| 27 Jun 2026 |
 hexa (signing key rotation when) | I did it after waking up today | 13:31:54 |
 emily | sure, after the software is done, given I need it for other things first anyway :) | 13:46:36 |
| emily | it's certainly something I'd be interested in working on in general, would be nice to have much better build security and provenance than we have now | 13:48:24 |
| emily | (ofc I am very grateful to all the work that goes into keeping our builders safe at present 💜) | 13:48:48 |
| emily | it's also frankly the kind of thing where FUD gives us an opportunity to turn lemons into lemonade | 13:49:58 |
 vcunat | BTW, does anyone understand why this graph is showing so weird since the new queue-runner?
grafana.nixos.org/d/MJw9PcAiz/hydra-jobs?viewPanel=panel-2
The numbers shown in the tooltip doesn't match what's on the Y-axis. Completely off 🤯 | 13:50:56 |
| vcunat | * BTW, does anyone understand why this graph is showing so weird since the new queue-runner?
https://grafana.nixos.org/d/MJw9PcAiz/hydra-jobs?viewPanel=panel-2
The numbers shown in the tooltip doesn't match what's on the Y-axis. Completely off 🤯 | 13:51:07 |
| vcunat | Ah, values stacking to each other in some surprising ways. | 13:52:28 |
| vcunat | * Ah, values stacking on top of each other in some surprising ways. | 13:52:39 |
| vcunat | hexa (signing key rotation when): or did you enable stacking in this graph on purpose? (or someone else) Maybe it's confusing just my mind. | 13:55:31 |
| hexa (signing key rotation when) | oh, yeah. it's stacked for better visibility, sorry | 13:56:01 |
| hexa (signing key rotation when) | we can disable it, if it's too confusing | 13:56:15 |
| hexa (signing key rotation when) | check again | 13:56:55 |
| vcunat | I figured out how to change it. But I think the main block was me not realizing that these are "stacked area" charts, after being used to (overlaping) lines. | 13:59:14 |
| vcunat | * I figured out how to change it. But I think the main block was me not realizing that these are "stacked area" charts, after being used to (overlapping) lines. | 13:59:17 |
| vcunat | Certainly not a big deal, mostly a matter of getting used to, I expect. | 13:59:48 |
| hexa (signing key rotation when) | I think we should go with what is easier to parse | 14:00:11 |
| vcunat | The two graphs stack in different order. | 14:00:23 |
| hexa (signing key rotation when) | yeah, syncing graphs is annoying | 14:00:34 |
| hexa (signing key rotation when) | especially with coloring | 14:00:39 |
| vcunat | But at least colors seem consistent, magically. | 14:00:43 |
| vcunat | I'm slightly missing the big-parallel visuals. | 14:01:18 |
| vcunat | Since we still have separate *-linux builders for those. | 14:01:47 |
| vcunat | It does happen that either is over- or under-loaded. | 14:02:04 |
| hexa (signing key rotation when) | yeah, I wonder if these are not provided as labels any more | 14:02:06 |
| hexa (signing key rotation when) | I also contemplated making big-parallel a supported and not a mandatory feature | 14:02:26 |
| vcunat | Instead we got the -v1...-v4 labels which we do not utilize. | 14:02:48 |
| hexa (signing key rotation when) | so the big-parallel builders become a lucky fast path for random builds, when no big-parallel builds are around | 14:02:54 |
| emily | wouldn't that risk using up the slots for normal builds (which are way more plentiful) and having no big-parallel slots? | 14:03:24 |
| hexa (signing key rotation when) | yeah, but if builds take long with many more cores, aren't they suited for big-parallel builders anyway? | 14:03:45 |
