Add service and service monitor manifests

Signed-off-by: ghokun <[email protected]>

Author: ghokun

manifests/device-plugin.yml

@@ -21,13 +21,13 @@ metadata:
 spec:
   selector:
     matchLabels:
-      name: kuartis-virtual-gpu-device-plugin
+      app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin
   updateStrategy:
     type: RollingUpdate
   template:
     metadata:
       labels:
-        name: kuartis-virtual-gpu-device-plugin
+        app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin
     spec:
       hostIPC: true
       nodeSelector:
@@ -45,7 +45,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
       containers:
-        - image: ghcr.io/kuartis/kuartis-virtual-gpu-device-plugin:0.4.10
+        - image: ghcr.io/kuartis/kuartis-virtual-gpu-device-plugin:0.5.0
           name: kuartis-virtual-gpu-device-plugin-ctr
           command:
             - /usr/bin/virtual-gpu-device-plugin
@@ -89,3 +89,40 @@ spec:
         - name: dockershimsock
           hostPath:
             path: /var/run/dockershim.sock
+---
+apiVersion: v1
+kind: Service
+metadata:
+  labels:
+    app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin
+  name: kuartis-virtual-gpu-device-plugin
+  namespace: kube-system
+spec:
+  ports:
+    - name: metrics
+      port: 8080
+      targetPort: 8080
+      protocol: TCP
+  selector:
+    app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin
+  sessionAffinity: None
+  type: ClusterIP
+---
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  labels:
+    app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin
+  name: kuartis-virtual-gpu-device-plugin
+  namespace: kube-system
+spec:
+  endpoints:
+    - interval: 15s
+      path: /metrics
+      port: metrics
+  namespaceSelector:
+    matchNames:
+      - kube-system
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: kuartis-virtual-gpu-device-plugin

pkg/gpu/nvidia/metrics.go

@@ -26,8 +26,7 @@ const (
 
 var node = os.Getenv("NODE_NAME")
 
-var metricsFormat = `
-# HELP gpu_memory_usage_per_container Shows the GPU memory usage per container.
+var metricsFormat = `# HELP gpu_memory_usage_per_container Shows the GPU memory usage per container.
 # TYPE gpu_memory_usage_per_container gauge
 {{- range $m := . }}
 gpu_memory_usage_per_container{pid="{{ $m.Pid }}",gpuindex="{{ $m.GpuIndex }}",gpuuuid="{{ $m.GpuUUID }}",node="{{ $m.Node }}",namespace="{{ $m.Namespace }}",pod="{{ $m.Pod }}",poduid="{{ $m.PodUid }}",container="{{ $m.Container }}",containerid="{{ $m.ContainerId }}"} {{ $m.UsedGpuMemory }}
@@ -86,7 +85,6 @@ func collectMetrics(w http.ResponseWriter, r *http.Request) {
 			ContainerId: container.GetId(),
 		}
 	}
-	log.Printf("Current map %+v", containerMap)
 	collected := []metric{}
 	for i := 0; i < getDeviceCount(); i++ {
 		d, ret := nvml.DeviceGetHandleByIndex(i)
@@ -96,7 +94,7 @@ func collectMetrics(w http.ResponseWriter, r *http.Request) {
 		log.Printf("Found %d processes on GPU %d", len(processes), i)
 		for _, process := range processes {
 			containerId := getContainerId(process.Pid)
-			if container, ok := containerMap[strings.TrimSpace(containerId)]; ok {
+			if container, ok := containerMap[containerId]; ok {
 				log.Printf("Using %s Found container %+v for process: %d", containerId, container, process.Pid)
 				collected = append(collected, metric{
 					Pid:           process.Pid,
@@ -145,6 +143,5 @@ func getContainerId(pid uint32) string {
 	}
 	proc := string(data)
 	containerId := proc[strings.LastIndex(proc, "/")+1:]
-	log.Printf("Found container id %s for process: %d", containerId, pid)
-	return containerId
+	return strings.TrimSpace(containerId)
 }