| 27 Nov 2023 |
 cole-h | In reply to @artturin:matrix.org
prs from 2+ hours haven't been evalled Yep, unfortunately that’s because Nixpkgs still takes on the order of ~10+ minutes for each evaluation. We used to be able to get by with 5-7 machines, but nowadays even 9 is too slow. I think I investigated this slowdown before and posted here, but I don’t remember | 13:37:58 |
 Lily Foster | In reply to @cole-h:matrix.org
Yep, unfortunately that’s because Nixpkgs still takes on the order of ~10+ minutes for each evaluation. We used to be able to get by with 5-7 machines, but nowadays even 9 is too slow. I think I investigated this slowdown before and posted here, but I don’t remember There's an ambitious PR at https://github.com/NixOS/nixpkgs/pull/269403 which we could possibly replace our outpath calculation with to get a bit improvement. I haven't looked at the specifics of it though | 13:40:47 |
| cole-h | Hm! | 13:41:24 |
| Lily Foster | In reply to @cole-h:matrix.org
Yep, unfortunately that’s because Nixpkgs still takes on the order of ~10+ minutes for each evaluation. We used to be able to get by with 5-7 machines, but nowadays even 9 is too slow. I think I investigated this slowdown before and posted here, but I don’t remember * There's an ambitious PR at https://github.com/NixOS/nixpkgs/pull/269403 which we could possibly replace our outpath calculation with to get a big improvement. I haven't looked at the specifics of it though | 13:41:34 |
| Lily Foster | I was hopefully going to give that PR a close look sometime today (not that I would know what I'm doing, but at least to have some understanding of the technical changes) | 13:43:40 |
 Artturin | A few people said that the eval ram usage went up in the last few days | 15:18:01 |
 teto | Artturin brought to my attention hte problem of nix eval going up ( I witnessed it myself before). Do we know why that is (due to release workload )? Is there a ticket to track this down ? Does ofborg use maintainers/scripts/rebuild-amount.sh ? I wonder if that's what ofborg used ? The script could use some polish https://github.com/NixOS/nixpkgs/issues/269616, I sometimes see fancy benchmarks on eval metrics, what do people use ? | 22:48:57 |
| Lily Foster | In reply to @keiichi:matrix.org
Artturin brought to my attention hte problem of nix eval going up ( I witnessed it myself before). Do we know why that is (due to release workload )? Is there a ticket to track this down ? Does ofborg use maintainers/scripts/rebuild-amount.sh ? I wonder if that's what ofborg used ? The script could use some polish https://github.com/NixOS/nixpkgs/issues/269616, I sometimes see fancy benchmarks on eval metrics, what do people use ? It does not use that script. It uses its own outpaths.nix that it copies into the nixpkgs checkout and then evaluates. See above in the scrollback for a PR that could really help what ofborg does though | 22:52:48 |
| teto | saw that one a few days ago, looked exciting | 22:56:49 |
| 29 Nov 2023 |
 piegames | If I have a package set, and add a tests attribute to it, and modify multiple packages in the package set in a single commit with an appropriate title, will OfBorg then build that test? | 11:55:57 |
| 1 Dec 2023 |
|  moritz.hedtke set their display name to Moritz Hedtke. | 11:08:13 |
| 8 Dec 2023 |
 K900 | Are the evaluators overloaded again? | 12:00:54 |
| K900 | My kernels PR this morning got all green pretty quick | 12:01:02 |
| K900 | But https://github.com/NixOS/nixpkgs/pull/272870 is stuck | 12:01:04 |
| Lily Foster | In reply to @k900:0upti.me
But https://github.com/NixOS/nixpkgs/pull/272870 is stuck It seems to be progressing? What was stuck? | 12:08:41 |
| K900 | Oh yeah it just got picked up I guess | 12:08:59 |
| 15 Dec 2023 |
|  fgaz joined the room. | 15:46:42 |
| 16 Dec 2023 |
|  kenji joined the room. | 13:23:31 |
| 19 Dec 2023 |
 @infinisil:matrix.org | The eval queue is rising.. | 21:13:11 |
 hexa | Redacted or Malformed Event | 22:11:21 |
| hexa | Redacted or Malformed Event | 22:11:22 |
| hexa | Redacted or Malformed Event | 22:11:37 |
| hexa | Redacted or Malformed Event | 22:13:02 |
| hexa | looks like there was a blip, but we still have 9 evaluators | 22:14:36 |
| cole-h | In reply to @infinisil:matrix.org
The eval queue is rising.. A side effect of evals taking 10m+ a pop 😅 | 22:45:30 |
| @infinisil:matrix.org | https://github.com/NixOS/nixpkgs/pull/269403 will be able to help, but it needs to get done first 🙃 | 22:46:43 |
| cole-h | I've got my eye on it. I also tried out nix-eval-jobs manually with outpaths.nix at one point but it took more than 24 hours (at which point I killed it) | 22:48:26 |
| cole-h | Maybe I was holding it wrong | 22:48:37 |
| Lily Foster | In reply to @cole-h:matrix.org
Maybe I was holding it wrong probably, as i did not experience that. i've experimented with it as well but the semantics were a little different in result. i intend to more thoroughly at least attempt to make a PoC with ofborg using it soon for comparison | 23:03:42 |
| 20 Dec 2023 |
 @janik0:matrix.org | Did ofborg die? It feels like the eval for most prs is stalled, but I don't really know where to look. | 19:42:33 |
