Anyone with experience using NixOS to create a Lemmy instance, please share any caveats and troubleshooting tips you have.

  • yiliu@informis.land
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 year ago

    FWIW, I’m running NixOS but gave up on running the Lemmy module. I gave up when I realized that Lemmy seems to need superuser access to the Postgresql server, to install plugins or whatever.

    So instead, I used Arion to make a docker-compose image, running in podman. Works great so far.

  • neoney@lemmy.neoney.dev
    link
    fedilink
    English
    arrow-up
    1
    ·
    edit-2
    1 year ago

    Oh hello, my instance is on NixOS!
    There’s a NixOS module for lemmy, but it’s kind of broken right now and in the middle of a PR. If you want, you can replace the current nix package with the one from the fork, and successfully deploy it, however, if I were you, I’d just wait a week or so.

    The configuration right now requires a bunch of hacks, and even doing some non-reproducible stuff.

    • ericjmorey@lemmy.worldOP
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      That PR link returns a 404.

      This is my way of learning Nix/NixOS so I’d rather give it a go before the PR is merged.

          • neoney@lemmy.neoney.dev
            link
            fedilink
            English
            arrow-up
            1
            ·
            edit-2
            1 year ago

            Okay, I’ve actually figured simplifying it out, it’s not that bad. Let me share my config:

            First, since the PR isn’t in nixos-unstable yet, I’m adding the fork as a flake input

            inputs = {
              nixpkgs-lemmy.url = "github:CobaltCause/nixpkgs/lemmy-module-improvements";
            };
            

            then, in my system configuration, I add this:

            # Not sure if this is required, maybe caddy auto-allows it
            networking.firewall.interfaces.eth0.allowedTCPPorts = [443 80];
            
            # Override the lemmy module with the one from the PR
            disabledModules = ["services/web-apps/lemmy.nix"];
            imports = [
              "${inputs.nixpkgs-lemmy}/nixos/modules/services/web-apps/lemmy.nix"
            ];
            
            services.lemmy = {
              database.createLocally = true;
              database.uri = "postgres:///lemmy?host=/run/postgresql&user=lemmy";
              enable = true;
              settings = {
                hostname = "<YOUR_HOSTNAME>";
              };
              caddy.enable = true;
            };
            

            and, that’s it!
            However, I’m not sure if it will cleanly deploy, as you might get an error while starting.
            If so, please check postgresql logs sudo journalctl -fu postgresql. The error will most likely be something like this:
            […] [10289] ERROR: permission denied: "RI_ConstraintTrigger_a_16639" is a system trigger […]

            If that happens, you need to manually run the migration until the fix is merged into Lemmy. Here’s how I did it:

            1. sudo su - postgres
            2. psql -v ON_ERROR_STOP=1 lemmy postgres
            3. (in psql) SET ROLE lemmy;
            4. Paste the SQL code from here: https://github.com/LemmyNet/lemmy/issues/2784#issuecomment-1578337686

            After that’s done, you can exit the postgres CLI by typing \q, exit the postgres user and just simply sudo systemctl restart lemmy which should start properly now, and be accessible to the outside network.
            Open it and it will give you the initial setup screen. Good luck!