kubelet: add operations count and error count metrics to network plugin manager

Author: AnishShah

pkg/kubelet/dockershim/network/metrics/metrics.go

@@ -28,9 +28,11 @@ import (
 
 const (
 	// NetworkPluginOperationsKey is the key for operation count metrics.
-	NetworkPluginOperationsKey = "network_plugin_operations"
+	NetworkPluginOperationsKey = "network_plugin_operations_total"
 	// NetworkPluginOperationsLatencyKey is the key for the operation latency metrics.
 	NetworkPluginOperationsLatencyKey = "network_plugin_operations_duration_seconds"
+	// NetworkPluginOperationsErrorsKey is the key for the operations error metrics.
+	NetworkPluginOperationsErrorsKey = "network_plugin_operations_errors_total"
 
 	// Keep the "kubelet" subsystem for backward compatibility.
 	kubeletSubsystem = "kubelet"
@@ -49,6 +51,28 @@ var (
 		},
 		[]string{"operation_type"},
 	)
+
+	// NetworkPluginOperations collects operation counts by operation type.
+	NetworkPluginOperations = metrics.NewCounterVec(
+		&metrics.CounterOpts{
+			Subsystem:      kubeletSubsystem,
+			Name:           NetworkPluginOperationsKey,
+			Help:           "Cumulative number of network plugin operations by operation type.",
+			StabilityLevel: metrics.ALPHA,
+		},
+		[]string{"operation_type"},
+	)
+
+	// NetworkPluginOperationsErrors collects operation errors by operation type.
+	NetworkPluginOperationsErrors = metrics.NewCounterVec(
+		&metrics.CounterOpts{
+			Subsystem:      kubeletSubsystem,
+			Name:           NetworkPluginOperationsErrorsKey,
+			Help:           "Cumulative number of network plugin operation errors by operation type.",
+			StabilityLevel: metrics.ALPHA,
+		},
+		[]string{"operation_type"},
+	)
 )
 
 var registerMetrics sync.Once
@@ -57,6 +81,8 @@ var registerMetrics sync.Once
 func Register() {
 	registerMetrics.Do(func() {
 		legacyregistry.MustRegister(NetworkPluginOperationsLatency)
+		legacyregistry.MustRegister(NetworkPluginOperations)
+		legacyregistry.MustRegister(NetworkPluginOperationsErrors)
 	})
 }
 

pkg/kubelet/dockershim/network/plugins.go

@@ -382,45 +382,57 @@ func (pm *PluginManager) podUnlock(fullPodName string) {
 
 // recordOperation records operation and duration
 func recordOperation(operation string, start time.Time) {
+	metrics.NetworkPluginOperations.WithLabelValues(operation).Inc()
 	metrics.NetworkPluginOperationsLatency.WithLabelValues(operation).Observe(metrics.SinceInSeconds(start))
 }
 
+// recordError records errors for metric.
+func recordError(operation string) {
+	metrics.NetworkPluginOperationsErrors.WithLabelValues(operation).Inc()
+}
+
 func (pm *PluginManager) GetPodNetworkStatus(podNamespace, podName string, id kubecontainer.ContainerID) (*PodNetworkStatus, error) {
-	defer recordOperation("get_pod_network_status", time.Now())
+	const operation = "get_pod_network_status"
+	defer recordOperation(operation, time.Now())
 	fullPodName := kubecontainer.BuildPodFullName(podName, podNamespace)
 	pm.podLock(fullPodName).Lock()
 	defer pm.podUnlock(fullPodName)
 
 	netStatus, err := pm.plugin.GetPodNetworkStatus(podNamespace, podName, id)
 	if err != nil {
+		recordError(operation)
 		return nil, fmt.Errorf("networkPlugin %s failed on the status hook for pod %q: %v", pm.plugin.Name(), fullPodName, err)
 	}
 
 	return netStatus, nil
 }
 
 func (pm *PluginManager) SetUpPod(podNamespace, podName string, id kubecontainer.ContainerID, annotations, options map[string]string) error {
-	defer recordOperation("set_up_pod", time.Now())
+	const operation = "set_up_pod"
+	defer recordOperation(operation, time.Now())
 	fullPodName := kubecontainer.BuildPodFullName(podName, podNamespace)
 	pm.podLock(fullPodName).Lock()
 	defer pm.podUnlock(fullPodName)
 
 	klog.V(3).Infof("Calling network plugin %s to set up pod %q", pm.plugin.Name(), fullPodName)
 	if err := pm.plugin.SetUpPod(podNamespace, podName, id, annotations, options); err != nil {
+		recordError(operation)
 		return fmt.Errorf("networkPlugin %s failed to set up pod %q network: %v", pm.plugin.Name(), fullPodName, err)
 	}
 
 	return nil
 }
 
 func (pm *PluginManager) TearDownPod(podNamespace, podName string, id kubecontainer.ContainerID) error {
-	defer recordOperation("tear_down_pod", time.Now())
+	const operation = "tear_down_pod"
+	defer recordOperation(operation, time.Now())
 	fullPodName := kubecontainer.BuildPodFullName(podName, podNamespace)
 	pm.podLock(fullPodName).Lock()
 	defer pm.podUnlock(fullPodName)
 
 	klog.V(3).Infof("Calling network plugin %s to tear down pod %q", pm.plugin.Name(), fullPodName)
 	if err := pm.plugin.TearDownPod(podNamespace, podName, id); err != nil {
+		recordError(operation)
 		return fmt.Errorf("networkPlugin %s failed to teardown pod %q network: %v", pm.plugin.Name(), fullPodName, err)
 	}