load their environments. Resque-pool can load the environment once and almost instantaneously fork all of the workers.

How to use

To configure resque-pool, you can use either resque-pool.yml or config/resque-pool.yml. To use resque-pool, require its rake tasks (resque/pool/tasks) in your rake file, and call the resque:pool task.

The YAML file supports both using root level defaults as well as environment specific overrides (RACK_ENV, RAILS_ENV, and RESQUE_ENV environment variables will be used to determine environment). For example, to use resque-pool with rails, in config/resque-pool.yml:

foo: 1
bar: 2
"foo,bar,baz": 1

production:
  "foo,bar,baz": 4

and in lib/tasks/resque.rake:

require 'resque/pool/tasks'
# this task will get called before resque:pool:setup
# preload the rails environment in the pool master
task "resque:setup" => :environment do
  # generic worker setup, e.g. Hoptoad for failed jobs
end
task "resque:pool:setup" do
  # close any sockets or files in pool master
  ActiveRecord::Base.connection.disconnect!
  # and re-open them in the resque worker parent
  Resque::Pool.after_prefork do |job|
    ActiveRecord::Base.establish_connection
  end
end

Then you can start the queues via:

rake resque:pool RAILS_ENV=production VERBOSE=1

This will start up seven worker processes, one exclusively for the foo queue, two exclusively for the bar queue, and four workers looking at all queues in priority. This is similar to if you ran the following:

rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=foo &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=bar &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=bar &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=foo,bar,baz &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=foo,bar,baz &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=foo,bar,baz &
rake resque:work RAILS_ENV=production VERBOSE=1 QUEUES=foo,bar,baz &

Resque already forks for its own child processes. The pool master will stay around monitoring the resque worker parents, giving three levels: a single pool master, many worker parents, and a worker child per worker (when the actual job is being processed). For example, ps -ef f | grep [r]esque (in Linux) might return something like the following:

rails    13858     1  0 13:44 ?        S      0:02 resque-pool-master: managing [13867, 13875, 13871, 13872, 13868, 13870, 13876]
rails    13867 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Waiting for foo
rails    13868 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Waiting for bar
rails    13870 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Waiting for bar
rails    13871 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Waiting for foo,bar,baz
rails    13872 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Forked 7481 at 1280343254
rails     7481 13872  0 14:54 ?        S      0:00      \_ resque-1.9.9: Processing foo since 1280343254
rails    13875 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Waiting for foo,bar,baz
rails    13876 13858  0 13:44 ?        S      0:00  \_ resque-1.9.9: Forked 7485 at 1280343255
rails     7485 13876  0 14:54 ?        S      0:00      \_ resque-1.9.9: Processing bar since 1280343254

An example startup script, which redirects STDOUT and STDERR and creates a pid file, is given in the examples directory.

SIGNALS

The pool master responds to the following signals:

• HUP - reload the config file, e.g. to change the number of workers per queue list
• QUIT - send QUIT to each worker parent and shutdown the master after all workers are done.
• INT - send QUIT to each worker parent and immediately shutdown master
• TERM - send TERM to each worker parent and immediately shutdown master
• WINCH - send QUIT to each worker, but keep master running (send HUP to reload config and restart workers)
• USR1/USR2/CONT - send the signal on to all worker parents (see Resque docs).

After a HUP, workers that are no longer needed will be gracefully shutdown via QUIT.

Other Features

Workers will watch the pool master, and gracefully shutdown if the master process dies (for whatever reason) before them.

You can specify an alternate config file by setting the RESQUE_POOL_CONFIG environment variable like so:

rake resque:pool RESQUE_ENV=production RESQUE_POOL_CONFIG=/path/to/my/config.yml

TODO

• do appropriate logging (e.g. all to one logfile, each queue to its own logfile, or each worker to its own logfile). Logfile location must be configurable, but default to log/resque-pool.log. Of course, since resque "logs" by writing to $stdout, this is really no more than redirecting stdout to the appropriate logfile.
• (optionally) daemonize, setting a PID file somewhere. configurable, of course, but default to tmp/pids/resque-pool.pid.
• recover gracefully from a malformed config file (on startup and HUP)
• procline for malformed config file, graceful shutdown... and other states?
• figure out a good automated way to test this (cucumber or rspec?)
• clean up the code (I stole most of it from unicorn, and it's still a bit bastardized); excessive use of vim foldmarkers are a code smell.
• rdoc
• incorporate resque-batchworker features? (v2.0)
• web interface for adding and removing workers (etc) (v2.0)

Contributors

• John Schult (config file can be split by environment)