| 20 Jun 2026 |
 K900 | I wonder if less but bigger jobs will be better now | 10:03:20 |
 hexa | Redacted or Malformed Event | 10:03:24 |
| hexa | https://github.com/NixOS/infra/pull/1093 | 10:09:46 |
| hexa | 6x20C (down from 10x20C) for aarch64-linux | 10:10:04 |
| hexa | 6x24C (up from 4x24C) for x86_64-linux | 10:10:17 |
| hexa | that's ~1.5 times the core count of the machine | 10:10:47 |
| K900 | Now that we have pressure info, we should really start taking that into account when scheduling | 10:10:49 |
| hexa | we kinda do | 10:11:01 |
| K900 | Would make life a lot easier I think | 10:11:03 |
| hexa | in that we dont' schedule new jobs beyond 70% cpu pressure | 10:11:20 |
| K900 | I mean like, instead of relying on big-parallel and friends | 10:11:30 |
| K900 | We can probably have better heuristics | 10:11:36 |
| hexa | oh, you mean mixed load? | 10:11:38 |
| K900 | I think mixed load is what organically falls out if we start tracking per-job costs | 10:12:04 |
 raitobezarius | Redacted or Malformed Event | 10:13:49 |
| hexa | yes, it would. | 10:14:49 |
| hexa | the scheduler supports custom scheduling functions | 10:15:01 |
| hexa | https://github.com/NixOS/hydra/blob/master/nixos-modules/queue-runner-module.nix#L47-L63 | 10:15:29 |
|  Harshal joined the room. | 12:36:12 |
| K900 | Did someone restart Hydra again | 13:08:41 |
| K900 | We will never get a kernel at this rate | 13:08:43 |
| hexa | ● hydra-queue-runner-dev.service - Hydra Queue Runner main service
Loaded: loaded (/etc/systemd/system/hydra-queue-runner-dev.service; enabled; preset: ignored)
Active: active (running) since Sat 2026-06-20 08:35:31 UTC; 4h 33min ago
| 13:09:06 |
| hexa | the builds are still running | 13:11:47 |
| hexa | the queue-runner is still doing | 13:12:00 |
| K900 | https://hydra.nixos.org/build/331682971#tabs-buildsteps | 13:12:57 |
| hexa | yeah, it did release-26.05 6.18.36 first | 13:13:19 |
| hexa | together with the browsers | 13:13:26 |
| hexa | once that was done it had a free slot for the unstable-small 6.1836 | 13:14:26 |
| K900 | The 1h40 one was running ~3 hours ago | 13:14:39 |
| hexa | hm | 13:14:53 |
