| 22 Jun 2026 |
 Sergei Zimmerman (xokdvium) | Hydra stil seems to be doing quite poorly:
https://hydra.nixos.org/build/332391533/log sh: line 1: g++: command not found. Don't know what that's about, GC issues again?
| 21:56:27 |
| 23 Jun 2026 |
 vmfunc | sh: line 1: g++: command not found, huh?? | 04:27:09 |
 Vladimír Čunát | FYI: nixos-unstable is on a 7 days old commit now, but this time queue-runner isn't the issue; we have a different pair of blockers: https://github.com/NixOS/nixpkgs/pull/530382#issuecomment-4776516421 | 06:52:57 |
| Vladimír Čunát | * FYI: nixos-unstable is on a 7 days old commit now, but this time infra isn't the issue; we have a different pair of blockers: https://github.com/NixOS/nixpkgs/pull/530382#issuecomment-4776516421 | 06:53:13 |
| Vladimír Čunát |
GC issues again?
Step 22 here looks like some GC issue as well:
https://hydra.nixos.org/build/332396502#tabs-buildsteps
(when I fetch this /nix/store path from cache.nixos.org, the file is there)
| 07:41:38 |
 Mic92 | @xokdvium:matrix.org: there was still a TOCTOU bug in my temproots fix, I added now a regression test for the future: https://github.com/NixOS/hydra/commit/3bcea35e40abece41eee094ea1cb0e8407ed7325 | 07:57:21 |
| Mic92 | @hexa:lossy.network: I got some work to do and I am out this evening. In case you got time, I have also some new builder limit options that should be wired | 08:01:04 |
| Mic92 | The last 3 commits in the branch | 08:01:29 |
 hexa | ready for prod? | 08:02:07 |
| hexa | 
Download | 08:10:16 |
| hexa | builds stuck in connecting for hours | 08:10:25 |
| Mic92 | @hexa:lossy.network: yes. I kept the other patches still in the staging branch that I want to test. | 08:12:29 |
| hexa | alright | 08:12:40 |
| Mic92 | Check if grpc listening socket is still up | 08:12:52 |
| Mic92 | Maybe same issue as the metrics server | 08:13:21 |
| hexa | root@sweeping-filly:~/ > netstat | grep mimas
tcp6 0 0 sweeping-filly.b.59366 mimas.nixos.org.https ESTABLISHED
tcp6 0 0 sweeping-filly.b.58946 mimas.nixos.org.https ESTABLISHED
| 08:14:48 |
| hexa | aren't they multiplexed into the same grpc stream? | 08:14:58 |
| hexa | [root@mimas:~]# ss | grep 2a01:4f8:261:1351::1
tcp ESTAB 0 0 [2a01:4f8:2220:11c8::1]:https [2a01:4f8:261:1351::1]:58946
tcp ESTAB 0 0 [2a01:4f8:2220:11c8::1]:https [2a01:4f8:261:1351::1]:59366\
| 08:15:35 |
| hexa | at least thye agree about that part | 08:15:40 |
| hexa | 58 fds open on queue-runner on elated-minsky | 08:16:21 |
| hexa | looks sensible | 08:16:24 |
| Mic92 | Also looks like tonic adds all the logic. | 08:17:52 |
| Mic92 | ss might be hard to debug because of socket activation. You can restart and I debug this next time it happen. | 08:19:34 |
| hexa | i'll do the updates in a bit anyway | 08:21:39 |
| hexa | vcunat max build duration? | 08:36:17 |
| hexa | 12h? 1d? | 08:37:24 |
| hexa | also, we should mark more jobs big-parallel, e.g. nodejs | 08:37:51 |
| Vladimír Čunát | In reply to @hexa:lossy.network
12h? 1d? Something like that, I guess | 08:41:42 |
| hexa | I'd say if your job takes 12h it should be big-parallel instead | 08:42:08 |
| hexa | but let's start with 1d to not be too disruptive for now | 08:42:21 |
