Change prometheus metrics type from summary to histogram (antrea-io#1202

)

The summary types are tagged for deprecation, Kubernetes recommended to
use histograms instead of summaries. The main advantages of histogram
types are aggregation and inexpensive.

In this commit, we changed three Antrea controller metrics from summary
to histogram type. They are DurationAppliedToGroupSyncing,
DurationAddressGroupSyncing, and DurationInternalNetworkPolicySyncing.

Fixes antrea-io#905

Co-authored-by: Yongming Ding <[email protected]>

pkg/controller/metrics/prometheus.go

@@ -38,20 +38,20 @@ var (
 		Help:           "The total number of internal-networkpolicy processed",
 		StabilityLevel: metrics.STABLE,
 	})
-	DurationAppliedToGroupSyncing = metrics.NewSummary(&metrics.SummaryOpts{
+	DurationAppliedToGroupSyncing = metrics.NewHistogram(&metrics.HistogramOpts{
 		Name:           "antrea_controller_applied_to_group_sync_duration_milliseconds",
 		Help:           "The duration of syncing applied-to-group",
-		StabilityLevel: metrics.STABLE,
+		StabilityLevel: metrics.ALPHA,
 	})
-	DurationAddressGroupSyncing = metrics.NewSummary(&metrics.SummaryOpts{
+	DurationAddressGroupSyncing = metrics.NewHistogram(&metrics.HistogramOpts{
 		Name:           "antrea_controller_address_group_sync_duration_milliseconds",
 		Help:           "The duration of syncing address-group",
-		StabilityLevel: metrics.STABLE,
+		StabilityLevel: metrics.ALPHA,
 	})
-	DurationInternalNetworkPolicySyncing = metrics.NewSummary(&metrics.SummaryOpts{
+	DurationInternalNetworkPolicySyncing = metrics.NewHistogram(&metrics.HistogramOpts{
 		Name:           "antrea_controller_network_policy_sync_duration_milliseconds",
 		Help:           "The duration of syncing internal-networkpolicy",
-		StabilityLevel: metrics.STABLE,
+		StabilityLevel: metrics.ALPHA,
 	})
 	LengthAppliedToGroupQueue = metrics.NewGauge(&metrics.GaugeOpts{
 		Name:           "antrea_controller_length_applied_to_group_queue",