Eliranw/run 28124 utilization metrics (#148)

Add check for workloadKind to generate utilizationmetrics

Author: eliranw

internal/status-updater/handlers/pod/gpu_usage_calculator.go

@@ -71,7 +71,7 @@ func calculateGpuUsageFromPodType(dynamicclient dynamic.Interface, pod *v1.Pod,
 	switch podType {
 	case "train":
 		return generateGpuUsageStatus(topology.Range{Min: 80, Max: 100}, gpuFraction, totalGpuMemory, false)
-	case "build", "interactive-preemptible":
+	case "build", "interactive-preemptible", "interactive", "distributed":
 		return generateGpuUsageStatus(topology.Range{Min: 0, Max: 0}, gpuFraction, totalGpuMemory, false)
 	case "inference":
 		return generateGpuUsageStatus(topology.Range{Min: 0, Max: 0}, gpuFraction, totalGpuMemory, true)
@@ -104,6 +104,20 @@ func calculateUtilizationFromAnnotation(annotationValue string) (*topology.Range
 }
 
 func getPodType(dynamicClient dynamic.Interface, pod *v1.Pod) (string, error) {
+	if workloadKind, ok := pod.Labels["workloadKind"]; ok {
+		switch workloadKind {
+		case "TrainingWorkload":
+			return "train", nil
+		case "DistributedWorkload":
+			return "distributed", nil
+		case "InferenceWorkload":
+			return "inference", nil
+		case "InteractiveWorkload":
+			return "interactive", nil
+		}
+	}
+
+	// Fallback to existing PodGroup lookup
 	podGroupName := pod.Annotations[constants.AnnotationPodGroupName]
 	if podGroupName == "" {
 		return "", fmt.Errorf("pod %s has no constants.PodGroupNameAnnotation annotation", pod.Name)

test/integration/kind-cluster-config.yaml

@@ -3,13 +3,6 @@ apiVersion: kind.x-k8s.io/v1alpha4
 name: dra-cluster
 featureGates:
   DynamicResourceAllocation: true
-containerdConfigPatches:
-# Enable CDI as described in
-# https://tags.cncf.io/container-device-interface#containerd-configuration
-- |-
-  [plugins."io.containerd.grpc.v1.cri"]
-    enable_cdi = true
-    cdi_spec_dirs = ["/etc/cdi", "/var/run/cdi"]
 nodes:
 - role: control-plane
   image: kindest/node:v1.34.0
@@ -19,22 +12,3 @@ nodes:
     apiServer:
       extraArgs:
         runtime-config: "resource.k8s.io/v1beta1=true"
-    scheduler:
-      extraArgs:
-        v: "1"
-    controllerManager:
-      extraArgs:
-        v: "1"
-  - |
-    kind: InitConfiguration
-    nodeRegistration:
-      kubeletExtraArgs:
-        v: "4"
-- role: worker
-  image: kindest/node:v1.34.0
-  kubeadmConfigPatches:
-  - |
-    kind: JoinConfiguration
-    nodeRegistration:
-      kubeletExtraArgs:
-        v: "4"