Re: December 2024 stabilization week CANCELED!
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 18 Dec 2024 06:15:40 UTC
On Wed, Dec 18, 2024 at 08:29:46AM +0800, George Neville-Neil wrote: G> > due to shortage of human resources and due to reservation of certain hardware G> > for the Christmas events, we won't be able to properly A/B test the December G> > snapshot at Netflix as we usually do. Given that testing performed by Netflix G> > is the largest sieve to catch regressions, and we still don't have any second G> > large party participating in the stabweek, me together with David dhw@ decided G> > that this month the stabweek to be skipped. G> G> How much hardware is required for this? Could this be done with new machines either in the cluster or at Sentex (the test lab) or does this have to be done at NetFlix because that's where the systems are best set up to do this? If more hardware is needed you should talk to Ed at the Foundation and get some ordered. I don't think there is anything in the world comparable to the A/B testing that David does at Netflix. Although, the test covers only some subset of FreeBSD functionality: TCP (including advanced stacks), sendfile(2), kernel TLS and a small set of device drivers, it does that at a very large scale on production traffic. You can't repeat that at home :) But the point is not about substituting the David's test with an other test (next week or anytime). The point of the stabweeks cycle is that several parties pick the same revision and extensively test it with their own tests. I was really hoping that more large companies that use FreeBSD would join the stabweek cycle, that the FreeBSD cluster updates would also use the same revision, that more individuals will follow the stabweek. Unfortunately, little participation is observed right now. This December, if we exclude Netflix/dhw@ test from the stabweek, it would be only a handful of individuals updating their desktops, which in my opinion, isn't enough to announce a stabweek and the associated freeze. But for any future months any additional testing resources are very much appreciated. E.g., I'd be glad to see Peter's Holm stress2 running on the Sentex test lab during Monday of stabweek. If you can provide more coverage in future stabweeks - please join. -- Gleb Smirnoff