Enable VLAB in CI (#20)

.github/workflows/ci.yaml

@@ -105,13 +105,72 @@ jobs:
         with:
           limit-access-to-actor: true
 
+  vlab:
+    runs-on: ["vlab"]
+    # needs:
+    #   - cache-tools
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup Go
+        uses: actions/setup-go@v4
+        with:
+          go-version: stable
+          cache: true
+
+      - name: Setup git for private Go modules
+        env:
+          GH_ACCESS_TOKEN: ${{ secrets.GH_ACCESS_TOKEN }}
+        run: git config --global url.https://[email protected]/.insteadOf https://github.com/
+
+      - name: Build hhfab
+        run: |
+          make hhfab
+
+      - name: Login to ghcr.io
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Login to lab registry
+        uses: docker/login-action@v3
+        with:
+          registry: m.l.hhdev.io:31000
+          username: lab
+          password: ${{ secrets.LAB_REGISTRY_TOKEN }}
+
+      - name: hhfab init and build
+        run: |
+          export XDG_CONFIG_HOME=$(pwd)/.config
+          export HHFAB_BASE_REPO=m.l.hhdev.io:31000/githedgehog
+
+          bin/hhfab init -p vlab -v
+          bin/hhfab build -v
+          ls -lah .hhfab
+
+          bin/hhfab vlab up -v --dry-run
+          sudo bin/hhfab vlab up -v --run-complete hack/vlab-test.sh
+      - name: Setup tmate session for debug
+        if: ${{ failure() && github.event_name == 'workflow_dispatch' && inputs.debug_enabled }}
+        uses: mxschmitt/action-tmate@v3
+        timeout-minutes: 30
+        with:
+          limit-access-to-actor: true
+
   publish:
     # runs-on: ["lab", "simple", "jammy"]
     runs-on: ubuntu-latest
     if: startsWith(github.event.ref, 'refs/tags/') && github.event_name == 'push'
     needs:
       # - cache-tools
       - build
+      - vlab
 
     steps:
       - name: Checkout repository

cmd/hhfab/main.go

@@ -267,6 +267,14 @@ func main() {
 								Name:  "compact",
 								Usage: "run more lightweight vms, small risks",
 							},
+							&cli.BoolFlag{
+								Name:  "install-complete",
+								Usage: "run installer and complete vlab (for testing)",
+							},
+							&cli.StringFlag{
+								Name:  "run-complete",
+								Usage: "run installer, run provided script and than complete vlab (for testing)",
+							},
 						},
 						Before: func(ctx *cli.Context) error {
 							return setupLogger(verbose, brief)
@@ -284,7 +292,8 @@ func main() {
 
 							killStaleVMs := cCtx.Bool("kill-stale-vms")
 
-							return errors.Wrap(svc.StartServer(killStaleVMs, cCtx.Bool("compact")), "error starting vlab")
+							return errors.Wrap(svc.StartServer(killStaleVMs, cCtx.Bool("compact"),
+								cCtx.Bool("install-complete"), cCtx.String("run-complete")), "error starting vlab")
 						},
 					},
 					{

hack/vlab-test.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+set -ex
+
+function control {
+    bin/hhfab vlab ssh --vm control-1 PATH=/opt/bin "$@"
+}
+
+control kubectl fabric vpc create --name vpc-1 --subnet 10.90.0.1/24
+control kubectl fabric vpc attach --vpc vpc-1 --conn server-1--mclag--switch-1--switch-2
+control kubectl fabric vpc attach --vpc vpc-1 --conn server-2--mclag--switch-1--switch-2
+
+control kubectl get agent -o wide

pkg/fab/vlab/service.go

@@ -48,6 +48,8 @@ type Service struct {
 type Config struct {
 	DryRun            bool
 	Compact           bool
+	InstallComplete   bool
+	RunComplete       string
 	Basedir           string
 	Wiring            *wiring.Data
 	ControlIgnition   string
@@ -173,8 +175,10 @@ func (svc *Service) checkForStaleVMs(ctx context.Context, killStaleVMs bool) err
 	return nil
 }
 
-func (svc *Service) StartServer(killStaleVMs bool, compact bool) error {
+func (svc *Service) StartServer(killStaleVMs bool, compact bool, installComplete bool, runComplete string) error {
 	svc.cfg.Compact = compact
+	svc.cfg.InstallComplete = installComplete
+	svc.cfg.RunComplete = runComplete
 
 	for _, cmd := range RequiredCommands {
 		_, err := exec.LookPath(cmd)

pkg/fab/vlab/vm.go

@@ -85,6 +85,10 @@ func (vm *VM) Run(ctx context.Context, eg *errgroup.Group) {
 
 func (vm *VM) RunVM(ctx context.Context) func() error {
 	return func() error {
+		if vm.Cfg.DryRun {
+			return nil
+		}
+
 		// <-vm.TPMReady
 		time.Sleep(3 * time.Second) // TODO please, no!
 
@@ -171,10 +175,6 @@ func (vm *VM) RunVM(ctx context.Context) func() error {
 			return errors.Errorf("error running control vm: no ssh port found")
 		}
 
-		if vm.Cfg.DryRun {
-			return nil
-		}
-
 		return errors.Wrapf(execCmd(ctx, vm.Basedir, true, "qemu-system-x86_64", []string{}, args...), "error running vm")
 	}
 }
@@ -185,12 +185,12 @@ func (vm *VM) RunInstall(ctx context.Context) func() error {
 			return nil
 		}
 
-		if vm.Cfg.DryRun {
+		if vm.Installed.Is() {
+			slog.Debug("Control node is already installed", "name", vm.Name)
 			return nil
 		}
 
-		if vm.Installed.Is() {
-			slog.Debug("Control node is already installed", "name", vm.Name)
+		if vm.Cfg.DryRun {
 			return nil
 		}
 
@@ -256,6 +256,27 @@ func (vm *VM) RunInstall(ctx context.Context) func() error {
 			return errors.Wrapf(err, "error marking control node as installed")
 		}
 
+		if vm.Cfg.InstallComplete {
+			// TODO do graceful shutdown
+			slog.Info("Exiting after control node installation as requested")
+			os.Exit(0)
+		}
+
+		if vm.Cfg.RunComplete != "" {
+			slog.Info("Running script after control node installation as requested")
+
+			err = execCmd(ctx, "", false, vm.Cfg.RunComplete, []string{
+				"KUBECONFIG=" + filepath.Join(vm.Cfg.Basedir, "kubeconfig.yaml"),
+			})
+			if err != nil {
+				return errors.Wrapf(err, "error running script after control node installation")
+			}
+
+			// TODO do graceful shutdown
+			slog.Info("Exiting after script after control node installation as requested")
+			os.Exit(0)
+		}
+
 		return nil
 	}
 
@@ -271,6 +292,10 @@ func (vm *VM) RunInstall(ctx context.Context) func() error {
 
 func (vm *VM) RunTPM(ctx context.Context) func() error {
 	return func() error {
+		if vm.Cfg.DryRun {
+			return nil
+		}
+
 		err := execCmd(ctx, vm.Basedir, true, "swtpm", []string{}, "socket", "--tpm2", "--tpmstate", "dir=tpm",
 			"--ctrl", "type=unixio,path=tpm.sock.ctrl", "--server", "type=unixio,path=tpm.sock", "--pid", "file=tpm.pid",
 			"--log", "level=1", "--flags", "startup-clear")
@@ -492,7 +517,7 @@ func execCmd(ctx context.Context, basedir string, quiet bool, name string, env [
 
 	outputs := []io.Writer{logFile}
 
-	if !quiet {
+	if !quiet || slog.Default().Enabled(ctx, slog.LevelDebug) {
 		outputs = append(outputs, os.Stdout)
 	}