Add Prometheus metrics to count ANP and ACNP Status updates (#1801)

Too frequent Status updates could generate too many versions of the CRD,
that would need to be stored in etcd until the next compaction by
kube-apiserver. Too many updates could also cause fragmentation of the
database. It is useful to have access to the number of updates over time
in production clusters.

docs/prometheus-integration.md

@@ -172,10 +172,14 @@ tables.
 
 #### Antrea Controller Metrics
 
+- **antrea_controller_acnp_status_updates:** The total number of actual
+status updates performed for Antrea ClusterNetworkPolicy Custom Resources
 - **antrea_controller_address_group_processed:** The total number of
 address-group processed
 - **antrea_controller_address_group_sync_duration_milliseconds:** The duration
 of syncing address-group
+- **antrea_controller_anp_status_updates:** The total number of actual status
+updates performed for Antrea NetworkPolicy Custom Resources
 - **antrea_controller_applied_to_group_processed:** The total number of
 applied-to-group processed
 - **antrea_controller_applied_to_group_sync_duration_milliseconds:** The

pkg/controller/metrics/prometheus.go

@@ -89,6 +89,20 @@ var (
 		Help:           "The length of InternalNetworkPolicyQueue",
 		StabilityLevel: metrics.STABLE,
 	})
+	AntreaNetworkPolicyStatusUpdates = metrics.NewCounter(&metrics.CounterOpts{
+		Namespace:      metricNamespaceAntrea,
+		Subsystem:      metricSubsystemController,
+		Name:           "anp_status_updates",
+		Help:           "The total number of actual status updates performed for Antrea NetworkPolicy Custom Resources",
+		StabilityLevel: metrics.ALPHA,
+	})
+	AntreaClusterNetworkPolicyStatusUpdates = metrics.NewCounter(&metrics.CounterOpts{
+		Namespace:      metricNamespaceAntrea,
+		Subsystem:      metricSubsystemController,
+		Name:           "acnp_status_updates",
+		Help:           "The total number of actual status updates performed for Antrea ClusterNetworkPolicy Custom Resources",
+		StabilityLevel: metrics.ALPHA,
+	})
 )
 
 // Initialize Prometheus metrics collection.
@@ -122,4 +136,10 @@ func InitializePrometheusMetrics() {
 	if err := legacyregistry.Register(LengthInternalNetworkPolicyQueue); err != nil {
 		klog.Errorf("Failed to register antrea_controller_length_network_policy_queue with Prometheus: %s", err.Error())
 	}
+	if err := legacyregistry.Register(AntreaNetworkPolicyStatusUpdates); err != nil {
+		klog.Errorf("Failed to register antrea_controller_anp_status_updates with Prometheus: %s", err.Error())
+	}
+	if err := legacyregistry.Register(AntreaClusterNetworkPolicyStatusUpdates); err != nil {
+		klog.Errorf("Failed to register antrea_controller_acnp_status_updates with Prometheus: %s", err.Error())
+	}
 }

pkg/controller/networkpolicy/status_controller.go

@@ -34,6 +34,7 @@ import (
 	antreaclientset "github.com/vmware-tanzu/antrea/pkg/client/clientset/versioned"
 	secinformers "github.com/vmware-tanzu/antrea/pkg/client/informers/externalversions/security/v1alpha1"
 	seclisters "github.com/vmware-tanzu/antrea/pkg/client/listers/security/v1alpha1"
+	"github.com/vmware-tanzu/antrea/pkg/controller/metrics"
 	antreatypes "github.com/vmware-tanzu/antrea/pkg/controller/types"
 )
 
@@ -328,6 +329,7 @@ func (c *networkPolicyControl) UpdateAntreaNetworkPolicyStatus(namespace, name s
 	if anp.Status == *status {
 		return nil
 	}
+	metrics.AntreaNetworkPolicyStatusUpdates.Inc()
 	toUpdate := anp.DeepCopy()
 	toUpdate.Status = *status
 	_, err = c.antreaClient.SecurityV1alpha1().NetworkPolicies(namespace).UpdateStatus(context.TODO(), toUpdate, v1.UpdateOptions{})
@@ -344,6 +346,7 @@ func (c *networkPolicyControl) UpdateAntreaClusterNetworkPolicyStatus(name strin
 	if cnp.Status == *status {
 		return nil
 	}
+	metrics.AntreaClusterNetworkPolicyStatusUpdates.Inc()
 	toUpdate := cnp.DeepCopy()
 	toUpdate.Status = *status
 	_, err = c.antreaClient.SecurityV1alpha1().ClusterNetworkPolicies().UpdateStatus(context.TODO(), toUpdate, v1.UpdateOptions{})