fix: fix operator tests (#3904)

Signed-off-by: Julien Mancuso <jmancuso@nvidia.com>

deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go

@@ -734,11 +734,15 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									MainContainer: &corev1.Container{
 										Image: "test-image:latest",
 										Command: []string{
-											"sh",
-											"-c",
+											"some",
+											"dynamo",
+											"command",
 										},
 										Args: []string{
-											"some dynamo command",
+											"--tensor-parallel-size",
+											"4",
+											"--pipeline-parallel-size",
+											"1",
 										},
 										Env: []corev1.EnvVar{
 											{
@@ -817,8 +821,8 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									{
 										Name:    commonconsts.MainContainerName,
 										Image:   "test-image:latest",
-										Command: []string{"sh", "-c"},
-										Args:    []string{"ray start --head --port=6379 && some dynamo command"},
+										Command: []string{"/bin/sh", "-c"},
+										Args:    []string{"ray start --head --port=6379 && some dynamo command --tensor-parallel-size 4 --pipeline-parallel-size 1"},
 										Env: []corev1.EnvVar{
 											{Name: "DYN_NAMESPACE", Value: "default"},
 											{Name: "DYN_PARENT_DGD_K8S_NAME", Value: "test-lws-deploy"},
@@ -931,7 +935,7 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									{
 										Name:    commonconsts.MainContainerName,
 										Image:   "test-image:latest",
-										Command: []string{"sh", "-c"},
+										Command: []string{"/bin/sh", "-c"},
 										Args:    []string{"ray start --address=$(LWS_LEADER_ADDRESS):6379 --block"},
 										Env: []corev1.EnvVar{
 											{Name: "DYN_NAMESPACE", Value: "default"},

deploy/cloud/operator/internal/dynamo/backend_vllm.go

@@ -106,7 +106,7 @@ func injectRayDistributedLaunchFlags(container *corev1.Container, role Role, ser
 		leaderHostname := multinodeDeployer.GetLeaderHostname(serviceName)
 		container.Args = []string{fmt.Sprintf("ray start --address=%s:%s --block", leaderHostname, VLLMPort)}
 	}
-	container.Command = []string{"sh", "-c"} // ensure cmd is a shell
+	container.Command = []string{"/bin/sh", "-c"} // ensure cmd is a shell
 }
 
 func injectDataParallelLaunchFlags(container *corev1.Container, role Role, serviceName string, multinodeDeployer MultinodeDeployer) {

deploy/cloud/operator/internal/dynamo/backend_vllm_test.go

@@ -303,7 +303,7 @@ func TestUpdateVLLMMultinodeArgs(t *testing.T) {
 			name:              "leader prepends ray start --head",
 			role:              RoleLeader,
 			multinodeDeployer: &GroveMultinodeDeployer{},
-			initialContainer:  &corev1.Container{Args: []string{"python3", "-m", "dynamo.vllm", tensorParallelSizeFlag, "16"}, Resources: corev1.ResourceRequirements{Limits: corev1.ResourceList{corev1.ResourceName("nvidia.com/gpu"): resource.MustParse("8")}}},
+			initialContainer:  &corev1.Container{Command: []string{"python3"}, Args: []string{"-m", "dynamo.vllm", tensorParallelSizeFlag, "16"}, Resources: corev1.ResourceRequirements{Limits: corev1.ResourceList{corev1.ResourceName("nvidia.com/gpu"): resource.MustParse("8")}}},
 			expectedArgs:      []string{fmt.Sprintf("ray start --head --port=%s && python3 -m dynamo.vllm %s 16", VLLMPort, tensorParallelSizeFlag)},
 		},
 		{

deploy/cloud/operator/internal/dynamo/graph_test.go

@@ -2417,11 +2417,17 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 									MainContainer: &corev1.Container{
 										Image: "worker-image",
 										Command: []string{
-											"/bin/sh",
-											"-c",
+											"python3",
+											"-m",
+											"dynamo.vllm",
 										},
 										Args: []string{
-											"python3 -m dynamo.vllm --custom-flag custom-value",
+											"--custom-flag",
+											"custom-value",
+											"--tensor-parallel-size",
+											"4",
+											"--pipeline-parallel-size",
+											"1",
 										},
 										StartupProbe: &corev1.Probe{
 											ProbeHandler: corev1.ProbeHandler{
@@ -2598,7 +2604,7 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 													"-c",
 												},
 												Args: []string{
-													"ray start --head --port=6379 && python3 -m dynamo.vllm --custom-flag custom-value",
+													"ray start --head --port=6379 && python3 -m dynamo.vllm --custom-flag custom-value --tensor-parallel-size 4 --pipeline-parallel-size 1",
 												},
 												Ports: []corev1.ContainerPort{
 													{
@@ -3345,7 +3351,7 @@ func TestGeneratePodSpecForComponent_VLLM(t *testing.T) {
 				ComponentType: commonconsts.ComponentTypeWorker,
 				ExtraPodSpec: &common.ExtraPodSpec{
 					MainContainer: &corev1.Container{
-						Args: []string{"python3", "-m", "dynamo.vllm"},
+						Args: []string{"python3", "-m", "dynamo.vllm", "--tensor-parallel-size", "4", "--pipeline-parallel-size", "1"},
 					},
 				},
 			},
@@ -3359,6 +3365,11 @@ func TestGeneratePodSpecForComponent_VLLM(t *testing.T) {
 			name: "VLLM multinode worker",
 			component: &v1alpha1.DynamoComponentDeploymentSharedSpec{
 				ComponentType: commonconsts.ComponentTypeWorker,
+				ExtraPodSpec: &common.ExtraPodSpec{
+					MainContainer: &corev1.Container{
+						Args: []string{"python3", "-m", "dynamo.vllm", "--tensor-parallel-size", "4", "--pipeline-parallel-size", "1"},
+					},
+				},
 			},
 			backendFramework:  BackendFrameworkVLLM,
 			role:              RoleWorker,
@@ -4757,7 +4768,7 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 
 	tests := []struct {
 		name                   string
-		component              *v1alpha1.DynamoComponentDeploymentOverridesSpec
+		component              *v1alpha1.DynamoComponentDeploymentSharedSpec
 		expectError            bool
 		expectedResourceClaims []corev1.ResourceClaim
 		expectedPodClaims      []corev1.PodResourceClaim
@@ -4765,8 +4776,7 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 	}{
 		{
 			name: "component with resource claims",
-			component: &v1alpha1.DynamoComponentDeploymentOverridesSpec{
-				DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
+			component: &v1alpha1.DynamoComponentDeploymentSharedSpec{
 				ComponentType: commonconsts.ComponentTypeWorker,
 				Resources: &common.Resources{
 					Requests: &common.ResourceItem{
@@ -4818,7 +4828,6 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 					},
 				},
 			},
-			},
 			expectError: false,
 			expectedResourceClaims: []corev1.ResourceClaim{
 				{
@@ -4853,8 +4862,7 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 		},
 		{
 			name: "component with multiple resource claims",
-			component: &v1alpha1.DynamoComponentDeploymentOverridesSpec{
-				DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
+			component: &v1alpha1.DynamoComponentDeploymentSharedSpec{
 				ComponentType: commonconsts.ComponentTypeWorker,
 				Resources: &common.Resources{
 					Claims: []corev1.ResourceClaim{
@@ -4886,7 +4894,6 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 					},
 				},
 			},
-			},
 			expectError: false,
 			expectedResourceClaims: []corev1.ResourceClaim{
 				{
@@ -4909,8 +4916,7 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 		},
 		{
 			name: "component without resource claims",
-			component: &v1alpha1.DynamoComponentDeploymentOverridesSpec{
-				DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
+			component: &v1alpha1.DynamoComponentDeploymentSharedSpec{
 				ComponentType: commonconsts.ComponentTypeFrontend,
 				Resources: &common.Resources{
 					Requests: &common.ResourceItem{
@@ -4919,7 +4925,6 @@ func TestGenerateBasePodSpec_ResourceClaims(t *testing.T) {
 					},
 				},
 			},
-			},
 			expectError:            false,
 			expectedResourceClaims: nil,
 			expectedPodClaims:      nil,

docs/kubernetes/dynamo_operator.md

@@ -39,7 +39,8 @@ helm fetch https://helm.ngc.nvidia.com/nvidia/ai-dynamo/charts/dynamo-platform-$
 helm install dynamo-platform dynamo-platform-${RELEASE_VERSION}.tgz --namespace ${NAMESPACE} --create-namespace
 ```
 
-For namespace-restricted installations (shared clusters):
+For namespace-restricted installations (shared clusters), you'll need to install the Dynamo platform in each namespace you want to deploy to.
+Namespace restriction is enabled by setting the `dynamo-operator.namespaceRestriction.enabled` flag to `true`.
 ```bash
 helm install dynamo-platform dynamo-platform-${RELEASE_VERSION}.tgz \
   --namespace ${NAMESPACE} \