Merge pull request 'fix: clan machines install on machines without hardware configuration' (#2983) from fix/systemless-installs into main

Reviewed-on: https://git.clan.lol/clan/clan-core/pulls/2983
Reviewed-by: kenji <aks.kenji@protonmail.com>

checks/flake-module.nix

@@ -12,6 +12,7 @@ in
     ./flash/flake-module.nix
     ./impure/flake-module.nix
     ./installation/flake-module.nix
+    ./installation-without-system/flake-module.nix
     ./morph/flake-module.nix
     ./nixos-documentation/flake-module.nix
   ];
@@ -49,7 +50,7 @@ in
           flakeOutputs =
             lib.mapAttrs' (
               name: config: lib.nameValuePair "nixos-${name}" config.config.system.build.toplevel
-            ) self.nixosConfigurations
+            ) (lib.filterAttrs (n: _v: n != "test-install-machine-without-system") self.nixosConfigurations)
             // lib.mapAttrs' (n: lib.nameValuePair "package-${n}") self'.packages
             // lib.mapAttrs' (n: lib.nameValuePair "devShell-${n}") self'.devShells
             // lib.mapAttrs' (name: config: lib.nameValuePair "home-manager-${name}" config.activation-script) (

checks/installation-without-system/flake-module.nix

@@ -0,0 +1,211 @@
+{
+  self,
+  lib,
+  inputs,
+  ...
+}:
+{
+  # The purpose of this test is to ensure `clan machines install` works
+  # for machines that don't have a hardware config yet.
+
+  # If this test starts failing it could be due to the `facter.json` being out of date
+  # you can get a new one by adding
+  # client.fail("cat test-flake/machines/test-install-machine/facter.json >&2")
+  # to the installation test.
+  clan.machines.test-install-machine-without-system = {
+    fileSystems."/".device = lib.mkDefault "/dev/vda";
+    boot.loader.grub.device = lib.mkDefault "/dev/vda";
+
+    imports = [ self.nixosModules.test-install-machine-without-system ];
+  };
+  clan.machines.test-install-machine-with-system = {
+    facter.reportPath = "${inputs.test-fixtures}/nixos-vm-facter-json/facter.json";
+
+    fileSystems."/".device = lib.mkDefault "/dev/vda";
+    boot.loader.grub.device = lib.mkDefault "/dev/vda";
+
+    imports = [ self.nixosModules.test-install-machine-without-system ];
+  };
+  flake.nixosModules = {
+    test-install-machine-without-system =
+      { lib, modulesPath, ... }:
+      {
+        imports = [
+          (modulesPath + "/testing/test-instrumentation.nix") # we need these 2 modules always to be able to run the tests
+          (modulesPath + "/profiles/qemu-guest.nix")
+          ../lib/minify.nix
+        ];
+
+        networking.hostName = "test-install-machine";
+
+        environment.etc."install-successful".text = "ok";
+
+        boot.consoleLogLevel = lib.mkForce 100;
+        boot.kernelParams = [ "boot.shell_on_fail" ];
+
+        # disko config
+        boot.loader.grub.efiSupport = lib.mkDefault true;
+        boot.loader.grub.efiInstallAsRemovable = lib.mkDefault true;
+        clan.core.vars.settings.secretStore = "vm";
+        clan.core.vars.generators.test = {
+          files.test.neededFor = "partitioning";
+          script = ''
+            echo "notok" > $out/test
+          '';
+        };
+        disko.devices = {
+          disk = {
+            main = {
+              type = "disk";
+              device = "/dev/vda";
+
+              preCreateHook = ''
+                test -e /run/partitioning-secrets/test/test
+              '';
+
+              content = {
+                type = "gpt";
+                partitions = {
+                  boot = {
+                    size = "1M";
+                    type = "EF02"; # for grub MBR
+                    priority = 1;
+                  };
+                  ESP = {
+                    size = "512M";
+                    type = "EF00";
+                    content = {
+                      type = "filesystem";
+                      format = "vfat";
+                      mountpoint = "/boot";
+                      mountOptions = [ "umask=0077" ];
+                    };
+                  };
+                  root = {
+                    size = "100%";
+                    content = {
+                      type = "filesystem";
+                      format = "ext4";
+                      mountpoint = "/";
+                    };
+                  };
+                };
+              };
+            };
+          };
+        };
+      };
+  };
+  perSystem =
+    {
+      pkgs,
+      lib,
+      ...
+    }:
+    let
+      dependencies = [
+        self
+        self.nixosConfigurations.test-install-machine-with-system.config.system.build.toplevel
+        self.nixosConfigurations.test-install-machine-with-system.config.system.build.diskoScript
+        self.nixosConfigurations.test-install-machine-with-system.config.system.clan.deployment.file
+        pkgs.stdenv.drvPath
+        pkgs.bash.drvPath
+        pkgs.nixos-anywhere
+        pkgs.bubblewrap
+      ] ++ builtins.map (i: i.outPath) (builtins.attrValues self.inputs);
+      closureInfo = pkgs.closureInfo { rootPaths = dependencies; };
+    in
+    {
+      # On aarch64-linux, hangs on reboot with after installation:
+      # vm-test-run-test-installation> (finished: waiting for the VM to power off, in 1.97 seconds)
+      # vm-test-run-test-installation>
+      # vm-test-run-test-installation> new_machine: must succeed: cat /etc/install-successful
+      # vm-test-run-test-installation> new_machine: waiting for the VM to finish booting
+      # vm-test-run-test-installation> new_machine: starting vm
+      # vm-test-run-test-installation> new_machine: QEMU running (pid 80)
+      # vm-test-run-test-installation> new_machine: Guest root shell did not produce any data yet...
+      # vm-test-run-test-installation> new_machine:   To debug, enter the VM and run 'systemctl status backdoor.service'.
+      checks = pkgs.lib.mkIf (pkgs.stdenv.isLinux && pkgs.stdenv.hostPlatform.system != "aarch64-linux") {
+        test-installation-without-system = (import ../lib/test-base.nix) {
+          name = "test-installation-without-system";
+          nodes.target = {
+            services.openssh.enable = true;
+            virtualisation.diskImage = "./target.qcow2";
+            virtualisation.useBootLoader = true;
+
+            # virtualisation.fileSystems."/" = {
+            #   device = "/dev/disk/by-label/this-is-not-real-and-will-never-be-used";
+            #   fsType = "ext4";
+            # };
+          };
+          nodes.installer =
+            { modulesPath, ... }:
+            {
+              imports = [
+                (modulesPath + "/../tests/common/auto-format-root-device.nix")
+              ];
+              services.openssh.enable = true;
+              users.users.root.openssh.authorizedKeys.keyFiles = [ ../lib/ssh/pubkey ];
+              system.nixos.variant_id = "installer";
+              environment.systemPackages = [ pkgs.nixos-facter ];
+              virtualisation.emptyDiskImages = [ 512 ];
+              virtualisation.diskSize = 8 * 1024;
+              virtualisation.rootDevice = "/dev/vdb";
+              # both installer and target need to use the same diskImage
+              virtualisation.diskImage = "./target.qcow2";
+              nix.settings = {
+                substituters = lib.mkForce [ ];
+                hashed-mirrors = null;
+                connect-timeout = lib.mkForce 3;
+                flake-registry = pkgs.writeText "flake-registry" ''{"flakes":[],"version":2}'';
+                experimental-features = [
+                  "nix-command"
+                  "flakes"
+                ];
+              };
+              system.extraDependencies = dependencies;
+            };
+          nodes.client = {
+            environment.systemPackages = [
+              self.packages.${pkgs.system}.clan-cli
+            ] ++ self.packages.${pkgs.system}.clan-cli.runtimeDependencies;
+            environment.etc."install-closure".source = "${closureInfo}/store-paths";
+            virtualisation.memorySize = 2048;
+            nix.settings = {
+              substituters = lib.mkForce [ ];
+              hashed-mirrors = null;
+              connect-timeout = lib.mkForce 3;
+              flake-registry = pkgs.writeText "flake-registry" ''{"flakes":[],"version":2}'';
+              experimental-features = [
+                "nix-command"
+                "flakes"
+              ];
+            };
+            system.extraDependencies = dependencies;
+          };
+
+          testScript = ''
+            client.start()
+            installer.start()
+
+            client.succeed("${pkgs.coreutils}/bin/install -Dm 600 ${../lib/ssh/privkey} /root/.ssh/id_ed25519")
+            client.wait_until_succeeds("timeout 2 ssh -o StrictHostKeyChecking=accept-new -v root@installer hostname")
+            client.succeed("cp -r ${../..} test-flake && chmod -R +w test-flake")
+            client.fail("test -f test-flake/machines/test-install-machine-without-system/hardware-configuration.nix")
+            client.fail("test -f test-flake/machines/test-install-machine-without-system/facter.json")
+            client.succeed("clan machines install --debug --flake test-flake --yes test-install-machine-without-system --target-host root@installer --update-hardware-config nixos-facter >&2")
+            try:
+              installer.shutdown()
+            except BrokenPipeError:
+              # qemu has already exited
+              pass
+
+            target.state_dir = installer.state_dir
+            target.start()
+            target.wait_for_unit("multi-user.target")
+            assert(target.succeed("cat /etc/install-successful").strip() == "ok")
+          '';
+        } { inherit pkgs self; };
+      };
+    };
+}

checks/installation/flake-module.nix

@@ -175,12 +175,19 @@
             client.succeed("${pkgs.coreutils}/bin/install -Dm 600 ${../lib/ssh/privkey} /root/.ssh/id_ed25519")
             client.wait_until_succeeds("timeout 2 ssh -o StrictHostKeyChecking=accept-new -v root@installer hostname")
             client.succeed("cp -r ${../..} test-flake && chmod -R +w test-flake")
+
+            # test that we can generate hardware configurations
+            client.fail("test -f test-flake/machines/test-install-machine/facter.json")
             client.fail("test -f test-flake/machines/test-install-machine/hardware-configuration.nix")
             client.succeed("clan machines update-hardware-config --flake test-flake test-install-machine root@installer >&2")
             client.succeed("test -f test-flake/machines/test-install-machine/facter.json")
             client.succeed("clan machines update-hardware-config --backend nixos-generate-config --flake test-flake test-install-machine root@installer>&2")
             client.succeed("test -f test-flake/machines/test-install-machine/hardware-configuration.nix")
-            client.succeed("clan machines install --debug --flake ${../..} --yes test-install-machine --target-host root@installer >&2")
+
+            # but we don't use them because they're not cached
+            client.succeed("rm test-flake/machines/test-install-machine/hardware-configuration.nix test-flake/machines/test-install-machine/facter.json")
+
+            client.succeed("clan machines install --debug --flake test-flake --yes test-install-machine --target-host root@installer >&2")
             try:
               installer.shutdown()
             except BrokenPipeError:

flake.lock

@@ -75,6 +75,7 @@
         "nixpkgs": "nixpkgs",
         "sops-nix": "sops-nix",
         "systems": "systems",
+        "test-fixtures": "test-fixtures",
         "treefmt-nix": "treefmt-nix"
       }
     },
@@ -114,6 +115,31 @@
         "type": "github"
       }
     },
+    "test-fixtures": {
+      "inputs": {
+        "flake-parts": [
+          "flake-parts"
+        ],
+        "nixpkgs": [
+          "nixpkgs"
+        ]
+      },
+      "locked": {
+        "lastModified": 1741504481,
+        "narHash": "sha256-Ndx7LCbLF2sgRvbiefKEe1rgL+cYlBANVRokg27DflI=",
+        "ref": "main",
+        "rev": "3508b7ed11dad068ffc8c9f0047a5c7d54644e2c",
+        "shallow": true,
+        "type": "git",
+        "url": "https://git.clan.lol/clan/test-fixtures"
+      },
+      "original": {
+        "ref": "main",
+        "shallow": true,
+        "type": "git",
+        "url": "https://git.clan.lol/clan/test-fixtures"
+      }
+    },
     "treefmt-nix": {
       "inputs": {
         "nixpkgs": [

flake.nix

@@ -19,6 +19,10 @@
 
     treefmt-nix.url = "github:numtide/treefmt-nix";
     treefmt-nix.inputs.nixpkgs.follows = "nixpkgs";
+
+    test-fixtures.url = "git+https://git.clan.lol/clan/test-fixtures?ref=main&shallow=1";
+    test-fixtures.inputs.flake-parts.follows = "flake-parts";
+    test-fixtures.inputs.nixpkgs.follows = "nixpkgs";
   };
 
   outputs =

pkgs/clan-cli/clan_cli/machines/install.py

@@ -3,6 +3,7 @@ import logging
 import os
 import sys
 from dataclasses import dataclass, field
+from enum import Enum
 from pathlib import Path
 from tempfile import TemporaryDirectory
 
@@ -25,6 +26,12 @@ from clan_cli.vars.generate import generate_vars
 log = logging.getLogger(__name__)
 
 
+class BuildOn(Enum):
+    AUTO = "auto"
+    LOCAL = "local"
+    REMOTE = "remote"
+
+
 @dataclass
 class InstallOptions:
     machine: Machine
@@ -33,7 +40,7 @@ class InstallOptions:
     debug: bool = False
     no_reboot: bool = False
     phases: str | None = None
-    build_on_remote: bool = False
+    build_on: BuildOn | None = None
     nix_options: list[str] = field(default_factory=list)
     update_hardware_config: HardwareConfig = HardwareConfig.NONE
     password: str | None = None
@@ -122,10 +129,8 @@ def install_machine(opts: InstallOptions) -> None:
         if opts.identity_file:
             cmd += ["-i", str(opts.identity_file)]
 
-        if opts.build_on_remote:
-            cmd.extend(["--build-on", "remote"])
-        else:
-            cmd.extend(["--build-on", "auto"])
+        if opts.build_on:
+            cmd += ["--build-on", opts.build_on.value]
 
         if h.port:
             cmd += ["--ssh-port", str(h.port)]
@@ -210,7 +215,7 @@ def install_command(args: argparse.Namespace) -> None:
                 debug=args.debug,
                 no_reboot=args.no_reboot,
                 nix_options=args.option,
-                build_on_remote=args.build_on_remote,
+                build_on=BuildOn(args.build_on) if args.build_on is not None else None,
                 update_hardware_config=HardwareConfig(args.update_hardware_config),
                 password=password,
                 identity_file=args.identity_file,
@@ -241,10 +246,10 @@ def register_install_parser(parser: argparse.ArgumentParser) -> None:
         help="Host key (.ssh/known_hosts) check mode.",
     )
     parser.add_argument(
-        "--build-on-remote",
-        action="store_true",
-        help="build the NixOS configuration on the remote machine",
-        default=False,
+        "--build-on",
+        choices=[x.value for x in BuildOn],
+        default=None,
+        help="where to build the NixOS configuration",
     )
     parser.add_argument(
         "--yes",

pkgs/clan-cli/clan_cli/machines/machines.py

@@ -4,15 +4,13 @@ import logging
 from dataclasses import dataclass, field
 from functools import cached_property
 from pathlib import Path
-from time import time
 from typing import TYPE_CHECKING, Any, Literal
 
-from clan_cli.cmd import RunOpts, run_no_stdout
 from clan_cli.errors import ClanError
 from clan_cli.facts import public_modules as facts_public_modules
 from clan_cli.facts import secret_modules as facts_secret_modules
 from clan_cli.flake import Flake
-from clan_cli.nix import nix_build, nix_config, nix_eval, nix_test_store
+from clan_cli.nix import nix_config, nix_test_store
 from clan_cli.ssh.host import Host
 from clan_cli.ssh.host_key import HostKeyCheck
 from clan_cli.ssh.parse import parse_deployment_address
@@ -64,41 +62,6 @@ class Machine:
             f"nixosConfigurations.{self.name}.pkgs.hostPlatform.system"
         )
 
-    @property
-    def can_build_locally(self) -> bool:
-        config = nix_config()
-        if self.system == config["system"] or self.system in config["extra-platforms"]:
-            return True
-
-        nix_code = f"""
-            let
-              flake = builtins.getFlake("path:{self.flake.store_path}?narHash={self.flake.hash}");
-            in
-              (flake.inputs.nixpkgs.legacyPackages.{self.system}.runCommandNoCC "clan-can-build-{int(time())}" {{ }} "touch $out").drvPath
-        """
-
-        unsubstitutable_drv = json.loads(
-            run_no_stdout(
-                nix_eval(
-                    [
-                        "--expr",
-                        nix_code,
-                    ]
-                ),
-                opts=RunOpts(prefix=self.name),
-            ).stdout.strip()
-        )
-
-        try:
-            run_no_stdout(
-                nix_build([f"{unsubstitutable_drv}^*"]), opts=RunOpts(prefix=self.name)
-            )
-        except Exception as e:
-            self.debug("failed to build test derivation", exc_info=e)
-            return False
-        else:
-            return True
-
     @property
     def deployment(self) -> dict:
         if self.cached_deployment is not None: