| 26 Dec 2025 |
 Janne | That would be great | 16:19:33 |
|  aktaboot changed their profile picture. | 19:25:55 |
 mynacol | For non-foundation use you can always just throw https://github.com/Defelo/nixpkgs-review-gha at a PR and also run nixos tests, just like with local nixpkgs-review. If individuals use it, this should be very much fine for Github, but is not feasible for running automatically on PRs by the NixOS org (except they get enterprise plan sponsored, I hear). | 22:32:33 |
 gabyx | ah nice: but not sure I quite understand that tool, this repo is a GHA setup which runs nixpkgs-review.
So you can trigger this repo's action (how?) and it will pull the PR, and build the attribute on all platforms? | 22:38:36 |
| mynacol | Read the readme :). It explains how you can run it. And just try it.
I found it a month ago and think it's ingenious. | 22:43:32 |
| gabyx | Jeah I probably got the answer. | 22:43:52 |
| gabyx | Its certainly good, also ingenious would be: a local tool which runs on qemu VMs, such that you dont need to debug stupid GHA yaml etc... | 22:50:11 |
| 27 Dec 2025 |
 Arian | We at work have a pretty neat self-hosted gha cluster that spawns up isolated NixOS vms for each job in seconds. But it is AWS based so it'll cost money :').
But for us it's a major money saver. We undercut GitHub enterprise pricing massively and we get fancy servers that are actually ... Fast | 12:27:49 |
|  jasper changed their display name from jappie @ 39c3 to jasper @ 39c3 ☎️ 62749. | 13:30:50 |
| 28 Dec 2025 |
| gabyx | Any body has a similar setup with Gitlab Runners? | 15:13:03 |
 hexa (signing key rotation when) | that will end you up in docker | 15:26:26 |
| hexa (signing key rotation when) | don't think gitlab supports anything else | 15:26:32 |
| hexa (signing key rotation when) | but yeah, you can have autoscaling groups iirc | 15:26:52 |
| gabyx | Shell executors, but jeah you are right. | 15:26:55 |
| gabyx | How do you spawn nixos vms, whats the mechanics here, is it the GHA runner which can do that? Or what ingredients is needed to make that work? kubernetes no right> that would end you up in containers as well? | 15:28:47 |
| Arian | Listen to GitHub webhooks and then do ec2:RunInstances call | 17:29:58 |
| Arian | That spawns a nixos VM with gha runner | 17:30:24 |
 vcunat | self-hosted GitLab certainly allows to have other stuff, probably via the shell executor. We have LXC that way. | 18:38:36 |
| 29 Dec 2025 |
|  rcouto joined the room. | 23:38:57 |
| 30 Dec 2025 |
|  amadaluzia[tde] changed their profile picture. | 12:58:11 |
| jasper changed their display name from jasper @ 39c3 ☎️ 62749 to jasper. | 23:38:40 |
| jasper | 23:40:10 |
| 1 Jan 2026 |
|  @adam:robins.wtf left the room. | 17:00:08 |
| 3 Jan 2026 |
| Arian | don't have access to my Yubikey atm to check. the AWS account Hand-waivy how many objects do we upload to the cache per month? Context: Was thinking of setting up an SNS topic of EventBridge bus to allow people to subscribe to upload events through webhooks e.g. for people who want to set up downstream caches and keep in sync with cache.nixos.org . Wanted to make a quick napkin calculation of the cost.
| 21:46:34 |
| hexa (signing key rotation when) | Redacted or Malformed Event | 22:15:06 |
| hexa (signing key rotation when) | Download NumberOfObjects-2025_01_08_22_00_00-2026_01_02_22_00_00-UTC.csv | 22:15:53 |
| hexa (signing key rotation when) | cc Arian | 22:15:56 |
| Arian | monthly cost for EventBridge would be $11 . Then each webhook destination would cost $2 per month
cc flokli
| 23:26:23 |
| Arian | * monthly cost for EventBridge would be $11 as we publish ~11m objects per month . Then each webhook destination would cost $2 per month
cc flokli
| 23:26:43 |
| Arian | Probably slightly cheaper if if we only notify on narinfo files. This is peanuts. Very feasible | 23:27:09 |
