I've been looking around for a tool to help me keep up with builds of many different projects. I'm a buildbot fan, but it isn't appropriate when you've got to deal with a large number of different projects.
cruisecontrol.rb is nice, but is simultaneously overkill (why is this thing written in rails?) and not a perfect match. It's extensible, though, so I've modified it to suit my needs. In particular, polling is just not an appropriate way to run a CI system.
So now I'm making use of github's post-receive hooks to trigger events indicating a new build should run. However, there wasn't an easy way to get this into cc.rb, so I created a new scheduler in my fork at github that uses beanstalkd as a simple queue to communicate between a dedicated mongrel web-hook receiver and the rest of the world.
- beanstalkd v 0.11 or greater
- And this project's cchook.rb
- dustin-beanstalk-client
- mongrel
- json
Once your beanstalk server is running, you'll need to edit config.yml to point to it. config.yml.example should make it obvious how this works. Note that you can have more than one beanstalkd running if for some reason you have so much build traffic coming through as to necessitate it.
Then, run it and configure your project's web hook to point to the URL of this server.
If a project is sending requests to cchook, but cchook isn't configured to know of the project, it'll be ignored (well, logged then ignored), so make them match.
On the cc.rb side, edit your project's cruise_config.rb (as found in ~/.cruise/projects/project_name/cruise_config.rb) to require the client gem and point to your tube.
For example, assume your beanstalk server is beanstalk-server and the queue you want to use for this project is called myproj-build, your config may look like this:
require 'beanstalk-client'
Project.configure do |project|
project.scheduler.close if project.scheduler.is_a?(BeanstalkScheduler)
queue = Beanstalk::Pool.new ['beanstalk-server:11300'], 'myproj-build'
project.scheduler = BeanstalkScheduler.new(project, queue)
end