Merge branch 'main' of https://github.com/ai-dynamo/dynamo into hzhou/metrics

Author: tedzhouhk

components/metrics/README.md

@@ -82,7 +82,7 @@ metrics --component VllmWorker --endpoint load_metrics
 
 **NOTE**: `load_metrics` is currently a
 [hard-coded](https://github.com/ai-dynamo/dynamo/blob/d5220c7b1151372ba3d2a061c7d0a7ed72724789/lib/llm/src/kv_router/publisher.rs#L108)
-endpoint name used for python-based workers that register a `KvMetricsPublisher`.
+endpoint name used for python-based workers that register a `WorkerMetricsPublisher`.
 
 ## Visualization
 

components/planner/README.md

@@ -17,7 +17,7 @@ limitations under the License.
 
 # Planner
 
-The planner is a component that monitors the state of the system and makes adjustments to the number of workers to ensure that the system is running efficiently. It can dynamically scale prefill/decode workers up and down based on a variety of KV metrics. You can find documentation and benchmarking examples in the [planner docs](../../docs/planner.md).
+The planner is a component that monitors the state of the system and makes adjustments to the number of workers to ensure that the system is running efficiently. It can dynamically scale prefill/decode workers up and down based on a variety of KV metrics. You can find documentation and benchmarking examples in the [planner docs](../../docs/guides/planner.md).
 
 ## Usage
 

container/Dockerfile.sglang

@@ -325,6 +325,9 @@ COPY --from=wheel_builder /opt/nvidia/nvda_nixl /opt/nvidia/nvda_nixl
 # Copy Cargo cache to avoid re-downloading dependencies
 COPY --from=wheel_builder $CARGO_HOME $CARGO_HOME
 
+# Copy rest of the code
+COPY . /workspace
+
 # Build C bindings, creates lib/bindings/c/include
 RUN cd /workspace/lib/bindings/c && cargo build --release --locked
 
@@ -365,8 +368,6 @@ ARG GENAI_PERF_VERSION
 # Install genai-perf for benchmarking
 RUN uv pip install genai-perf==$GENAI_PERF_VERSION
 
-COPY . /workspace
-
 ENTRYPOINT ["/opt/nvidia/nvidia_entrypoint.sh"]
 
 CMD []

container/Dockerfile.tensorrt_llm

@@ -300,6 +300,9 @@ COPY --from=wheel_builder /workspace /workspace
 # Copy Cargo cache to avoid re-downloading dependencies
 COPY --from=wheel_builder $CARGO_HOME $CARGO_HOME
 
+# Copy rest of the code
+COPY . /workspace
+
 # Build C bindings, creates lib/bindings/c/include
 RUN cd /workspace/lib/bindings/c && cargo build --release --locked
 

container/Dockerfile.vllm

@@ -403,6 +403,9 @@ COPY --from=wheel_builder /opt/nvidia/nvda_nixl /opt/nvidia/nvda_nixl
 # Copy Cargo cache to avoid re-downloading dependencies
 COPY --from=wheel_builder $CARGO_HOME $CARGO_HOME
 
+# Copy rest of the code
+COPY . /workspace
+
 # Build C bindings, creates lib/bindings/c/include
 #
 # TODO: In theory the 'cargo build' in earlier stage covers this, we "just" need to copy the
@@ -447,8 +450,6 @@ ARG GENAI_PERF_VERSION
 # Install genai-perf for benchmarking
 RUN uv pip install genai-perf==$GENAI_PERF_VERSION
 
-COPY . /workspace
-
 ENTRYPOINT ["/opt/nvidia/nvidia_entrypoint.sh"]
 
 CMD []

deploy/cloud/helm/platform/components/operator/charts/dynamo-crds/Chart.yaml renamed to deploy/cloud/helm/crds/Chart.yaml

@@ -14,7 +14,7 @@
 # limitations under the License.
 apiVersion: v2
 name: dynamo-crds
-description: A Helm chart for CRDs of dynamo operator
+description: A Helm chart for dynamo CRDs
 type: application
 version: 0.1.6
 dependencies: []

deploy/cloud/helm/platform/components/operator/charts/dynamo-crds/crds/nvidia.com_dynamocomponentdeployments.yaml renamed to deploy/cloud/helm/crds/templates/nvidia.com_dynamocomponentdeployments.yaml

@@ -19,6 +19,8 @@ kind: CustomResourceDefinition
 metadata:
   annotations:
     controller-gen.kubebuilder.io/version: v0.16.4
+    # never delete this resource with helm delete
+    helm.sh/resource-policy: keep
   name: dynamocomponentdeployments.nvidia.com
 spec:
   group: nvidia.com

deploy/cloud/helm/platform/components/operator/charts/dynamo-crds/crds/nvidia.com_dynamocomponents.yaml renamed to deploy/cloud/helm/crds/templates/nvidia.com_dynamocomponents.yaml

@@ -19,6 +19,8 @@ kind: CustomResourceDefinition
 metadata:
   annotations:
     controller-gen.kubebuilder.io/version: v0.16.4
+    # never delete this resource with helm delete
+    helm.sh/resource-policy: keep
   name: dynamocomponents.nvidia.com
 spec:
   group: nvidia.com

deploy/cloud/helm/platform/components/operator/charts/dynamo-crds/crds/nvidia.com_dynamographdeployments.yaml renamed to deploy/cloud/helm/crds/templates/nvidia.com_dynamographdeployments.yaml

@@ -19,6 +19,8 @@ kind: CustomResourceDefinition
 metadata:
   annotations:
     controller-gen.kubebuilder.io/version: v0.16.4
+    # never delete this resource with helm delete
+    helm.sh/resource-policy: keep
   name: dynamographdeployments.nvidia.com
 spec:
   group: nvidia.com

deploy/cloud/helm/deploy.sh

@@ -41,7 +41,7 @@ export ENABLE_LWS="${ENABLE_LWS:=false}"
 
 # Add command line options
 INTERACTIVE=false
-
+INSTALL_CRDS=false
 # Parse command line arguments
 while [[ $# -gt 0 ]]; do
   key="$1"
@@ -50,11 +50,16 @@ while [[ $# -gt 0 ]]; do
       INTERACTIVE=true
       shift
       ;;
+    --crds)
+      INSTALL_CRDS=true
+      shift
+      ;;
     --help)
       echo "Usage: $0 [options]"
       echo "Options:"
       echo "  --interactive       Run in interactive mode"
       echo "  --help              Show this help message"
+      echo "  --crds              Also install the CRDs"
       exit 0
       ;;
     *)
@@ -115,15 +120,6 @@ retry_command "$HELM_CMD repo add bitnami https://charts.bitnami.com/bitnami" 5
 retry_command "$HELM_CMD repo add minio https://charts.min.io/" 5 5 && \
 retry_command "$HELM_CMD repo update" 5 5
 
-cd platform
-cd components/operator
-retry_command "$HELM_CMD dependency update" 5 5
-cd ../..
-cd components/api-store
-retry_command "$HELM_CMD dependency update" 5 5
-cd ../..
-retry_command "$HELM_CMD dep update" 7 5
-cd ..
 
 # Generate the values file
 echo "Generating values file with:"
@@ -143,6 +139,7 @@ echo "INGRESS_CLASS: $INGRESS_CLASS"
 echo "ISTIO_GATEWAY: $ISTIO_GATEWAY"
 echo "DYNAMO_INGRESS_SUFFIX: $DYNAMO_INGRESS_SUFFIX"
 echo "VIRTUAL_SERVICE_SUPPORTS_HTTPS: $VIRTUAL_SERVICE_SUPPORTS_HTTPS"
+echo "INSTALL_CRDS: $INSTALL_CRDS"
 
 envsubst '${NAMESPACE} ${RELEASE_NAME} ${DOCKER_USERNAME} ${DOCKER_PASSWORD} ${DOCKER_SERVER} ${IMAGE_TAG} ${DYNAMO_INGRESS_SUFFIX} ${PIPELINES_DOCKER_SERVER} ${PIPELINES_DOCKER_USERNAME} ${PIPELINES_DOCKER_PASSWORD} ${DOCKER_SECRET_NAME} ${INGRESS_ENABLED} ${ISTIO_ENABLED} ${INGRESS_CLASS} ${ISTIO_GATEWAY} ${VIRTUAL_SERVICE_SUPPORTS_HTTPS} ${ENABLE_LWS}' < dynamo-platform-values.yaml > generated-values.yaml
 echo "generated file contents:"
@@ -157,6 +154,12 @@ cd platform
 retry_command "$HELM_CMD dep build" 5 5
 cd ..
 
+# Install/upgrade the helm chart for the CRDs
+if [ "$INSTALL_CRDS" = true ]; then
+  echo "Installing/upgrading helm chart for the CRDs..."
+  $HELM_CMD upgrade --install dynamo-crds crds/ --namespace default --wait --atomic
+fi
+
 # Install/upgrade the helm chart
 echo "Installing/upgrading helm chart..."
 $HELM_CMD upgrade --install $RELEASE_NAME platform/ \