I'm sketching an Architecture for a service in an openstack cluster, which consists of multiple machines (and there will be new machines every now and than). I would like to have one machine that shares it's /nix/store with all the other instanzes in the cluster. That is not a problem. I have a mono repository and would like to build all machines (not deploy) and nix-copy-closure the resulting closures to this machine. Meaning the build server (which is not the cache server) is in a loop while true; git pull ; colmena build ; nix-copy-closure ???? private_cache_server How to find all closures that need to be copied?

I want to use colmena. Because we would not deploy all machines instantly.

I just want to enrich the cache.

All machines have a very similar base setup, and updates should be pulled from the cache, to avoid uploading 50 times the same closures from my office to the data center.

In reply to @palo:xaos.space

All machines have a very similar base setup, and updates should be pulled from the cache, to avoid uploading 50 times the same closures from my office to the data center.

(I answered before you asked :D)

(and use the buildOnTarget option)

Or I push always all closures on a differnt build server.

Alright, next try to get this formatted correctly...

I recently updated to colmena 0.4.0 and cannot deploy from my Mac anymore:

~/.co/nixpkgs main ▷ nix develop -c colmena apply --on elara
[INFO ] Using flake: git+file:///Users/markuscisler/.config/nixpkgs
[INFO ] Enumerating nodes...
error:
       … while calling the 'attrNames' builtin

         at «string»:1:209:

            1| with builtins; let assets = getFlake "path:/tmp/nix-shell.9futSw/colmena-assets-0CsIT0?lastModified=1687506045&narHash=sha256-6Mfqk77CvYUqLBAWSiOuxaeT%2Bi2EF06vZHsFgILLuFU%3D"; hive = assets.processFlake; in attrNames hive.nodes
             |                                                                                                                                                                                                                 ^

       … while evaluating the attribute 'nodes'

         at /nix/store/vfamvq9dh5qiz0b3f6pbkkf77kqkmpr8-source/eval.nix:186:3:

          185|
          186|   nodes = listToAttrs (map (name: { inherit name; value = evalNode name (configsFor name); }) nodeNames);
             |   ^
          187|   toplevel =         lib.mapAttrs (_: v: v.config.system.build.toplevel) nodes;

       (stack trace truncated; use '--show-trace' to show the full trace)

       error: attribute 'currentSystem' missing

       at /nix/store/b90clwdlidf3s95m73jsby3ikfa56h91-source/pkgs/top-level/impure.nix:17:43:

           16|   # (build, in GNU Autotools parlance) platform.
           17|   localSystem ? { system = args.system or builtins.currentSystem; }
             |                                           ^
           18|
[ERROR] -----
[ERROR] Operation failed with error: Child process exited with error code: 1
Hint: Backtrace available - Use `RUST_BACKTRACE=1` environment variable to display a backtrace

The relevant parts from my flake.nix are:

      colmena = {
        meta = {
          specialArgs = { inherit inputs; };

          nixpkgs = import inputs.nixpkgs-stable { };
          nodeNixpkgs = {
            kerala = import inputs.nixpkgs-2305 { };
          };
        };
        defaults = import ./nixos/remote-defaults;

        elara = { name, nodes, ... }: {
          deployment = {
            buildOnTarget = true;
            #targetHost = "192.168.10.2";
            targetHost = "100.66.68.66";
          };
          nixpkgs.system = "x86_64-linux";
          imports = [ ./machines/elara.nix ];
        };
      };

Any hints on what is going on here?