fix(self-hosted): repair the CLI-driven local install path

src/clis/nvcf-cli/cmd/self_hosted_compute_plane.go

@@ -132,7 +132,7 @@ func runSelfHostedComputePlaneInstall(c *cobra.Command, _ []string) error {
 		Stdout:          c.OutOrStdout(),
 		Stderr:          c.ErrOrStderr(),
 		Ctx:             c.Context(),
-		ExtraEnv:        computePlaneInstallEnv(clusterName, ncaID),
+		ExtraEnv:        computePlaneInstallEnv(clusterName, ncaID, filepath.Dir(valuesPath)),
 	})
 }
 
@@ -234,10 +234,14 @@ func inferClusterNameFromValuesPath(path string) string {
 	return ""
 }
 
-func computePlaneInstallEnv(clusterName, ncaID string) []string {
+func computePlaneInstallEnv(clusterName, ncaID, outputDir string) []string {
 	return []string{
 		"CLUSTER_NAME=" + clusterName,
 		"NCA_ID=" + ncaID,
+		// The worker helmfile resolves the registration values at
+		// $OUTPUT_DIR/$CLUSTER_NAME-register-values.yaml via requiredEnv, so
+		// point it at the directory holding the --values file.
+		"OUTPUT_DIR=" + outputDir,
 	}
 }
 

src/clis/nvcf-cli/cmd/self_hosted_compute_plane_test.go

@@ -117,6 +117,9 @@ func TestComputePlaneInstallTemplatesUserValuesFile(t *testing.T) {
 	assert.Contains(t, out, "arg=--kube-context=gpu-context")
 	assert.Contains(t, out, "env:CLUSTER_NAME=gpu-from-values")
 	assert.Contains(t, out, "env:NCA_ID=nca-from-values")
+	// The worker helmfile reads $OUTPUT_DIR/$CLUSTER_NAME-register-values.yaml,
+	// so install must point OUTPUT_DIR at the directory of --values.
+	assert.Contains(t, out, "env:OUTPUT_DIR="+filepath.Dir(valuesFile))
 	assert.Contains(t, out, "arg="+stackDir)
 	assert.FileExists(t, fakeBin)
 }
@@ -928,6 +931,7 @@ done
 printf 'verb=%s\n' "$last"
 printf 'env:CLUSTER_NAME=%s\n' "$CLUSTER_NAME"
 printf 'env:NCA_ID=%s\n' "$NCA_ID"
+printf 'env:OUTPUT_DIR=%s\n' "$OUTPUT_DIR"
 `
 	require.NoError(t, os.WriteFile(fakeBin, []byte(body), 0o755))
 	t.Setenv("PATH", filepath.Dir(fakeBin)+":"+os.Getenv("PATH"))

src/clis/nvcf-cli/cmd/self_hosted_down.go

@@ -331,6 +331,10 @@ func runDownComputePlaneForCluster(c *cobra.Command, ctx context.Context, sink p
 	extra := []string{
 		"CLUSTER_NAME=" + clusterName,
 		"NCA_ID=" + downNCAID,
+		// The worker helmfile also requires OUTPUT_DIR to resolve
+		// $OUTPUT_DIR/$CLUSTER_NAME-register-values.yaml, the same
+		// directory readRegisterValuesYAML reads from below.
+		"OUTPUT_DIR=" + filepath.Join(resolved.Path, "out"),
 	}
 	unregisterClusterID := clusterName
 	if rv, err := readRegisterValuesYAML(resolved.Path, clusterName); err == nil {

src/clis/nvcf-cli/cmd/self_hosted_down_test.go

@@ -259,7 +259,7 @@ func installFakeHelmfile(t *testing.T) string {
 	logPath := filepath.Join(dir, "helmfile.log")
 	binPath := filepath.Join(dir, "helmfile")
 	script := `#!/bin/sh
-printf '%s\n' "$PWD|$*|CLUSTER_NAME=${CLUSTER_NAME}" >> "$NVCF_TEST_HELMFILE_LOG"
+printf '%s\n' "$PWD|$*|CLUSTER_NAME=${CLUSTER_NAME}|OUTPUT_DIR=${OUTPUT_DIR}" >> "$NVCF_TEST_HELMFILE_LOG"
 `
 	require.NoError(t, os.WriteFile(binPath, []byte(script), 0o755))
 	t.Setenv("NVCF_TEST_HELMFILE_LOG", logPath)
@@ -348,6 +348,9 @@ func TestDown_ClusterNameCleansControlPlaneWhenLastClusterRemoved(t *testing.T)
 	require.Len(t, invocations, 2, "last cluster removal must destroy compute and control planes")
 	assert.Contains(t, invocations[0], filepath.Join(computePlaneStack, "helmfile.d")+"/")
 	assert.Contains(t, invocations[0], "CLUSTER_NAME=test-cluster")
+	// The worker helmfile reads $OUTPUT_DIR/$CLUSTER_NAME-register-values.yaml
+	// at render time, so destroy must export it like install does.
+	assert.Contains(t, invocations[0], "OUTPUT_DIR="+filepath.Join(computePlaneStack, "out"))
 	assert.Contains(t, invocations[1], filepath.Join(controlPlaneStack, "helmfile.d")+"/")
 	assert.Contains(t, invocations[1], "--sequential-helmfiles")
 }

tests/bdd/features/multi-cluster-up.feature

@@ -31,6 +31,28 @@ Feature: Bring up a local multi-cluster NVCF stack with the CLI
       # teardown.
       And I copy the file "deploy/stacks/self-managed/secrets/secrets.yaml.template" to "deploy/stacks/self-managed/secrets/local-secrets.yaml"
       And I substitute "REPLACE_WITH_BASE64_DOCKER_CREDENTIAL" in file "deploy/stacks/self-managed/secrets/local-secrets.yaml" with base64 of "$oauthtoken:${NGC_API_KEY}"
+      # --env local also reads operator-authored environment values from
+      # both split stacks: deploy/stacks/<stack>/environments/local.yaml.
+      # Neither file is tracked, so author both from the BDD multi-cluster
+      # fixtures (they carry the alias-service URL shape the split
+      # topology needs). observability.profile is disabled because this
+      # workflow runs 'helmfile apply', whose diff phase validates
+      # rendered manifests against the live cluster (--dry-run=server);
+      # on a fresh cluster the ServiceMonitor CRDs do not exist yet and
+      # the diff fails before anything installs. The Helmfile workflow
+      # (helmfile sync) has no diff phase and keeps the default profile.
+      And I copy the file "tests/bdd/fixtures/self-managed-local-bdd-multi.yaml" to "deploy/stacks/self-managed/environments/local.yaml"
+      And I update yaml file "deploy/stacks/self-managed/environments/local.yaml" with keys:
+        | global.imagePullSecrets[0].name | nvcr-pull-secret                     |
+        | global.helm.sources.repository  | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | global.image.repository         | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | observability.profile           | disabled                             |
+      And I copy the file "tests/bdd/fixtures/nvcf-compute-plane-local-bdd-multi.yaml" to "deploy/stacks/nvcf-compute-plane/environments/local.yaml"
+      And I update yaml file "deploy/stacks/nvcf-compute-plane/environments/local.yaml" with keys:
+        | global.imagePullSecrets[0].name | nvcr-pull-secret                     |
+        | global.helm.sources.repository  | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | global.image.repository         | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | observability.profile           | disabled                             |
       # Conflict precheck: single-cluster ncp-local's k3d serverlb
       # claims 0.0.0.0:8080/8443/10081, and ncp-local-cp also
       # needs NATS on 4222 plus the worker callback port 10086.

tests/bdd/features/single-cluster-up.feature

@@ -31,6 +31,27 @@ Feature: Bring up a local single-cluster NVCF stack with the CLI
       # teardown, so the working tree stays clean.
       And I copy the file "deploy/stacks/self-managed/secrets/secrets.yaml.template" to "deploy/stacks/self-managed/secrets/local-secrets.yaml"
       And I substitute "REPLACE_WITH_BASE64_DOCKER_CREDENTIAL" in file "deploy/stacks/self-managed/secrets/local-secrets.yaml" with base64 of "$oauthtoken:${NGC_API_KEY}"
+      # --env local also reads operator-authored environment values from
+      # both split stacks: deploy/stacks/<stack>/environments/local.yaml.
+      # Neither file is tracked, so author both from the BDD fixtures.
+      # observability.profile is disabled because this workflow runs
+      # 'helmfile apply', whose diff phase validates rendered manifests
+      # against the live cluster (--dry-run=server); on a fresh cluster
+      # the ServiceMonitor CRDs do not exist yet and the diff fails
+      # before anything installs. The Helmfile workflow (helmfile sync)
+      # has no diff phase and keeps the default profile.
+      And I copy the file "tests/bdd/fixtures/self-managed-local-bdd.yaml" to "deploy/stacks/self-managed/environments/local.yaml"
+      And I update yaml file "deploy/stacks/self-managed/environments/local.yaml" with keys:
+        | global.imagePullSecrets[0].name | nvcr-pull-secret                     |
+        | global.helm.sources.repository  | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | global.image.repository         | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | observability.profile           | disabled                             |
+      And I copy the file "tests/bdd/fixtures/nvcf-compute-plane-local-bdd.yaml" to "deploy/stacks/nvcf-compute-plane/environments/local.yaml"
+      And I update yaml file "deploy/stacks/nvcf-compute-plane/environments/local.yaml" with keys:
+        | global.imagePullSecrets[0].name | nvcr-pull-secret                     |
+        | global.helm.sources.repository  | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | global.image.repository         | ${SAMPLE_NGC_ORG}/${SAMPLE_NGC_TEAM} |
+        | observability.profile           | disabled                             |
       # Conflict precheck: ncp-local-cp's k3d serverlb claims
       # 0.0.0.0:8080/8443/10081, NATS on 4222, and the worker
       # callback port 10086, overlapping host ports single-cluster

tests/bdd/fixtures/nvcf-cli-local.yaml

@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 base_http_url: "http://api.localhost:8080"
-base_grpc_url: "localhost:10081"
+base_grpc_url: "grpc.localhost:10081"
 invoke_url: "http://invocation.localhost:8080"
 api_keys_service_url: "http://api-keys.localhost:8080"
 icms_url: "http://sis.localhost:8080"

tests/bdd/fixtures_test.go

@@ -145,7 +145,7 @@ func TestNVCFCLILocalFixtureTargetsLocalGRPCGateway(t *testing.T) {
 	if err := yaml.Unmarshal(fixtureBytes, &fixture); err != nil {
 		t.Fatalf("parse local CLI fixture: %v", err)
 	}
-	if got, want := fixture["base_grpc_url"], "localhost:10081"; got != want {
+	if got, want := fixture["base_grpc_url"], "grpc.localhost:10081"; got != want {
 		t.Fatalf("base_grpc_url = %v, want %q", got, want)
 	}
 }

tests/bdd/godog_test.go

@@ -280,6 +280,8 @@ func TestSingleClusterUpFeatureFileWiresToSteps(t *testing.T) {
 	writeProfileHandoffArtifact(t, suite.Config.RepoRoot)
 	writeSingleClusterComputeRegisterValues(t, suite.Config.RepoRoot)
 	seedStackSecretsTemplate(t, suite.Config.RepoRoot)
+	seedHelmfileLocalBDDFixture(t, suite.Config.RepoRoot)
+	seedComputePlaneLocalBDDFixture(t, suite.Config.RepoRoot)
 
 	sc := steps.NewScenarioContext(suite)
 	featurePath := mustResolveFeaturePath(t, "single-cluster-up.feature")
@@ -361,6 +363,8 @@ func TestMultiClusterUpFeatureFileWiresToSteps(t *testing.T) {
 	writeMulticlusterProfileHandoffArtifact(t, suite.Config.RepoRoot)
 	writeMulticlusterComputeRegisterValues(t, suite.Config.RepoRoot, "nvcf-compute-plane", "ncp-local-compute-1")
 	seedStackSecretsTemplate(t, suite.Config.RepoRoot)
+	seedHelmfileLocalBDDMultiFixture(t, suite.Config.RepoRoot)
+	seedComputePlaneLocalBDDMultiFixture(t, suite.Config.RepoRoot)
 
 	sc := steps.NewScenarioContext(suite)
 	featurePath := mustResolveFeaturePath(t, "multi-cluster-up.feature")