In the past 10 months I've gone from one spare machine running WCG, to my main workstation running WCG, to building machines explicitly for running WCG, to the realization that I'll soon have a small fleet of machines which do nothing but crunch WCG workunits. In other words, I'm on the verge of having a compute farm in my house.

Since I'm a devops engineer, this naturally meant there was only one thing to do: set up a configuration management and orchestration system for these machines. It's called Homefarm, and if this sounds interesting to you, you can check out the code and docs here on Github.

It assumes one Raspberry Pi to act as a control node, and any number of x86_64 machines running Alpine Linux as compute nodes. I chose Alpine because of its compactness and focus on security. Management is done via Ansible playbooks.

As of this writing, the version is 0.9.1. I don't plan to make any changes other than bugfixes before 1.0.0, and work on new capabilities will pick up after that.

If you have highly customized BOINC client configurations, then Homefarm won't deliver what you're looking for right now (managing cc_config.xml is on the roadmap for post-1.0.0).

It does handle some things nicely though, including per-node project configurations and automated config of boinctui on the control node (so that it always knows about your compute nodes and can connect to them).

I don't expect anyone to run out and reinstall all their machines to test this, but please do look things over and let me know what you think, especially if you feel like there's something I should be handling, or if anything in the documentation is unclear. Thanks!