Merge pull request #73549 from haiyanmeng/runtimeclass

Add monitoring for RuntimeClass

pkg/kubelet/kuberuntime/instrumented_services.go

@@ -178,10 +178,15 @@ func (in instrumentedRuntimeService) Attach(req *runtimeapi.AttachRequest) (*run
 
 func (in instrumentedRuntimeService) RunPodSandbox(config *runtimeapi.PodSandboxConfig, runtimeHandler string) (string, error) {
 	const operation = "run_podsandbox"
-	defer recordOperation(operation, time.Now())
+	startTime := time.Now()
+	defer recordOperation(operation, startTime)
+	defer metrics.RunPodSandboxLatencies.WithLabelValues(runtimeHandler).Observe(metrics.SinceInMicroseconds(startTime))
 
 	out, err := in.service.RunPodSandbox(config, runtimeHandler)
 	recordError(operation, err)
+	if err != nil {
+		metrics.RunPodSandboxErrors.WithLabelValues(runtimeHandler).Inc()
+	}
 	return out, err
 }
 

pkg/kubelet/metrics/metrics.go

@@ -65,6 +65,10 @@ const (
 	ConfigUIDLabelKey             = "node_config_uid"
 	ConfigResourceVersionLabelKey = "node_config_resource_version"
 	KubeletConfigKeyLabelKey      = "node_config_kubelet_key"
+
+	// Metrics keys for RuntimeClass
+	RunPodSandboxLatenciesKey = "run_podsandbox_latencies"
+	RunPodSandboxErrorsKey    = "run_podsandbox_errors"
 )
 
 var (
@@ -210,6 +214,22 @@ var (
 			Help:      "This metric is true (1) if the node is experiencing a configuration-related error, false (0) otherwise.",
 		},
 	)
+	RunPodSandboxLatencies = prometheus.NewSummaryVec(
+		prometheus.SummaryOpts{
+			Subsystem: KubeletSubsystem,
+			Name:      RunPodSandboxLatenciesKey,
+			Help:      "Latencies in microseconds of the run_podsandbox operations. Broken down by RuntimeClass.",
+		},
+		[]string{"runtime_handler"},
+	)
+	RunPodSandboxErrors = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Subsystem: KubeletSubsystem,
+			Name:      RunPodSandboxErrorsKey,
+			Help:      "Cumulative number of the run_podsandbox operation errors by RuntimeClass.",
+		},
+		[]string{"runtime_handler"},
+	)
 )
 
 var registerMetrics sync.Once