add node labels and integration test support for compute-domain-dra-plugin

- Add computeDomainDevicePluginLabelKey to status-updater node labeling
- Add compute-domain-dra-plugin to integration test setup
- Enable computeDomainDraPlugin in integration test values

Author: enoodle

internal/status-updater/handlers/node/labels.go

@@ -12,9 +12,10 @@ import (
 )
 
 const (
-	dcgmExporterLabelKey = "nvidia.com/gpu.deploy.dcgm-exporter"
-	devicePluginLabelKey = "nvidia.com/gpu.deploy.device-plugin"
-	draPluginGpuLabelKey = "nvidia.com/gpu.deploy.dra-plugin-gpu"
+	dcgmExporterLabelKey              = "nvidia.com/gpu.deploy.dcgm-exporter"
+	devicePluginLabelKey              = "nvidia.com/gpu.deploy.device-plugin"
+	draPluginGpuLabelKey              = "nvidia.com/gpu.deploy.dra-plugin-gpu"
+	computeDomainDevicePluginLabelKey = "nvidia.com/gpu.deploy.compute-domain-dra-plugin"
 )
 
 // labelNode labels the node with required labels for the fake-gpu-operator to function.
@@ -25,6 +26,7 @@ func (p *NodeHandler) labelNode(node *v1.Node) error {
 	if !isFakeNode(node) {
 		labels[devicePluginLabelKey] = "true"
 		labels[draPluginGpuLabelKey] = "true"
+		labels[computeDomainDevicePluginLabelKey] = "true"
 	}
 
 	err := p.patchNodeLabels(node, labels)
@@ -38,9 +40,10 @@ func (p *NodeHandler) labelNode(node *v1.Node) error {
 // unlabelNode removes the labels from the node that were added by the fake-gpu-operator.
 func (p *NodeHandler) unlabelNode(node *v1.Node) error {
 	err := p.patchNodeLabels(node, map[string]interface{}{
-		dcgmExporterLabelKey: nil,
-		devicePluginLabelKey: nil,
-		draPluginGpuLabelKey: nil,
+		dcgmExporterLabelKey:              nil,
+		devicePluginLabelKey:              nil,
+		draPluginGpuLabelKey:              nil,
+		computeDomainDevicePluginLabelKey: nil,
 	})
 	if err != nil && !errors.IsNotFound(err) {
 		return fmt.Errorf("failed to unlabel node %s: %w", node.Name, err)

test/integration/setup.sh

@@ -53,7 +53,7 @@ if [[ "${SKIP_SETUP}" != "true" ]]; then
 
     echo "Loading images into kind cluster..."
     DOCKER_REPO_BASE="${DOCKER_REPO_BASE:-ghcr.io/run-ai/fake-gpu-operator}"
-    for component in dra-plugin-gpu status-updater status-exporter topology-server kwok-dra-plugin compute-domain-controller; do
+    for component in dra-plugin-gpu status-updater status-exporter topology-server kwok-dra-plugin compute-domain-controller compute-domain-dra-plugin; do
         IMAGE="${DOCKER_REPO_BASE}/${component}:${DOCKER_TAG}"
         echo "Loading ${IMAGE}..."
         kind load docker-image \
@@ -82,7 +82,8 @@ if [[ "${SKIP_SETUP}" != "true" ]]; then
         --set statusExporter.image.tag="${DOCKER_TAG}" \
         --set topologyServer.image.tag="${DOCKER_TAG}" \
         --set kwokDraPlugin.image.tag="${DOCKER_TAG}" \
-        --set computeDomainController.image.tag="${DOCKER_TAG}"
+        --set computeDomainController.image.tag="${DOCKER_TAG}" \
+        --set computeDomainDraPlugin.image.tag="${DOCKER_TAG}"
 
     echo "Waiting for status-updater pod to be ready..."
     kubectl wait --for=condition=Ready pod -l app=status-updater -n gpu-operator --timeout=120s
@@ -99,6 +100,9 @@ if [[ "${SKIP_SETUP}" != "true" ]]; then
     echo "Waiting for kwok-dra-plugin pod to be ready..."
     kubectl wait --for=condition=Ready pod -l app=kwok-dra-plugin -n gpu-operator --timeout=120s
 
+    echo "Waiting for compute-domain-dra-plugin daemonset to be ready..."
+    kubectl wait --for=condition=Ready pod -l app=compute-domain-dra-plugin -n gpu-operator --timeout=120s
+
     # Install KWOK controller for simulated nodes
     echo "Installing KWOK controller..."
     KWOK_VERSION="${KWOK_VERSION:-v0.7.0}"

test/integration/values.yaml

@@ -57,6 +57,11 @@ computeDomainController:
   image:
     pullPolicy: Never
 
+computeDomainDraPlugin:
+  enabled: true
+  image:
+    pullPolicy: Never
+
 # GPU topology configuration
 # Status-updater will use this to create topology ConfigMaps
 topology: