| 26 Nov 2022 |
 @linus:schreibt.jetzt | Afaik the evaluator will only evaluate one jobset at a time | 19:31:33 |
 magic_rb |
ah ok so that may be a path forward | 19:31:45 |
|  @omlet:matrix.org left the room. | 19:33:52 |
| @linus:schreibt.jetzt | In reply to @magic_rb:matrix.redalder.org
ah ok so that may be a path forward update: I was wrong, by default it does up to 4 evals at a time. But you can configure that using the max_concurrent_evals option. | 19:59:48 |
 das_j | there are magic environment variables that may or may not fix that | 21:00:08 |
| das_j | # evaluator_initial_heap_size is basically equivalent to GC_INITIAL_HEAP_SIZE and defaults to 384MiB
# evaluator_max_memory_size is… "If our RSS exceeds the maximum, exit. The master will start a new process.". defaults to 4096 (I assume MiB)
# evaluator_workers is how many parallel evaluator workers are used. defaults to 1
# max_concurrent_evals defines how many evaluations are run concurrently. defaults to 4
extraConfig = ''
email_notification = 1
evaluator_max_memory_size = 4096
evaluator_initial_heap_size = ${toString (2 * 1024 * 1024 * 1024)}
max_output_size = ${toString (16 * 1024 * 1024 * 1024)}
max_concurrent_evals = 3
evaluator_workers = 4
| 21:00:22 |
| das_j | (all of these are guisses) | 21:00:26 |
| das_j | * (all of these are guesses) | 21:00:29 |
| das_j | * there are magic settings that may or may not fix that | 21:00:50 |
| 28 Nov 2022 |
|  h7x4 changed their profile picture. | 20:05:25 |
| 30 Nov 2022 |
|  R̴̨͕͇͍̞̮̐̅͆̌̀̉̐͋̈́̃̀͒́̎̅̚̚̚͠͝Ĕ̵̡̛͖͖̟̙̫̱͈̘̞̭͍͍͑̌̄͑̓̋̓̀̈̏̈́͊̇͊͆̉͂̏̀̃̚͘͝͝ͅͅD̶̡̢͔̱̖̮͙͉̘̺͓͍̩̮͈͍͗̃̀̏͌͘͜ͅŚ̸̬̭̯̬͙͇͓̬̩̳̤͚͓̤̩̺͉͖̉͛̓̿̎͊̿̆́̐͂̇͌̄̇̓͘ͅͅT̴̞̫̘̝͇͔̟̪̪̦͂̔̎̀̎ͅŎ̷̡̬̹̪͈̭̣͈̭̭͉̦̖̝̘̪͖͔̥̦̘̻̳Ṋ̶̛̫͈̳̘͚̜̔̋͆̅̈́͊̑͊̉̌̈́̾͑̈́̚ͅË̸̡̨̨̛͇̜̖͔͖̻̟̗̠̙͓̘̗̥͉͇̜͑͆͊͑͑̀̓͒͜͝͝ joined the room. | 04:59:15 |
|  Florian | W3F left the room. | 13:12:16 |
| 1 Dec 2022 |
 jackdk | I am setting up a Hydra instance, and I would like the jobs that produce deployment artifacts to not trigger deploys unless all checks have passed. It is easy enough to add all the checks to the buildInputs of the artifact jobs, but that causes a lot of spurious rebuilding. Is there a better way to achieve what I'm trying to do? | 00:29:14 |
 cole-h | You're looking for an "aggregate" jobset. There are some examples in Nixpkgs, I believe (look for release.nix or similarly-named files in nixos/). Also present in the Nix (the tool) flake, and probably the Hydra flake / repo as well. | 00:40:14 |
| cole-h | * You're looking for an "aggregate" job. There are some examples in Nixpkgs, I believe (look for `release.nix` or similarly-named files in `nixos/`). Also present in the Nix (the tool) flake, and probably the Hydra flake / repo as well. | 00:40:40 |
 Sandro | Is it normal that my hydra queries narinfos from itself or did I mess something up? | 10:18:55 |
| das_j | It will use any substituter, including itself if configured as substituter | 10:20:23 |
| Sandro | Do I want this or does that slow down things? I have the feeling that sometimes jobs just stale a longer time which may be related to that. | 10:21:05 |
| Sandro | Also is warning: 7 concurrent database updates; PostgreSQL may be stalled normal? postgres itself seems to be happy. | 10:21:16 |
| das_j | In reply to @sandro:supersandro.de
Do I want this or does that slow down things? I have the feeling that sometimes jobs just stale a longer time which may be related to that. I don't think it makes any sense. It pushes the required store paths to builders anyway and substituting from itself will never yield any success | 10:21:38 |
| das_j | In reply to @sandro:supersandro.de
Also is warning: 7 concurrent database updates; PostgreSQL may be stalled normal? postgres itself seems to be happy. That's normal | 10:21:43 |
| Sandro | In reply to @janne.hess:helsinki-systems.de
I don't think it makes any sense. It pushes the required store paths to builders anyway and substituting from itself will never yield any success thanks for clarifying. Then I can turn it off and maybe hydra gets going faster | 10:25:05 |
| Sandro | it feels usually a bit slow for running one concurrent build | 10:25:20 |
| Sandro | Would you recommend to increase narinfo-cache-negative-ttl when you rebuild from stage0? Would there be anything else you would suggest to set? | 11:17:37 |
| @linus:schreibt.jetzt | maybe disable use of binary caches completely if you're doing that? | 11:18:00 |
| Sandro | I am only doing it for one job of 8 | 11:19:22 |
 ajs124 | In reply to @janne.hess:helsinki-systems.de
I don't think it makes any sense. It pushes the required store paths to builders anyway and substituting from itself will never yield any success by default, the builders do substitute tho.
it's like nix-copy-close -s. | 12:35:08 |
|  hexa changed their display name from hexa to hexa (22.11 now). | 13:08:37 |
| hexa changed their display name from hexa (22.11 now) to hexa. | 14:38:26 |
| Sandro | How do I reset an error message like?
Aborted: [31;1merror:[0m a '[35;1mx86_64-linux[0m' with features {[35;1mgccarch-skylake[0m} is required to build '[35;1m/nix/store/fykll0wrfya0zqw51nsi9jbi0lasyfaa-closure-info.drv[0m', but I am a '[35;1mx86_64-linux[0m' with features {[35;1mgccarch-haswell, gccarch-ivybridge, gccarch-sandybridge, gccarch-westmere[0m} (log, raw, tail)}
Clear failed builds cache does not seem to do anything. Also why is hydra trying to build this on localhost?
My hydra machines file looks like localhost x86_64-linux - 4 5 gccarch-haswell,gccarch-ivybridge,gccarch-sandybridge,gccarch-westmere - -
| 15:00:06 |
