need to update nix, there are a couple of places that hard-code the length of a rev
LnL: thank you for your PR yesterday, that looks much better
gchristensen: if it's still flaky you can increase the window, but I'm assuming updates will fail within 5 minutes unless something more is going on
yeah, good idea
niksnut gchristensen what do you think about adding a flake.nix to nixpkgs master? This would then be part of the 20.03 branch-off and allow downstream projects to experiment with flakes before the actual release of the nix flakes branch
I sort of feel like adding it to nixpkgs might be committing a bit too hard to it, but maybe it is a good idea
I'm trying an experiment where I try and ask for help on projects I want to get done, instead of just power through when I can :) thank you LnL, tilpner, hexa-, and flokli for helping with the prom & alertmanager troubles this week.
in the spirit of that: I need help with a small project to improve Hydra's performance. I need to find all the queries in Hydra which write to the `builds` table. You may be thinking "yuck, perl" but you won't need to wrlite any, and this really needs doing. I'm hoping for a volunteer to do some spelunking. After we've found them, some perl will need to be written, but I'll ask for help on that
separately, so no commitment ;)
imagine one were to look for those queries, what's the intent behind finding those queries?
right now the Build stable has columns: `job', `jobset`, `project` which are strings. Queries which filter this table can often take 30s+, sometimes 60s+ or even 90s+. I have a patch and migration which (1) adds `id` fields to the jobs, jobsets, and projects tables, and adds job_id, jobset_id, and project_id to the Builds table. those slow queries seem to take just a few seconds or even less
when using by-ID indexes.
so the goal is to, when writing to Builds, also write the `*_id` data and then update the slowest read queries. over time we may migrate all of them and drop the stringy ones, but that is out of scope for now
orivej has joined #nixos-dev
justanotheruser has quit [Ping timeout: 265 seconds]