[three]Bean

Some notes from the releng FAD

Jun 15, 2015 | categories: releng, fedora, koji View Comments

A number of us were funded to come together last weekend for a release engineering FAD in Westford. I'll save summaries of the whole event for writeups by others, but I will say that we planned work on the first day with an 'agile' exercise proposed and facilitated by imcleod. It worked well and mitigated any centrifugal forces acting on the FAD.

Read Adam Miller's post for a solid summary of the whole event. Here's just a run down of the things I personally got involved in:

  • The runroot plugin - we've been talking about it forever -- an Internet search reveals a mention of it from a meeting in 2007 <http://fedoraproject.org/wiki/Extras/SteeringCommittee/Meeting-20070315>. What finally made it a necessity is pungi4. We're hoping to move the compose to using that, and it in turn relies on the runroot plugin in koji to do most of its work.

    A portion of it was open sourced last year. I got it installed in staging and then found we were missing some more parts to it. We sorted those all out, submitted some patches upstream to koji and got a staging proof of concept working. As of last week, we have it in production and a new channel of builders dedicated to it.

  • Our staging koji instance was originally setup last year at the Bodhi2 FAD using the external repos feature of koji. This worked well enough for testing scratch builds and little things like that, but was insufficient for testing some larger parts of the releng process (the compose). We compared some options and have settled on a solution. Some pieces of that puzzle have been implemented but I've yet to finish putting it all together. Hopefully done this week.

  • As far as planning goes, I thought the koji 2.0 discussions were quite fruitful, and I'm really excited at how the architecture it shaping up.

  • We also had some good interstitial sessions talking about role and requirements for what has been heretofore called the ComposeDB. There's still more discussion to have before we can land on a planning document for it, but fwiw, we think we have a much better name for it than ComposeDB.

I'll end with noting that on the day before the FAD started, a number of us got together to plan for and hack on Fedora Hubs. decause and I stayed up late almost every night and wrote some pretty neat proofs of concept for it. There's good stuff running now in the prototype if you want to try to get it running on your local system. Fun, fun. Happy Hacking!

View Comments
blog comments powered by Disqus