feat: add subComponentType in DGD API and uptake in planner (#3200)

Signed-off-by: tmontfort <tmontfort@nvidia.com>
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
Co-authored-by: hongkuanz <hongkuanz@nvidia.com>

container/Dockerfile.vllm

@@ -216,20 +216,6 @@ COPY --from=dynamo_base /usr/local/bin/etcd/ /usr/local/bin/etcd/
 # Add ETCD and CUDA binaries to PATH so cicc and other CUDA tools are accessible
 ENV PATH=/usr/local/bin/etcd/:/usr/local/cuda/nvvm/bin:$PATH
 
-# Install prometheus
-ARG PROM_VERSION=3.4.1
-RUN ARCH=$(dpkg --print-architecture) && \
-    case "$ARCH" in \
-        amd64) PLATFORM=linux-amd64 ;; \
-        arm64) PLATFORM=linux-arm64 ;; \
-        *) echo "Unsupported architecture: $ARCH" && exit 1 ;; \
-    esac && \
-    curl -fsSL --retry 5 --retry-delay 5 "https://github.com/prometheus/prometheus/releases/download/v${PROM_VERSION}/prometheus-${PROM_VERSION}.${PLATFORM}.tar.gz" \
-    | tar -xz -C /tmp && \
-    mv "/tmp/prometheus-${PROM_VERSION}.${PLATFORM}/prometheus" /usr/local/bin/ && \
-    chmod +x /usr/local/bin/prometheus && \
-    rm -rf "/tmp/prometheus-${PROM_VERSION}.${PLATFORM}"
-
 # Copy UCX from dev image as plugin for NIXL
 # Copy NIXL source from devr image
 # Copy dynamo wheels for gitlab artifacts

deploy/cloud/helm/crds/templates/nvidia.com_dynamocomponentdeployments.yaml

@@ -683,7 +683,7 @@ spec:
                     Typically corresponds to a component defined in the packaged Dynamo artifacts.
                   type: string
                 dynamoNamespace:
-                  description: dynamo namespace of the service (allows to override the dynamo namespace of the service defined in annotations inside the dynamo archive)
+                  description: Dynamo namespace of the service (allows to override the Dynamo namespace of the service defined in annotations inside the Dynamo archive)
                   type: string
                 dynamoTag:
                   description: 'contains the tag of the DynamoComponent: for example, "my_package:MyService"'
@@ -816,8 +816,9 @@ spec:
                   type: object
                 extraPodSpec:
                   description: |-
-                    ExtraPodSpec merges additional fields into the generated PodSpec for advanced
-                    customization (tolerations, node selectors, affinity, etc.).
+                    ExtraPodSpec allows to override the main pod spec configuration.
+                    It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field
+                    that allows overriding the main container configuration.
                   properties:
                     activeDeadlineSeconds:
                       description: |-
@@ -10239,7 +10240,7 @@ spec:
                       type: object
                   type: object
                 serviceName:
-                  description: contains the name of the component
+                  description: The name of the component
                   type: string
                 sharedMemory:
                   description: SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
@@ -10253,6 +10254,9 @@ spec:
                       pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
                       x-kubernetes-int-or-string: true
                   type: object
+                subComponentType:
+                  description: SubComponentType indicates the sub-role of this component (for example, "prefill").
+                  type: string
               type: object
             status:
               description: Status reflects the current observed state of the component deployment.

deploy/cloud/helm/crds/templates/nvidia.com_dynamographdeployments.yaml

@@ -785,7 +785,7 @@ spec:
                         description: ComponentType indicates the role of this component (for example, "main").
                         type: string
                       dynamoNamespace:
-                        description: dynamo namespace of the service (allows to override the dynamo namespace of the service defined in annotations inside the dynamo archive)
+                        description: Dynamo namespace of the service (allows to override the Dynamo namespace of the service defined in annotations inside the Dynamo archive)
                         type: string
                       envFromSecret:
                         description: |-
@@ -915,8 +915,9 @@ spec:
                         type: object
                       extraPodSpec:
                         description: |-
-                          ExtraPodSpec merges additional fields into the generated PodSpec for advanced
-                          customization (tolerations, node selectors, affinity, etc.).
+                          ExtraPodSpec allows to override the main pod spec configuration.
+                          It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field
+                          that allows overriding the main container configuration.
                         properties:
                           activeDeadlineSeconds:
                             description: |-
@@ -10338,7 +10339,7 @@ spec:
                             type: object
                         type: object
                       serviceName:
-                        description: contains the name of the component
+                        description: The name of the component
                         type: string
                       sharedMemory:
                         description: SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
@@ -10352,6 +10353,9 @@ spec:
                             pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
                             x-kubernetes-int-or-string: true
                         type: object
+                      subComponentType:
+                        description: SubComponentType indicates the sub-role of this component (for example, "prefill").
+                        type: string
                     type: object
                   description: |-
                     Services allows per-service overrides of the component deployment settings.

deploy/cloud/helm/platform/components/operator/templates/prometheus.yaml

@@ -27,9 +27,14 @@ spec:
     any: true
   {{- end }}
   podMetricsEndpoints:
-  - interval: 30s
+  - interval: 5s
     path: /metrics
     port: http
+    relabelings:
+    - action: replace
+      sourceLabels:
+      - __meta_kubernetes_pod_label_nvidia_com_dynamo_namespace
+      targetLabel: dynamo_namespace
   selector:
     matchLabels:
       nvidia.com/dynamo-component-type: frontend
@@ -49,7 +54,7 @@ spec:
     any: true
   {{- end }}
   podMetricsEndpoints:
-  - interval: 30s
+  - interval: 5s
     path: /metrics
     port: system
   selector:
@@ -71,7 +76,7 @@ spec:
     any: true
   {{- end }}
   podMetricsEndpoints:
-  - interval: 30s
+  - interval: 5s
     path: /metrics
     port: metrics
   selector:

deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go

@@ -73,6 +73,9 @@ type DynamoComponentDeploymentSharedSpec struct {
 	// ComponentType indicates the role of this component (for example, "main").
 	ComponentType string `json:"componentType,omitempty"`
 
+	// SubComponentType indicates the sub-role of this component (for example, "prefill").
+	SubComponentType string `json:"subComponentType,omitempty"`
+
 	// Dynamo namespace of the service (allows to override the Dynamo namespace of the service defined in annotations inside the Dynamo archive)
 	DynamoNamespace *string `json:"dynamoNamespace,omitempty"`
 

deploy/cloud/operator/config/crd/bases/nvidia.com_dynamocomponentdeployments.yaml

@@ -683,7 +683,7 @@ spec:
                     Typically corresponds to a component defined in the packaged Dynamo artifacts.
                   type: string
                 dynamoNamespace:
-                  description: dynamo namespace of the service (allows to override the dynamo namespace of the service defined in annotations inside the dynamo archive)
+                  description: Dynamo namespace of the service (allows to override the Dynamo namespace of the service defined in annotations inside the Dynamo archive)
                   type: string
                 dynamoTag:
                   description: 'contains the tag of the DynamoComponent: for example, "my_package:MyService"'
@@ -816,8 +816,9 @@ spec:
                   type: object
                 extraPodSpec:
                   description: |-
-                    ExtraPodSpec merges additional fields into the generated PodSpec for advanced
-                    customization (tolerations, node selectors, affinity, etc.).
+                    ExtraPodSpec allows to override the main pod spec configuration.
+                    It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field
+                    that allows overriding the main container configuration.
                   properties:
                     activeDeadlineSeconds:
                       description: |-
@@ -10239,7 +10240,7 @@ spec:
                       type: object
                   type: object
                 serviceName:
-                  description: contains the name of the component
+                  description: The name of the component
                   type: string
                 sharedMemory:
                   description: SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
@@ -10253,6 +10254,9 @@ spec:
                       pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
                       x-kubernetes-int-or-string: true
                   type: object
+                subComponentType:
+                  description: SubComponentType indicates the sub-role of this component (for example, "prefill").
+                  type: string
               type: object
             status:
               description: Status reflects the current observed state of the component deployment.

deploy/cloud/operator/config/crd/bases/nvidia.com_dynamographdeployments.yaml

@@ -785,7 +785,7 @@ spec:
                         description: ComponentType indicates the role of this component (for example, "main").
                         type: string
                       dynamoNamespace:
-                        description: dynamo namespace of the service (allows to override the dynamo namespace of the service defined in annotations inside the dynamo archive)
+                        description: Dynamo namespace of the service (allows to override the Dynamo namespace of the service defined in annotations inside the Dynamo archive)
                         type: string
                       envFromSecret:
                         description: |-
@@ -915,8 +915,9 @@ spec:
                         type: object
                       extraPodSpec:
                         description: |-
-                          ExtraPodSpec merges additional fields into the generated PodSpec for advanced
-                          customization (tolerations, node selectors, affinity, etc.).
+                          ExtraPodSpec allows to override the main pod spec configuration.
+                          It is a k8s standard PodSpec. It also contains a MainContainer (standard k8s Container) field
+                          that allows overriding the main container configuration.
                         properties:
                           activeDeadlineSeconds:
                             description: |-
@@ -10338,7 +10339,7 @@ spec:
                             type: object
                         type: object
                       serviceName:
-                        description: contains the name of the component
+                        description: The name of the component
                         type: string
                       sharedMemory:
                         description: SharedMemory controls the tmpfs mounted at /dev/shm (enable/disable and size).
@@ -10352,6 +10353,9 @@ spec:
                             pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
                             x-kubernetes-int-or-string: true
                         type: object
+                      subComponentType:
+                        description: SubComponentType indicates the sub-role of this component (for example, "prefill").
+                        type: string
                     type: object
                   description: |-
                     Services allows per-service overrides of the component deployment settings.

deploy/cloud/operator/internal/consts/consts.go

@@ -37,6 +37,7 @@ const (
 	KubeLabelDynamoNamespace            = "nvidia.com/dynamo-namespace"
 	KubeLabelDynamoDeploymentTargetType = "nvidia.com/dynamo-deployment-target-type"
 	KubeLabelDynamoComponentType        = "nvidia.com/dynamo-component-type"
+	KubeLabelDynamoSubComponentType     = "nvidia.com/dynamo-sub-component-type"
 
 	KubeLabelValueFalse = "false"
 	KubeLabelValueTrue  = "true"

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

@@ -1154,6 +1154,10 @@ func (r *DynamoComponentDeploymentReconciler) generatePodTemplateSpec(ctx contex
 		podLabels[commonconsts.KubeLabelDynamoComponentType] = opt.dynamoComponentDeployment.Spec.ComponentType
 	}
 
+	if opt.dynamoComponentDeployment.Spec.SubComponentType != "" {
+		podLabels[commonconsts.KubeLabelDynamoSubComponentType] = opt.dynamoComponentDeployment.Spec.SubComponentType
+	}
+
 	podAnnotations := make(map[string]string)
 
 	kubeName := r.getKubeName(opt.dynamoComponentDeployment, opt.isStealingTrafficDebugModeEnabled)

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

@@ -698,9 +698,10 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 										Value: "test_value_from_dynamo_component_deployment_spec",
 									},
 								},
-								ComponentType:   string(commonconsts.ComponentTypeWorker),
-								ServiceName:     "test-lws-deploy-service",
-								DynamoNamespace: &[]string{"default"}[0],
+								ComponentType:    string(commonconsts.ComponentTypeWorker),
+								SubComponentType: "test-sub-component",
+								ServiceName:      "test-lws-deploy-service",
+								DynamoNamespace:  &[]string{"default"}[0],
 								Multinode: &v1alpha1.MultinodeSpec{
 									NodeCount: 2,
 								},
@@ -783,6 +784,7 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									"role":                               "leader",
 									"nvidia.com/label1":                  "label1",
 									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoSubComponentType:    "test-sub-component",
 									commonconsts.KubeLabelDynamoGraphDeploymentName: "",
 								},
 								Annotations: map[string]string{
@@ -893,6 +895,7 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									"role":                               "worker",
 									"nvidia.com/label1":                  "label1",
 									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoSubComponentType:    "test-sub-component",
 									commonconsts.KubeLabelDynamoGraphDeploymentName: "",
 								},
 								Annotations: map[string]string{

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

@@ -100,10 +100,6 @@ func (r *DynamoGraphDeploymentReconciler) Reconcile(ctx context.Context, req ctr
 	if err = r.Get(ctx, req.NamespacedName, dynamoDeployment); err != nil {
 		return ctrl.Result{}, client.IgnoreNotFound(err)
 	}
-	if err != nil {
-		// not found, nothing to do
-		return ctrl.Result{}, nil
-	}
 
 	defer func() {
 		if err != nil {
@@ -129,7 +125,7 @@ func (r *DynamoGraphDeploymentReconciler) Reconcile(ctx context.Context, req ctr
 
 		err = r.Status().Update(ctx, dynamoDeployment)
 		if err != nil {
-			logger.Error(err, "Unable to update the CRD status", "crd", req.NamespacedName)
+			logger.Error(err, "Unable to update the CRD status", "crd", req.NamespacedName, "state", state, "reason", reason, "message", message)
 		}
 		logger.Info("Reconciliation done")
 	}()

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

@@ -992,6 +992,9 @@ func generateLabels(component *v1alpha1.DynamoComponentDeploymentOverridesSpec,
 	if component.ComponentType != "" {
 		labels[commonconsts.KubeLabelDynamoComponentType] = component.ComponentType
 	}
+	if component.SubComponentType != "" {
+		labels[commonconsts.KubeLabelDynamoSubComponentType] = component.SubComponentType
+	}
 	setMetricsLabels(labels, dynamoDeployment)
 	if component.Labels != nil {
 		err := mergo.Merge(&labels, component.Labels, mergo.WithOverride)

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

@@ -62,9 +62,10 @@ func TestGenerateDynamoComponentsDeployments(t *testing.T) {
 						Services: map[string]*v1alpha1.DynamoComponentDeploymentOverridesSpec{
 							"service1": {
 								DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
-									DynamoNamespace: &[]string{"default"}[0],
-									ComponentType:   "frontend",
-									Replicas:        &[]int32{3}[0],
+									DynamoNamespace:  &[]string{"default"}[0],
+									ComponentType:    "frontend",
+									SubComponentType: "test-sub-component",
+									Replicas:         &[]int32{3}[0],
 									Resources: &common.Resources{
 										Requests: &common.ResourceItem{
 											CPU:    "1",
@@ -106,10 +107,11 @@ func TestGenerateDynamoComponentsDeployments(t *testing.T) {
 					},
 					Spec: v1alpha1.DynamoComponentDeploymentSpec{
 						DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
-							ServiceName:     "service1",
-							DynamoNamespace: &[]string{"default"}[0],
-							ComponentType:   "frontend",
-							Replicas:        &[]int32{3}[0],
+							ServiceName:      "service1",
+							DynamoNamespace:  &[]string{"default"}[0],
+							ComponentType:    "frontend",
+							SubComponentType: "test-sub-component",
+							Replicas:         &[]int32{3}[0],
 							Resources: &common.Resources{
 								Requests: &common.ResourceItem{
 									CPU:    "1",
@@ -1088,7 +1090,8 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 						Services: map[string]*v1alpha1.DynamoComponentDeploymentOverridesSpec{
 							"Frontend": {
 								DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
-									ComponentType: "frontend", // Frontend component
+									ComponentType:    "frontend", // Frontend component
+									SubComponentType: "test-sub-component",
 									ExtraPodMetadata: &common.ExtraPodMetadata{
 										Annotations: map[string]string{
 											"nvidia.com/annotation1": "annotation1",
@@ -1240,6 +1243,7 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 									commonconsts.KubeLabelDynamoSelector:            "test-dynamo-graph-deployment-frontend",
 									commonconsts.KubeLabelMetricsEnabled:            commonconsts.KubeLabelValueTrue,
 									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeFrontend,
+									commonconsts.KubeLabelDynamoSubComponentType:    "test-sub-component",
 									commonconsts.KubeLabelDynamoGraphDeploymentName: "test-dynamo-graph-deployment",
 									"nvidia.com/label1":                             "label1",
 									"nvidia.com/label2":                             "label2",
@@ -1642,8 +1646,9 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 											"nvidia.com/label2": "label2",
 										},
 									},
-									Replicas:      &[]int32{5}[0],
-									ComponentType: commonconsts.ComponentTypeWorker,
+									Replicas:         &[]int32{5}[0],
+									ComponentType:    commonconsts.ComponentTypeWorker,
+									SubComponentType: "test-sub-component",
 									ExtraPodSpec: &common.ExtraPodSpec{
 										MainContainer: &corev1.Container{
 											Image: "worker-image",
@@ -1767,6 +1772,7 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 								Name: "worker-ldr",
 								Labels: map[string]string{
 									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoSubComponentType:    "test-sub-component",
 									commonconsts.KubeLabelMetricsEnabled:            commonconsts.KubeLabelValueTrue,
 									commonconsts.KubeLabelDynamoSelector:            "test-dynamo-graph-deployment-worker-ldr",
 									commonconsts.KubeLabelDynamoGraphDeploymentName: "test-dynamo-graph-deployment",
@@ -1917,6 +1923,7 @@ func TestGenerateGrovePodCliqueSet(t *testing.T) {
 								Name: "worker-wkr",
 								Labels: map[string]string{
 									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoSubComponentType:    "test-sub-component",
 									commonconsts.KubeLabelMetricsEnabled:            commonconsts.KubeLabelValueTrue,
 									commonconsts.KubeLabelDynamoSelector:            "test-dynamo-graph-deployment-worker-wkr",
 									commonconsts.KubeLabelDynamoGraphDeploymentName: "test-dynamo-graph-deployment",

deploy/metrics/k8s/frontend-podmonitor.yaml

@@ -14,7 +14,12 @@ spec:
   podMetricsEndpoints:
     - port: http
       path: /metrics
-      interval: 2s
+      interval: 5s
+      relabelings:
+      - action: replace
+        sourceLabels:
+        - __meta_kubernetes_pod_label_nvidia_com_dynamo_namespace
+        targetLabel: dynamo_namespace
   namespaceSelector:
     matchNames:
       - ${NAMESPACE}

deploy/metrics/k8s/planner-podmonitor.yaml

@@ -14,7 +14,7 @@ spec:
   podMetricsEndpoints:
     - port: metrics
       path: /metrics
-      interval: 2s
+      interval: 5s
   namespaceSelector:
     matchNames:
       - $NAMESPACE
\ No newline at end of file

deploy/metrics/k8s/worker-podmonitor.yaml

@@ -14,7 +14,7 @@ spec:
   podMetricsEndpoints:
     - port: system
       path: /metrics
-      interval: 2s
+      interval: 5s
   namespaceSelector:
     matchNames:
       - ${NAMESPACE}

docs/kubernetes/sla_planner_deployment.md

@@ -9,10 +9,12 @@ Quick deployment guide for the disaggregated planner with automatic scaling.
 
 **Components:**
 - **Frontend**: Serves requests and exposes `/metrics`
-- **Prometheus**: Scrapes frontend metrics every 5 seconds
-- **Planner**: Queries Prometheus and adjusts worker scaling every 60 seconds
+- **Prometheus**: Scrapes frontend metrics every adjustment interval
+- **Planner**: Queries Prometheus and adjusts worker scaling every adjustment interval
 - **Workers**: prefill and backend workers handle inference
 
+The adjustment interval can be defined in the planner manifest as an argument. The default interval value can be found in this [file](/components/planner/src/dynamo/planner/defaults.py).
+
 ```mermaid
 flowchart LR
   Frontend --"/metrics"--> Prometheus
@@ -25,6 +27,7 @@ flowchart LR
 - Kubernetes cluster with GPU nodes
 - [Pre-Deployment Profiling](/docs/benchmarks/pre_deployment_profiling.md) completed and its results saved to `dynamo-pvc` PVC.
 - Prefill and decode worker uses the best parallelization mapping suggested by the pre-deployment profiling script.
+- [kube-prometheus-stack](/docs/kubernetes/metrics.md) installed and running.
 
 > [!NOTE]
 > **Important**: The profiling that occurs before Planner deployment requires additional Kubernetes manifests (ServiceAccount, Role, RoleBinding, PVC) that are not included in standard Dynamo deployments. Apply these manifests in the same namespace as `$NAMESPACE`. For a complete setup, start with the [Quick Start guide](/deploy/utils/README.md#quick-start), which provides a fully encapsulated deployment including all required manifests.
@@ -50,7 +53,6 @@ Expected pods (all should be `1/1 Running`):
 ```
 # For vLLM:
 vllm-disagg-planner-frontend-*            1/1 Running
-vllm-disagg-planner-prometheus-*          1/1 Running
 vllm-disagg-planner-planner-*             1/1 Running
 vllm-disagg-planner-backend-*             1/1 Running
 vllm-disagg-planner-prefill-*             1/1 Running
@@ -103,8 +105,8 @@ kubectl logs -n $NAMESPACE deployment/vllm-disagg-planner-planner --tail=10
 
 **Connection Issues:**
 ```bash
-# Verify Prometheus is accessible (runs on port 8000)
-kubectl port-forward -n $NAMESPACE deployment/vllm-disagg-planner-prometheus 9090:8000
+# Verify Prometheus is accessible
+kubectl port-forward svc/prometheus-kube-prometheus-prometheus -n monitoring 9090:9090
 curl "http://localhost:9090/api/v1/query?query=up"
 ```
 
@@ -119,3 +121,11 @@ curl http://localhost:8000/metrics | grep nv_llm_http_service
 - Large models can take 10+ minutes to initialize
 - Check worker logs: `kubectl logs -n $NAMESPACE deployment/vllm-disagg-planner-backend`
 - Ensure GPU resources are available for workers
+
+**Unknown Field subComponentType:**
+
+If you encounter the following error when attempting to apply the deployment:
+```bash
+Error from server (BadRequest): error when creating "components/backends/vllm/deploy/disagg.yaml": DynamoGraphDeployment in version "v1alpha1" cannot be handled as a DynamoGraphDeployment: strict decoding error: unknown field "spec.services.DecodeWorker.subComponentType", unknown field "spec.services.PrefillWorker.subComponentType"
+```
+This is because the `subComponentType` field has only been added in newer versions of the DynamoGraphDeployment CRD (> 0.5.0). You can upgrade the CRD version by following the instructions [here](/docs/kubernetes/installation_guide.md).

tests/planner/perf_test_configs/disagg_8b_planner.yaml

@@ -55,9 +55,6 @@ spec:
       envFromSecret: hf-token-secret
       componentType: planner
       replicas: 1
-      envs:
-        - name: PROMETHEUS_PORT
-          value: "8000"
       livenessProbe:
         exec:
           command:
@@ -98,47 +95,11 @@ spec:
               --adjustment-interval=60
               --prometheus-port=9085
               --no-correction
-    Prometheus:
-      dynamoNamespace: vllm-disagg-planner
-      componentType: main
-      replicas: 1
-      envs:
-        - name: PYTHONPATH
-          value: "/workspace/components/planner/src"
-        - name: PROMETHEUS_PORT
-          value: "8000"
-      livenessProbe:
-        exec:
-          command:
-            - /bin/sh
-            - -c
-            - "exit 0"
-        periodSeconds: 60
-        timeoutSeconds: 30
-        failureThreshold: 10
-      readinessProbe:
-        exec:
-          command:
-            - /bin/sh
-            - -c
-            - "exit 0"
-        initialDelaySeconds: 30
-        periodSeconds: 60
-        timeoutSeconds: 30
-        failureThreshold: 10
-      extraPodSpec:
-        mainContainer:
-          image: my-registry/vllm-runtime:my-tag
-          workingDir: /workspace/components/backends/vllm
-          command:
-            - /bin/sh
-            - -c
-          args:
-            - "python3 -m dynamo.planner.prometheus"
     VllmDecodeWorker:
       dynamoNamespace: vllm-disagg-planner
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: decode
       replicas: 1
       livenessProbe:
         httpGet:
@@ -195,6 +156,7 @@ spec:
       dynamoNamespace: vllm-disagg-planner
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: prefill
       replicas: 1
       livenessProbe:
         httpGet:

tests/planner/scaling/disagg_planner.yaml

@@ -11,8 +11,6 @@ spec:
       value: '{"Prometheus":{"global":{"scrape_interval":"5s"},"scrape_configs":[{"job_name":"prometheus","static_configs":[{"targets":["localhost:9090"]}]},{"job_name":"frontend","static_configs":[{"targets":["vllm-disagg-planner-frontend:8000"]}]}]}}'
     - name: DYNAMO_NAMESPACE
       value: "vllm-disagg-planner"
-    - name: PROMETHEUS_PORT
-      value: "8000"
   services:
     Frontend:
       dynamoNamespace: vllm-disagg-planner
@@ -63,45 +61,11 @@ spec:
               --itl=0.01
               --load-predictor=constant
               --no-correction
-    Prometheus: # NOTE: this is set on Prometheus to ensure a service is created for the Prometheus component. This is a workaround and should be managed differently.
-      dynamoNamespace: vllm-disagg-planner
-      componentType: frontend
-      replicas: 1
-      envs:
-        - name: PYTHONPATH
-          value: "/workspace/components/planner/src"
-      livenessProbe:
-        exec:
-          command:
-            - /bin/sh
-            - -c
-            - "exit 0"
-        periodSeconds: 60
-        timeoutSeconds: 30
-        failureThreshold: 10
-      readinessProbe:
-        exec:
-          command:
-            - /bin/sh
-            - -c
-            - "exit 0"
-        initialDelaySeconds: 30
-        periodSeconds: 60
-        timeoutSeconds: 30
-        failureThreshold: 10
-      extraPodSpec:
-        mainContainer:
-          image: my-registry/vllm-runtime:my-tag
-          workingDir: /workspace/components/backends/vllm
-          command:
-            - /bin/sh
-            - -c
-          args:
-            - "python3 -m dynamo.planner.prometheus"
     VllmDecodeWorker:
       dynamoNamespace: vllm-disagg-planner
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: decode
       replicas: 1
       resources:
         limits:
@@ -125,6 +89,7 @@ spec:
       dynamoNamespace: vllm-disagg-planner
       envFromSecret: hf-token-secret
       componentType: worker
+      subComponentType: prefill
       replicas: 1
       resources:
         limits: