kube-proxy implement dual stack metrics

Signed-off-by: Daman Arora <[email protected]>
Co-authored-by: Antonio Ojea <[email protected]>

Author: aojea

pkg/proxy/iptables/proxier.go

@@ -466,7 +466,7 @@ func CleanupLeftovers(ctx context.Context, ipt utiliptables.Interface) (encounte
 		err = ipt.Restore(utiliptables.TableNAT, natLines, utiliptables.NoFlushTables, utiliptables.RestoreCounters)
 		if err != nil {
 			logger.Error(err, "Failed to execute iptables-restore", "table", utiliptables.TableNAT)
-			metrics.IPTablesRestoreFailuresTotal.Inc()
+			metrics.IPTablesRestoreFailuresTotal.WithLabelValues(string(ipt.Protocol())).Inc()
 			encounteredError = true
 		}
 	}
@@ -493,7 +493,7 @@ func CleanupLeftovers(ctx context.Context, ipt utiliptables.Interface) (encounte
 		// Write it.
 		if err := ipt.Restore(utiliptables.TableFilter, filterLines, utiliptables.NoFlushTables, utiliptables.RestoreCounters); err != nil {
 			logger.Error(err, "Failed to execute iptables-restore", "table", utiliptables.TableFilter)
-			metrics.IPTablesRestoreFailuresTotal.Inc()
+			metrics.IPTablesRestoreFailuresTotal.WithLabelValues(string(ipt.Protocol())).Inc()
 			encounteredError = true
 		}
 	}
@@ -527,7 +527,7 @@ func (proxier *Proxier) Sync() {
 	if proxier.healthzServer != nil {
 		proxier.healthzServer.QueuedUpdate(proxier.ipFamily)
 	}
-	metrics.SyncProxyRulesLastQueuedTimestamp.SetToCurrentTime()
+	metrics.SyncProxyRulesLastQueuedTimestamp.WithLabelValues(string(proxier.ipFamily)).SetToCurrentTime()
 	proxier.syncRunner.Run()
 }
 
@@ -539,7 +539,7 @@ func (proxier *Proxier) SyncLoop() {
 	}
 
 	// synthesize "last change queued" time as the informers are syncing.
-	metrics.SyncProxyRulesLastQueuedTimestamp.SetToCurrentTime()
+	metrics.SyncProxyRulesLastQueuedTimestamp.WithLabelValues(string(proxier.ipFamily)).SetToCurrentTime()
 	proxier.syncRunner.Loop(wait.NeverStop)
 }
 
@@ -813,11 +813,11 @@ func (proxier *Proxier) syncProxyRules() {
 	// Keep track of how long syncs take.
 	start := time.Now()
 	defer func() {
-		metrics.SyncProxyRulesLatency.Observe(metrics.SinceInSeconds(start))
+		metrics.SyncProxyRulesLatency.WithLabelValues(string(proxier.ipFamily)).Observe(metrics.SinceInSeconds(start))
 		if tryPartialSync {
-			metrics.SyncPartialProxyRulesLatency.Observe(metrics.SinceInSeconds(start))
+			metrics.SyncPartialProxyRulesLatency.WithLabelValues(string(proxier.ipFamily)).Observe(metrics.SinceInSeconds(start))
 		} else {
-			metrics.SyncFullProxyRulesLatency.Observe(metrics.SinceInSeconds(start))
+			metrics.SyncFullProxyRulesLatency.WithLabelValues(string(proxier.ipFamily)).Observe(metrics.SinceInSeconds(start))
 		}
 		proxier.logger.V(2).Info("SyncProxyRules complete", "elapsed", time.Since(start))
 	}()
@@ -833,7 +833,7 @@ func (proxier *Proxier) syncProxyRules() {
 			proxier.logger.Info("Sync failed", "retryingTime", proxier.syncPeriod)
 			proxier.syncRunner.RetryAfter(proxier.syncPeriod)
 			if tryPartialSync {
-				metrics.IPTablesPartialRestoreFailuresTotal.Inc()
+				metrics.IPTablesPartialRestoreFailuresTotal.WithLabelValues(string(proxier.ipFamily)).Inc()
 			}
 			// proxier.serviceChanges and proxier.endpointChanges have already
 			// been flushed, so we've lost the state needed to be able to do
@@ -1528,10 +1528,10 @@ func (proxier *Proxier) syncProxyRules() {
 		"-j", "ACCEPT",
 	)
 
-	metrics.IPTablesRulesTotal.WithLabelValues(string(utiliptables.TableFilter)).Set(float64(proxier.filterRules.Lines()))
-	metrics.IPTablesRulesLastSync.WithLabelValues(string(utiliptables.TableFilter)).Set(float64(proxier.filterRules.Lines()))
-	metrics.IPTablesRulesTotal.WithLabelValues(string(utiliptables.TableNAT)).Set(float64(proxier.natRules.Lines() + skippedNatRules.Lines() - deletedChains))
-	metrics.IPTablesRulesLastSync.WithLabelValues(string(utiliptables.TableNAT)).Set(float64(proxier.natRules.Lines() - deletedChains))
+	metrics.IPTablesRulesTotal.WithLabelValues(string(utiliptables.TableFilter), string(proxier.ipFamily)).Set(float64(proxier.filterRules.Lines()))
+	metrics.IPTablesRulesLastSync.WithLabelValues(string(utiliptables.TableFilter), string(proxier.ipFamily)).Set(float64(proxier.filterRules.Lines()))
+	metrics.IPTablesRulesTotal.WithLabelValues(string(utiliptables.TableNAT), string(proxier.ipFamily)).Set(float64(proxier.natRules.Lines() + skippedNatRules.Lines() - deletedChains))
+	metrics.IPTablesRulesLastSync.WithLabelValues(string(utiliptables.TableNAT), string(proxier.ipFamily)).Set(float64(proxier.natRules.Lines() - deletedChains))
 
 	// Sync rules.
 	proxier.iptablesData.Reset()
@@ -1563,7 +1563,7 @@ func (proxier *Proxier) syncProxyRules() {
 		} else {
 			proxier.logger.Error(err, "Failed to execute iptables-restore")
 		}
-		metrics.IPTablesRestoreFailuresTotal.Inc()
+		metrics.IPTablesRestoreFailuresTotal.WithLabelValues(string(proxier.ipFamily)).Inc()
 		return
 	}
 	success = true
@@ -1572,17 +1572,17 @@ func (proxier *Proxier) syncProxyRules() {
 	for name, lastChangeTriggerTimes := range endpointUpdateResult.LastChangeTriggerTimes {
 		for _, lastChangeTriggerTime := range lastChangeTriggerTimes {
 			latency := metrics.SinceInSeconds(lastChangeTriggerTime)
-			metrics.NetworkProgrammingLatency.Observe(latency)
+			metrics.NetworkProgrammingLatency.WithLabelValues(string(proxier.ipFamily)).Observe(latency)
 			proxier.logger.V(4).Info("Network programming", "endpoint", klog.KRef(name.Namespace, name.Name), "elapsed", latency)
 		}
 	}
 
-	metrics.SyncProxyRulesNoLocalEndpointsTotal.WithLabelValues("internal").Set(float64(serviceNoLocalEndpointsTotalInternal))
-	metrics.SyncProxyRulesNoLocalEndpointsTotal.WithLabelValues("external").Set(float64(serviceNoLocalEndpointsTotalExternal))
+	metrics.SyncProxyRulesNoLocalEndpointsTotal.WithLabelValues("internal", string(proxier.ipFamily)).Set(float64(serviceNoLocalEndpointsTotalInternal))
+	metrics.SyncProxyRulesNoLocalEndpointsTotal.WithLabelValues("external", string(proxier.ipFamily)).Set(float64(serviceNoLocalEndpointsTotalExternal))
 	if proxier.healthzServer != nil {
 		proxier.healthzServer.Updated(proxier.ipFamily)
 	}
-	metrics.SyncProxyRulesLastTimestamp.SetToCurrentTime()
+	metrics.SyncProxyRulesLastTimestamp.WithLabelValues(string(proxier.ipFamily)).SetToCurrentTime()
 
 	// Update service healthchecks.  The endpoints list might include services that are
 	// not "OnlyLocal", but the services list will not, and the serviceHealthServer