!RROtHmAaQIkiJzJZZE:nixos.org

NixOS Infrastructure

384 Members
Next Infra call: 2024-07-11, 18:00 CEST (UTC+2) | Infra operational issues backlog: https://github.com/orgs/NixOS/projects/52 | See #infra-alerts:nixos.org for real time alerts from Prometheus.117 Servers

Load older messages


SenderMessageTime
10 Feb 2025
@k900:0upti.meK900I'm going to guess that nej is crashing somehow 17:28:55
@k900:0upti.meK900And stopping mid-line17:29:01
@hexa:lossy.networkhexaswitched to a generation before nej17:33:47
@k900:0upti.meK900It ded17:34:23
@hexa:lossy.networkhexait is a the stopping step 😄 17:34:53
@hexa:lossy.networkhexaat least the errory is pretty17:35:12
@hexa:lossy.networkhexaand we're back17:37:20
@k900:0upti.meK900I've restarted the eval again 17:46:49
@hexa:lossy.networkhexahttps://github.com/NixOS/hydra/issues/143717:47:02
@k900:0upti.meK900Well eval hasn't died yet 18:00:05
@k900:0upti.meK900At least 18:00:07
11 Feb 2025
@infinisil:matrix.orginfinisil

From here

Lix is the Nix implementation which runs hydra.nixos.org

I'm doubting the accurancy of that:

  • https://hydra.nixos.org/ says "using nix-2.24.8" at the bottom
  • No mention of Lix in https://github.com/nixos/hydra
  • Only this singular use in infra, but I don't think that's used for actual evals, and that line was introduced without any discussion

Am I missing something?

22:08:15
@k900:0upti.meK900It's used on the builders because cppnix had a weird random segfault bug and that was the easiest thing we could do to get the builds going22:09:52
@k900:0upti.meK900The coordinator has to be cppnix because that's the only thing our hydra branch works with22:10:15
@infinisil:matrix.orginfinisilAh so evals are still done using Nix22:12:04
@infinisil:matrix.orginfinisil K900: Can you link me to the segfault issue? 22:12:26
@k900:0upti.meK900Not off the top of my head22:12:38
@k900:0upti.meK900 @hexa (signing key rotation when) or @John Ericson might have it? 22:13:16
@Ericson2314:matrix.orgJohn Ericson infinisil: K900 I do not know about this segfault issue 22:15:18
@Ericson2314:matrix.orgJohn Ericsonor this mixed lix nix deployment22:15:28
@Ericson2314:matrix.orgJohn Ericsonheh22:15:33
@infinisil:matrix.orginfinisilOpened a PR to document this: https://github.com/NixOS/infra/pull/55422:28:25
@roberthensing:matrix.orgRobert Hensing (roberth)same. A core dump and/or gdb stack trace would be very helpful22:48:31
@arianvp:matrix.orgArianWe have core dump collection enabled at work where we're running 2.19. So we have coredumps for these specific crashes available I think.22:57:24
@arianvp:matrix.orgArianThis is the module we use for that. It's super simplistic. But works: https://github.com/arianvp/nixos-stuff/blob/master/configs/utm/coredump-upload.nix23:02:14
@arianvp:matrix.orgArianAre we collecting coredumps of the builders already today? Could hook that up to something like sentry or bugsnag or whatever and auto triage this and remove the human in the loop23:08:48
@roberthensing:matrix.orgRobert Hensing (roberth)2.19 is not officially supported anymore, so you may want to update those builders. Nonetheless, outdated coredumps may be valid input for a current bug23:30:30
@hexa:lossy.networkhexawow, this seems to have been blown out of proportions23:50:39
@hexa:lossy.networkhexathis was indeed the case23:51:52
@hexa:lossy.networkhexaand I might just have forgotten to ever send any backtrace upstream23:52:17

Show newer messages


Back to Room ListRoom Version: 6