| 4 Nov 2021 |
 cole-h | Can you do that if you create it from the Hydra web UI? | 19:54:25 |
 Mic92 | cole-h: you are right. That's not a thing. That's a weird limitation. | 19:59:54 |
 @grahamc:nixos.org | ma27 has a PR which makes it possible to add inputs to a flake. I was hesitant because I think the UI is destined to be confusing, but that hasn't stopped much in Hydra before :P | 20:32:36 |
 ma27 | I added a big warning above the form if flakes are supposed to be used. So it's ugly, but the behavior shouldn't be too unexpected :)
using this patch for a few months btw and it seems to work fine for me :) | 20:43:22 |
| Mic92 | well. I think I can live with the current approach: https://github.com/Mic92/dotfiles/blob/master/nixos/eve/modules/hydra/jobs.tf | 20:44:38 |
| ma27 | yup, that's what I planned to do in my setup at some point anyway, but I still think it makes sense to support that in the UI as well %) | 20:49:31 |
| Mic92 | wow, the github pr integration is super hacky: https://github.com/cleverca22/hydra-configs/blob/eae3caea99ca01af1b9641e375112f0036a30e70/not-os/default.nix#L44 | 21:56:26 |
 cleverca22 | yep | 21:56:55 |
| cleverca22 | hydra has a special input type, that will fetch the pull list and pass it to you as json | 21:57:07 |
| cleverca22 | your nix code is then responsible for incorporating that into the declarative jobsets | 21:57:28 |
| Mic92 | ma27: Would this PR enable github pull request with flakes? https://github.com/NixOS/hydra/pull/922/files I mean I would the same json file that one cannot use it in flakes. | 22:09:03 |
| ma27 | it works fine for gitea status, IIRC the github pr plugin uses a similar mechanism, so most likely yes :) | 22:11:10 |
| Mic92 | the github pullrequest plugin generates a file like this: https://github.com/cleverca22/hydra-configs/blob/eae3caea99ca01af1b9641e375112f0036a30e70/sample-pr.json | 22:13:45 |
| 5 Nov 2021 |
| @grahamc:nixos.org | Mic92: https://github.com/DeterminateSystems/hydra-github-jobsets-generator | 01:04:28 |
|  ReplayCoding joined the room. | 02:04:46 |
| Mic92 | grahamc: I see. That seems a bit less work. Still what do you think about a new plugin that creates these jobs based on the webhook that hydra already receives? | 08:21:05 |
| 6 Nov 2021 |
|  Tseb joined the room. | 09:17:21 |
| ma27 | btw, hydra-notify's prometheus exporter only exposes plugin metrics, correct? Wouldn't it make sense then to also add https://github.com/NixOS/nixos-org-configurations/blob/2cd3261e37e3ac4e5befaac5891df4848f29365f/delft/prometheus/hydra-queue-runner-reexporter.py to nixos/hydra | 14:03:32 |
| @grahamc:nixos.org | yes, I'd like to add a prometheus exporter to the queue runner | 14:06:19 |
| @grahamc:nixos.org | instead of having the monitoring be at arms-length where you don't actually get valid "up" metrics | 14:06:36 |
| 7 Nov 2021 |
| Mic92 set a profile picture. | 17:25:33 |
| 8 Nov 2021 |
|  freshgum joined the room. | 16:01:18 |
 fabianhjr | nixpkgs:glibc-2.34 seems stuck | 17:40:53 |
| fabianhjr | Queued 2 days ago and only 7 failures and everything else (79,659 derivations) still queued | 17:41:45 |
| @grahamc:nixos.org | I'm looking at adding visibility to RunCommand to Hydra's web UI, with:
- render command / start / stop / status under the Evaluation
- viewable logs
- ability to cancel / retry a task
since this has fairly different security properties I'm thinking it should probably be its own plugin so people don't start accidentally spew sensitive logs to the internet on a random update.
any opinions or suggestions on this?
| 19:07:20 |
| @grahamc:nixos.org | perhaps with its own daemon process so it doesn't necessarily run as hydra-notify's regular user | 19:08:05 |
| 10 Nov 2021 |
|  griff79 joined the room. | 22:47:22 |
| 11 Nov 2021 |
|  EdLin joined the room. | 07:45:43 |
| EdLin left the room. | 08:01:11 |
 samueldr | could eval "failures" (throws) be attached to the attribute name that generated them, such that when you visit e.g. the sd_image_new_kernel job and it's not part of the current eval because of an eval-time failure it's shown on the page? | 21:57:36 |
