| 20 Jun 2026 |
 hexa (signing key rotation when) | I'm sure the number will go down quickly, but | 22:41:42 |
| hexa (signing key rotation when) | new eval and now all older evals have gained the 250k queued jobs, too??? | 22:42:07 |
| hexa (signing key rotation when) | is that the weird | 22:44:49 |
| hexa (signing key rotation when) | Redacted or Malformed Event | 22:44:58 |
| hexa (signing key rotation when) |
fetchurl: add TeX historic archive mirrors
| 22:45:03 |
| hexa (signing key rotation when) | oh well, the eval also took suspiciously long | 22:46:30 |
| hexa (signing key rotation when) |
2026-06-20 23:53:46 (took 2 hours, 15 minutes, 2 seconds)
| 22:46:33 |
| hexa (signing key rotation when) | when the previous one took 23 minutes | 22:46:38 |
| hexa (signing key rotation when) | did it do stupid things? | 22:46:47 |
| hexa (signing key rotation when) | 
Download | 22:48:20 |
| hexa (signing key rotation when) | looks like it is now realizing it has already built those jobs | 22:48:30 |
| hexa (signing key rotation when) | peaks at 149 steps/min | 22:49:09 |
| hexa (signing key rotation when) | which would be a whopping 216k steps/day … bonkers | 22:49:25 |
| hexa (signing key rotation when) | Redacted or Malformed Event | 22:49:43 |
| hexa (signing key rotation when) | does that mean it needs a day to check the validity of these jobs? 🤔 | 22:50:35 |
| hexa (signing key rotation when) | that would be pretty bad | 22:50:40 |
| hexa (signing key rotation when) | 
Download | 22:55:58 |
| hexa (signing key rotation when) | that was the "before" picture | 22:56:03 |
| hexa (signing key rotation when) | so it also restarted failed builds? | 22:56:40 |
| 21 Jun 2026 |
 Mic92 | I actually thought I made that part faster, because before it was doing S3 checks for every store path and than I added an DB Check to make it faster. But I never compared it with what the old scheduler was doing. | 03:27:11 |
 K900 | https://grafana.nixos.org/goto/ffptmam7o9a80c?orgId=1 | 16:48:31 |
| K900 | If anyone wants to see the effect of the new queue runner on farm utilization | 16:48:42 |
| K900 | Definitely seems like we're getting way better low end | 16:49:40 |
 vcunat | Right now I'm more looking at the difficulties of making channels advance, as lots of non-sensical abortions are there ("dependency failed" but not really) - and restarts don't help: https://hydra.nixos.org/build/331700985 | 17:49:48 |
| vcunat | I think what does help is to make the builder build it by an explicit command and then restart the job in Hydra's UI. | 17:50:14 |
| vcunat | But we can't keep doing that all the time. | 17:50:36 |
| vcunat | (nixos-26.05 is 10 days old commit now, so I'm trying to work around such blockers manually at least.) | 17:51:23 |
| hexa (signing key rotation when) | can we merge staging-next-26.05? | 17:52:20 |
| hexa (signing key rotation when) | then we'll at least look at a fresh eval instead of maybe some regressions that we have already overcome | 17:52:41 |
| vcunat | But then I'll be manually fixing anew. | 17:53:43 |
