| 23 Sep 2021 |
 Rick (Mindavi) | Yeah | 15:49:28 |
 mbprtpmnr | Big problem here .... :))) ```
Sep 23 18:51:50 nixos-server kernel: Out of memory: Killed process 69971 (hydra-eval-jobs) total-vm:759624kB, anon-rss:666904kB, file-rss:0kB, shmem-rss:0kB, UID:122 pgtables:1476kB oom_score_adj:0
| 15:53:03 |
| mbprtpmnr | * Big problem here .... :)))
Sep 23 18:51:50 nixos-server kernel: Out of memory: Killed process 69971 (hydra-eval-jobs) total-vm:759624kB, anon-rss:666904kB, file-rss:0kB, shmem-rss:0kB, UID:122 pgtables:1476kB oom_score_adj:0 | 15:53:33 |
 das_j | https://downloadmoreram.com/ | 15:53:53 |
| mbprtpmnr | In reply to @janne.hess:helsinki-systems.de
https://downloadmoreram.com/ Yeah, sure... :))) | 15:54:46 |
| das_j | yeah :D but the only way of preventing this is getting more RAM | 15:55:27 |
| das_j | Nix evals are quite memory hungry | 15:55:33 |
| das_j | well also depends on the job you execute. evaluating all packages and tests takes a lot more memory than just evaluating a subset, of course | 15:56:06 |
| mbprtpmnr | I think I need to resize my VPS... | 15:57:03 |
| das_j | by a lot probably. we are evaluating our own systems and a subset of packages and we are really using our 128GB of memory | 15:58:03 |
| das_j | well, we also do parallel evals, so not doing that will probably help :D | 15:58:21 |
 ajs124 | and we give our postgres a bunch of memory | 15:58:40 |
| das_j | yeah, to make the web interface more bearable | 15:58:56 |
| mbprtpmnr | In reply to @janne.hess:helsinki-systems.de
well, we also do parallel evals, so not doing that will probably help :D parallel evals = distributed builds ...? I just ask because for me sounds the same...? | 15:59:21 |
| ajs124 | also made a bunch of errors disappear. although that probably wasn't the memory but the connection limits etc. | 15:59:21 |
| das_j | no, evals != builds | 15:59:50 |
| mbprtpmnr | well, i have 2 vps with nixos, can i make parallel eval? | 16:00:14 |
| das_j | you can't | 16:00:22 |
| das_j | it will always evaluate on one host | 16:00:27 |
| das_j | you can do multiple parallel evals on the host, but it's always limited to one host | 16:00:48 |
 cransom | evaluating nixpkgs is going to be a ginormous memory hog compared to a single system config. | 16:00:54 |
| mbprtpmnr | oooh, parallel evals on the same machine, not distributed...understood... | 16:01:26 |
| Rick (Mindavi) | In reply to @mbprtpmnr:matrix.org
Big problem here .... :)))
Sep 23 18:51:50 nixos-server kernel: Out of memory: Killed process 69971 (hydra-eval-jobs) total-vm:759624kB, anon-rss:666904kB, file-rss:0kB, shmem-rss:0kB, UID:122 pgtables:1476kB oom_score_adj:0 Thought something like that would happen :) | 16:01:48 |
| Rick (Mindavi) | Yeah, as I said, nixpkgs is not a very good idea :) | 16:02:46 |
| das_j | you can do nixpkgs, but the small variant. that's a lot more doable | 16:03:19 |
| das_j | instead of release.nix you take release-small.nix | 16:03:28 |
| Rick (Mindavi) | Maybe try patchelf for a start, I think the hydra manual also describes that | 16:03:42 |
| mbprtpmnr | In reply to @rick:matrix.ciphernetics.nl
Yeah, as I said, nixpkgs is not a very good idea :) but...with fu+ how can i eval my flake system configs, is verry small ... | 16:03:41 |
| Rick (Mindavi) | I don't know, maybe someone else can help with that | 16:04:41 |
| mbprtpmnr | I see that Hydra can build and eval also flakes... | 16:05:40 |
