I run a build bot to identify build + install issues of Gentoo Linux software packages*.
Currently 8 chroot images are running in parallel at a 64bit hardened Gentoo Linux host (i7-3930, 12 cores, 64 GB RAM, 2x 3 TB HDD). Each image is set up from a recent stage3 with a randomized combination of (~)amd64 + profile + USE flag set.
Within each image all packages will be tried to be emerged in an arbitrary order. The repository at the host is shared to all images and is synced every 3 hours. Added or modified packages are put on top of each work queue. Once a day @system is updated in each image. No parallel make jobs and no parallel emerge jobs are made (1 make job at N images rules over <=N make jobs at 1 image). No manual unmerge or any other action is usually made.
More than 5,500 packages are emerged daily, less than 1.5% do fail. The calculated coverage of the repository with 8 images is 85% after 7 days.
An image is replaced by a fresh new one usually after 12-21 days (based on this model). The old image is kept around for few weeks.
The Portage File List is fed too. Github hosts the source code.
*just to have fun, but to redeem too to a Linux distribution I do use and trust since 2003. The Gentoo project has its own tinderbox cluster with a slightly different approach.
The tinderbox runs in each chroot:
"qsearch --all | sort --random-sort | xargs -n 1 emerge --update"
and just parses the output.
back to my home page