add node ping total count metric

docs/ovn-ovs-monitor.md

@@ -87,10 +87,12 @@ OVS_Monitor | | ovsdb/vswitchd metrics
  14 | pinger_external_dns_latency_ms | The latency ms histogram the node request external dns
  15 | pinger_pod_ping_latency_ms | The latency ms histogram for pod peer ping
  16 | pinger_pod_ping_lost_total | The lost count for pod peer ping
- 17 | pinger_node_ping_latency_ms | The latency ms histogram for pod ping node
- 18 | pinger_node_ping_lost_total | The lost count for pod ping node
- 19 | pinger_external_ping_latency_ms | The latency ms histogram for pod ping external address
- 20 | pinger_node_external_lost_total | The lost count for pod ping external address
+ 17 | pinger_pod_ping_count_total | The total count for pod peer ping
+ 18 | pinger_node_ping_latency_ms | The latency ms histogram for pod ping node
+ 19 | pinger_node_ping_lost_total | The lost count for pod ping node
+ 20 | pinger_node_ping_count_total | The total count for pod ping node
+ 21 | pinger_external_ping_latency_ms | The latency ms histogram for pod ping external address
+ 22 | pinger_node_external_lost_total | The lost count for pod ping external address
  Kube-OVN-Controller | | Controller metrics
  1 | rest_client_request_latency_seconds | Request latency in seconds. Broken down by verb and URL
  2 | rest_client_requests_total | Number of HTTP requests, partitioned by status code, method, and host

pkg/pinger/metrics.go

@@ -144,6 +144,18 @@ var (
 			"target_node_ip",
 			"target_pod_ip",
 		})
+	podPingTotalCounter = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "pinger_pod_ping_count_total",
+			Help: "The total count for pod peer ping",
+		}, []string{
+			"src_node_name",
+			"src_node_ip",
+			"src_pod_ip",
+			"target_node_name",
+			"target_node_ip",
+			"target_pod_ip",
+		})
 	nodePingLatencyHistogram = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Name:    "pinger_node_ping_latency_ms",
@@ -168,6 +180,17 @@ var (
 			"target_node_name",
 			"target_node_ip",
 		})
+	nodePingTotalCounter = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "pinger_node_ping_count_total",
+			Help: "The total count for pod ping node",
+		}, []string{
+			"src_node_name",
+			"src_node_ip",
+			"src_pod_ip",
+			"target_node_name",
+			"target_node_ip",
+		})
 	externalPingLatencyHistogram = prometheus.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Name:    "pinger_external_ping_latency_ms",
@@ -617,8 +640,10 @@ func init() {
 	prometheus.MustRegister(externalDnsRequestLatencyHistogram)
 	prometheus.MustRegister(podPingLatencyHistogram)
 	prometheus.MustRegister(podPingLostCounter)
+	prometheus.MustRegister(podPingTotalCounter)
 	prometheus.MustRegister(nodePingLatencyHistogram)
 	prometheus.MustRegister(nodePingLostCounter)
+	prometheus.MustRegister(nodePingTotalCounter)
 	prometheus.MustRegister(externalPingLatencyHistogram)
 	prometheus.MustRegister(externalPingLostCounter)
 
@@ -720,7 +745,7 @@ func SetExternalDnsUnhealthyMetrics(nodeName string) {
 	externalDnsUnhealthyGauge.WithLabelValues(nodeName).Set(1)
 }
 
-func SetPodPingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, targetNodeIP, targetPodIP string, latency float64, lost int) {
+func SetPodPingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, targetNodeIP, targetPodIP string, latency float64, lost, total int) {
 	podPingLatencyHistogram.WithLabelValues(
 		srcNodeName,
 		srcNodeIP,
@@ -737,9 +762,17 @@ func SetPodPingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, targetN
 		targetNodeIP,
 		targetPodIP,
 	).Add(float64(lost))
+	podPingTotalCounter.WithLabelValues(
+		srcNodeName,
+		srcNodeIP,
+		srcPodIP,
+		targetNodeName,
+		targetNodeIP,
+		targetPodIP,
+	).Add(float64(total))
 }
 
-func SetNodePingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, targetNodeIP string, latency float64, lost int) {
+func SetNodePingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, targetNodeIP string, latency float64, lost, total int) {
 	nodePingLatencyHistogram.WithLabelValues(
 		srcNodeName,
 		srcNodeIP,
@@ -754,6 +787,13 @@ func SetNodePingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetNodeName, target
 		targetNodeName,
 		targetNodeIP,
 	).Add(float64(lost))
+	nodePingTotalCounter.WithLabelValues(
+		srcNodeName,
+		srcNodeIP,
+		srcPodIP,
+		targetNodeName,
+		targetNodeIP,
+	).Add(float64(total))
 }
 
 func SetExternalPingMetrics(srcNodeName, srcNodeIP, srcPodIP, targetAddress string, latency float64, lost int) {

pkg/pinger/ping.go

@@ -103,7 +103,8 @@ func pingNodes(config *Configuration) error {
 						config.PodName,
 						no.Name, addr.Address,
 						float64(stats.AvgRtt)/float64(time.Millisecond),
-						int(math.Abs(float64(stats.PacketsSent-stats.PacketsRecv))))
+						int(math.Abs(float64(stats.PacketsSent-stats.PacketsRecv))),
+						int(float64(stats.PacketsSent)))
 				}(addr.Address, no.Name)
 			}
 		}
@@ -154,7 +155,8 @@ func pingPods(config *Configuration) error {
 					nodeIP,
 					podIp,
 					float64(stats.AvgRtt)/float64(time.Millisecond),
-					int(math.Abs(float64(stats.PacketsSent-stats.PacketsRecv))))
+					int(math.Abs(float64(stats.PacketsSent-stats.PacketsRecv))),
+					int(float64(stats.PacketsSent)))
 			}(pod.Status.PodIP, pod.Name, pod.Status.HostIP, pod.Spec.NodeName)
 		}
 	}