Resque is widely used Redis-backed Ruby library for creating background jobs. ResqueStarter is a tool to start and manage multiple resque workers supporting graceful shutdown and restart, leveraging the Copy-on-write (COW).

PID   COMMAND
14814 resque_starter # preload app
14815  \_ resque:work[1]
14816  \_ resque:work[2]
14817  \_ resque:work[3]

Resque provides resque:workers task to run multiple resque workers, but it is only for development purpose as code comments says. It also provides an example configuration of god as resque.god, but it does not allow us to leverage Copy-on-write (CoW) to save memory of preloaded application.

Add this line to your application's Gemfile:

gem 'resque_starter'

And then execute:

$ bundle

Or install it yourself as:

$ gem install resque_starter

Please see example/resque.conf.rb

You can configure logger, pid file, number of concurrency, dequeue interval, queue lists.

bundle exec resque_starter -c /path/to/resque.conf.rb

Resque starter responds to a few different signals:

• TERM / INT - Quick shutdown, kills all workers immediately then exit
• QUIT - Graceful shutdown, waits for workers to finish processing then exit
• USR1 - Send USR1 to all workers, which immediately kill worker's child but don't exit
• USR2 - Send USR2 to all workers, which don't start to process any new jobs
• CONT - Send CONT to all workers, which start to process new jobs again after a USR2
• TTIN - Increment the number of worker processes by one
• TTOU - Decrement the number of worker processes by one with QUIT

Resque starter itself does not support graceful restart, yet. But, graceful restart can be done with server-starter.

Example configuration is available at server-starter/example/resque. See start_server and config/resque.conf.rb files.

HOW IT WORKS

On receiving HUP signal, server starter creates a new resque_starter (master) process. The new resque_starter (master) process forks a new resque worker. On after_fork, send TTOU to old resque_starter (master) process to gracefully shutdown one old resque worker. By repeating this procedure, new resque_starter process can be gracefully restarted. The number of working resque workers will be suppressed up to concurrency + 1 in this way.

ILLUSTRATION

On bootup:

PID   COMMAND
14813 server_starter
14814  \_ resque_starter
14815      \_ resque worker[0]
14816      \_ resque worker[1]
14817      \_ resque worker[2]

Send HUP:

PID   COMMAND
14813 server_starter
14814  \_ resque_starter (old)
14815      \_ resque worker[0] (dies)
14816      \_ resque worker[1]
14817      \_ resque worker[2]
14818  \_ resque_starter (new)
14819      \_ resque worker[0]

Finally:

PID   COMMAND
14813 server_starter
14818  \_ resque_starter (new)
14819      \_ resque worker[0]
14820      \_ resque worker[1]
14821      \_ resque worker[2]

1. Fork it ( https://github.com/sonots/resque_starter/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request