| 22 Dec 2024 |
 Janne Heß | cole-h: could you take a look at this? https://pad.lassul.us/ofborg-sustainability?view#RabbitMQ-Graph and skim through whether it seems correct. I'm honestly not sure | 13:55:15 |
| Janne Heß | And do you know how/if the lowprior queues are used? | 15:26:05 |
| Janne Heß | * And do you know how/if the lowprior queues are used? There are no exchanges attached | 15:26:25 |
|  allrealmsoflife joined the room. | 15:55:10 |
 Tristan Ross | Has anyone thought about moving Ofborg to parallel evals? I think it's a great concept and looks like it's more efficient. I think it's something worth looking into. Hydra doing the same thing as well would be great. | 17:37:09 |
| Tristan Ross | * Has anyone thought about moving Ofborg to parallel/split evals? I think it's a great concept and looks like it's more efficient. I think it's something worth looking into. Hydra doing the same thing as well would be great. | 17:39:51 |
 ma27 | Hydra is doing that already with hydra-eval-jobs/nix-eva-ljobs. | 17:42:11 |
| Janne Heß | In reply to @rosscomputerguy:matrix.org
Has anyone thought about moving Ofborg to parallel/split evals? I think it's a great concept and looks like it's more efficient. I think it's something worth looking into. Hydra doing the same thing as well would be great. Both yes and no 😁 so something new has to be done by the end of the year since we won't have that absurd capacity anymore for evaluating all that. So while your solution would be worth looking into, i will frist try and leverage what the new eval github actions give us. If we can just wait for them to finish and use their results to figure out what to build, we can skip all evaluation tasks | 17:53:24 |
| Tristan Ross | Gotcha | 17:56:03 |
| Janne Heß | But that requires some engineering first. We need to change the AMQP Queue design and the webhook receiver to also handle GHA Actions | 17:56:46 |
| Janne Heß | * But that requires some engineering first. We need to change the AMQP Queue design and the webhook receiver to also handle GHA Events | 17:56:54 |
| Tristan Ross | I'm just thinking how Ofborg and Hydra can handle nixpkgs growing significantly with enabling pkgsLLVM evals. Previously, Ofborg jobs have crashed due to OOM. | 17:56:58 |
| Janne Heß | In reply to @rosscomputerguy:matrix.org
I'm just thinking how Ofborg and Hydra can handle nixpkgs growing significantly with enabling pkgsLLVM evals. Previously, Ofborg jobs have crashed due to OOM. Sounds like a GHA problem, then 🤡 (as long as everything works as I hope it does) | 17:57:55 |
| Tristan Ross | Lol yeah, it's not fun having nix-env fail with no message. It's just nix-env:. The newer split evals seems to handle pkgsLLVM a lot better. If Ofborg and Hydra can handle it just as well as GHA then I think that PR won't be blocked by infra availability then. | 17:59:29 |
| Janne Heß | For hydra, we can jump to the Hydra room so I can tell you that I have no idea how that could be solved | 17:59:35 |
| Tristan Ross | Lol | 17:59:43 |
| Janne Heß | In reply to @rosscomputerguy:matrix.org
Lol yeah, it's not fun having nix-env fail with no message. It's just nix-env:. The newer split evals seems to handle pkgsLLVM a lot better. If Ofborg and Hydra can handle it just as well as GHA then I think that PR won't be blocked by infra availability then. Yeah, ofborg should in some way be solved by the end of the year | 18:00:30 |
 Jonas Chevalier | 
Download image.png | 18:01:43 |
| Jonas Chevalier | Janne Heß: ^ | 18:01:47 |
| Jonas Chevalier | the webhook is installed on the org-level | 18:01:56 |
| Janne Heß | In reply to @zimbatm:numtide.com
the webhook is installed on the org-level Crazy | 18:02:11 |
| Jonas Chevalier | yes, I think this has some security implications | 18:02:32 |
| Jonas Chevalier | eg: it also sends private repo events | 18:02:55 |
| Janne Heß | But thank you, I will mimic that hook in a nixpkgs fork and if testing goes right, we can migrate to a repo hook | 18:03:15 |
| Tristan Ross | In reply to @janne.hess:helsinki-systems.de
Yeah, ofborg should in some way be solved by the end of the year Cool, just lmk when that stops blocking https://github.com/NixOS/nixpkgs/pull/330567. | 18:03:30 |
| Janne Heß | Also, "send me everything" sounds like a lot more load than necessary | 18:03:43 |
| Janne Heß | In reply to @rosscomputerguy:matrix.org
Cool, just lmk when that stops blocking https://github.com/NixOS/nixpkgs/pull/330567. I will try to remember, but no guarantees | 18:04:00 |
| Jonas Chevalier | Awesome, thanks for taking ownership of Ofborg! | 18:05:40 |
 cole-h | In reply to @janne.hess:helsinki-systems.de
And do you know how/if the lowprior queues are used? There are no exchanges attached They're used as a dumping ground when the queue is too full. Theoretically we could hook up a "if main queue is empty, start taking from here"
But this was my bandaid fix for slow Mac builds | 19:44:57 |
| Janne Heß | In reply to @cole-h:matrix.org
They're used as a dumping ground when the queue is too full. Theoretically we could hook up a "if main queue is empty, start taking from here"
But this was my bandaid fix for slow Mac builds Interesting! Are they implemented somewhere in RabbitMQ? Because they don't seem to be referenced in the code | 20:13:43 |
