| 10 Oct 2025 |
 Vladimír Čunát | uep: I don't understand what you mean. Something seems wrong in particular? | 05:59:46 |
 uep | those grey numbers for unfinished jobs.. | 06:00:25 |
| uep | they were finished | 06:00:33 |
| Vladimír Čunát | So this is trunk-combined showing and someone did restart-all-failed retrospectively. | 06:00:34 |
| uep | yeah, i assume a lot of them are the same job | 06:01:00 |
| Vladimír Čunát | nixos-unstable is c9b6fb798541 for several hours at least. | 06:01:05 |
| uep | i know.. i've been running c9b6 since yesterday | 06:02:09 |
| Vladimír Čunát | The channel scripts have extra code to avoid the possibility of channels going backwards - exactly for similar situations. | 06:02:50 |
| uep | so, i assumed hydra had been restarted and forgotten some of those jobs since clearly they had already run | 06:03:19 |
| uep | but explicit restarts makes sense too | 06:03:46 |
| Vladimír Čunát | This doesn't happen automatically. | 06:04:06 |
| uep | * but explicit job restarts makes sense too | 06:04:10 |
| Vladimír Čunát | But I don't know who did it. No audits for that. | 06:04:16 |
| uep | yeah, to be clear, i wasn't worried about the channel going backwards, what i meant about going backwards was just the build count progress. | 06:07:54 |
| uep | but if that's because (say) there are flaky python tests again and just rerunning lets more things land in cache, ok | 06:08:44 |
| Vladimír Čunát | Hydra doesn't allow you to restart succeeded builds. | 06:09:09 |
| Vladimír Čunát | So this is a relatively common situation for us. I was just trying to find out if there's something more anomalous, but it doesn't sound so. | 06:11:52 |
| uep | yeah, i thought it was hydra being flaky rather than deliberately retrying failed, possibly flaky, builds for evals that were already complete.. because it's been like this all week. | 06:13:31 |
| uep | never mind, thanks for the clarification | 06:13:48 |
