| 19 Feb 2025 |
 gmacon | Looks like you need enable = config.roles ? etcd; since on the non-etcd hosts, config.roles will be an attrset that does not contain an etcd key, so you can't get the value of that key. Another option would be to ensure that roles always contains every possible role, but set to false for the roles this host doesn't have. | 20:18:22 |
 pcarrier | well I thought I did exactly that through its default value | 20:19:22 |
| pcarrier | options = {
roles = mkOption {
type = types.attrsOf types.bool;
default = {
bastion = false;
spawner = false;
etcd = false;
webserver = false;
};
};
};
| 20:20:05 |
| pcarrier | that's OK, I'll switch to a single role per box | 20:22:35 |
| pcarrier | perfect :) | 20:37:06 |
| pcarrier | Redacted or Malformed Event | 20:43:00 |
| pcarrier | * I'm trying to make nix run .#deploy args... run the colmena version as imported, can't find how to access the binary to put in a pkgs.writeShellScriptBin though | 20:43:20 |
| gmacon | Ah; the assignment overrides the default value instead of being merged with it, you need types.submodule with each role having its own default to do that. | 20:48:17 |
 kevinpthorne | From the cheap seats: Why not use an array of enum type? The kubernetes pkgs does it and makes it clean to check
Roles = [ “bastion” “etc” ]; for example | 20:50:11 |
| pcarrier | maybe I should ask for what I'm actually after: how can I run colmena unstable? | 20:56:44 |
| pcarrier | * maybe I should ask for what I'm actually after: how can I run colmena unstable, ideally through instructions in my flake? | 21:02:01 |
| pcarrier | * maybe I should ask for what I'm actually after: how can I run colmena unstable, ideally through instructions in my flake? (I was hoping to expose a wrapper à la nix run .#c) | 21:02:23 |
| pcarrier |
deployTool = pkgs.writeShellScriptBin "deployTool" ''
${colmena.packages.${system}.colmena}/bin/colmena "$@"
'';
fails with:
| 21:08:47 |
| pcarrier | * deployTool = pkgs.writeShellScriptBin "deployTool" ''
${colmena.packages.${system}.colmena}/bin/colmena "$@"
'';
fails with:
error: builder for '/nix/store/h8hz7zb1vllgws7sw9fbhm1cjpbrmpww-nix-eval-jobs-2.14.0-colmena.drv' failed with exit code 1;
last 8 log lines:
> unpacking sources
> unpacking source archive /nix/store/wj7x28pyf2230189fw788mm31z8zmzm3-source
> source root is source
> patching sources
> applying patch /nix/store/v5farz7ph5afn6hw2hi0r343lhhpv8dd-nix-eval-jobs-stable.patch
> patching file src/nix-eval-jobs.cc
> Hunk #1 FAILED at 342.
> 1 out of 1 hunk FAILED -- saving rejects to file src/nix-eval-jobs.cc.rej
For full logs, run:
nix-store -l /nix/store/h8hz7zb1vllgws7sw9fbhm1cjpbrmpww-nix-eval-jobs-2.14.0-colmena.drv
| 21:09:14 |
| 20 Feb 2025 |
| gmacon | This looks like the correct approach to me, but it seems that the build is broken with the versions you have pinned. Maybe nix flake update? | 13:44:38 |
| 24 Feb 2025 |
|  Saturn changed their profile picture. | 22:37:07 |
| 25 Feb 2025 |
|  @sumner:nevarro.space left the room. | 15:11:53 |
| 26 Feb 2025 |
| pcarrier | Is it normal for node enumeration to take tens of seconds and evaluations to take ~3 mins for about 10 nodes? | 19:25:36 |
| pcarrier | * Is it normal for node enumeration to take tens of seconds and evaluations to take ~3 mins for about 10 nodes? running unstable from a few days ago | 19:26:06 |
| pcarrier | * how can I run colmena unstable, ideally through instructions in my flake? (I was hoping to expose a wrapper à la nix run .#c) | 19:26:31 |
| pcarrier | * Is it normal for node enumeration to take tens of seconds for less than 30 nodes and for evaluations to take ~3 mins for about 10 nodes? running unstable from a few days ago | 19:26:57 |
| pcarrier | * Is it normal for node enumeration to take tens of seconds for less than 30 nodes and for evaluations to take ~3 mins for about 10 nodes? most builds+deploys are really fast in comparison. running unstable from a few days ago. | 19:27:18 |
| gmacon | I've found that the heuristic for --node-eval-limit is too optimistic and so my machine swaps like crazy during evaluation if I forget to lower it. With a reasonable value, I see about 30s eval time for 10 nodes. | 19:40:17 |
| pcarrier | no swapping AFAICT (I do have 96GB of RAM) | 19:41:43 |
| gmacon | well, then... if you only have 30 nodes, that's no problem (the heuristic limit will be 190) | 20:29:00 |
| 27 Feb 2025 |
|  Unit 1721344 [polygon] (it/its, es/ihr) changed their profile picture. | 08:24:02 |
| Unit 1721344 [polygon] (it/its, es/ihr) changed their display name from Unit 1721344 [polygon] (it/she) to Unit 1721344 [polygon] (it/its, es/ihr). | 22:21:08 |
| 3 Mar 2025 |
|  recoup3948 joined the room. | 16:03:11 |
| 4 Mar 2025 |
|  Valodim joined the room. | 20:23:22 |
| 5 Mar 2025 |
|  rntpts joined the room. | 23:02:45 |
