| 22 May 2025 |
 Jeremy Fleischman (jfly) | In reply to @joerg:thalheim.io
What is the best way to reach garbas these days? I met him at Planet Nix. He works at Flox, and they have a community Slack: https://go.flox.dev/slack | 16:01:05 |
 Domen Kožar | maybe it was namespace doing an image rollout | 17:24:01 |
 lassulus | Mic92 hexa (signing key rotation when) Jeremy Fleischman (jfly) : can one of you take care of https://github.com/NixOS/infra/pull/691 ? we would need that for the nixcon ticket sale to go live | 17:39:26 |
 K900 | Going to start a new unstable-small to get Hydra building the new kernels | 18:31:31 |
| K900 | Is Hydra dead | 20:12:44 |
 Tristan Ross | Seems like it, I can't connect to it | 20:13:12 |
| Tristan Ross | Oh, it loaded after like a minute and a half. | 20:13:58 |
 adamcstephens | It was being very slow for me earlier | 20:19:12 |
 leona | that very much might have been me. i tried to use the old script to find broken builds, but will now use another approach. in it's current way the script just breaks hydra | 20:21:20 |
 hexa | I hope you merged the 5.10 fix | 20:30:22 |
| K900 | I did | 20:30:29 |
 @sudoforge:matrix.org | shhh, the beast is sleeping. don't wake it | 20:41:26 |
| leona | Could someone please change stableBranch to true for the 25.05 hydra jobsets? :) | 21:30:45 |
| hexa | done | 22:04:54 |
| hexa | that also triggered an eval | 22:05:20 |
| leona | thank you! | 22:06:55 |
| 23 May 2025 |
 Vladimír Čunát | Huh, I didn't realize that the Hydra's queue runner is dumb enough to attempt building exactly the same derivation twice in parallel. Right now:
https://hydra.nixos.org/build/298195286
https://hydra.nixos.org/build/298186259
(note the timestamps etc.) The only luck here is that they got scheduled to the same builder machine and Nix on the machine does deduplicate that. | 08:08:20 |
| Vladimír Čunát | That might be happening a lot now, as 25.05 does not differ much from unstable/master yet - and quite commonly nixpkgs:trunk has significant job overlap nixos:trunk-combined. | 08:11:23 |
| Vladimír Čunát | * That might be happening a lot now, as 25.05 does not differ much from unstable/master yet - and quite commonly nixpkgs:trunk has significant job overlap with nixos:trunk-combined. | 08:11:32 |
| Vladimír Čunát | Here a darwin case where it scheduled to different machines and really duplicates:
https://hydra.nixos.org/build/298195288
https://hydra.nixos.org/build/298204485 | 08:13:41 |
| Vladimír Čunát | (floorp being a nice test case right now) | 08:13:58 |
| Vladimír Čunát | * Here a darwin case where it scheduled to different machines and really duplicates the whole work:
https://hydra.nixos.org/build/298195288
https://hydra.nixos.org/build/298204485 | 08:15:14 |
|  mdaniels5757 joined the room. | 17:00:24 |
 Alyssa Ross | What's the current Hydra output size limit? | 17:16:25 |
| hexa | max_output_size = 4121225472 # 3 << 30 + 900000000 = 3 GiB + 0.9 GB
| 17:17:05 |
| hexa | from nixos/infra in build/hydra.nix | 17:17:11 |
| Alyssa Ross | wonderful | 17:17:22 |
| Alyssa Ross | I remembered it as 2G | 17:17:27 |
| hexa | it grew with the graphical ISO over the years | 17:17:44 |
| hexa | and now with the combined ISO we've reached that limit | 17:18:14 |
