Roffe/metrics polish (#595)

* update metrics docs & dashboard
* renamed `namespace` label to `svc_namespace` for service metrics as it would be overwritten by most Prometheus setups
* Made histograms for all the controller sync times for better visualization
* added `controller_routes_sync_time`, `controller_bgp_advertisements_sent` & `controller_policy_chains_sync_time` metrics

docs/metrics.md

@@ -22,6 +22,8 @@ The default values unless other specified is
 By enabling [Kubernetes SD](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#<kubernetes_sd_config>) in Prometheus configuration & adding required annotations Prometheus can automaticly discover & scrape kube-router metrics
 
 ## Version notes
+kube-router v0.2.4 received a metrics overhaul where some metrics were changed into histograms, additional metrics was also added. Please make sure you are using the latest dashboard version with versions => v0.2.4
+
 kube-router 0.1.0-rc2 and upwards supports the runtime configuration for controlling where to expose the metrics. If you are using a older version, metrics path & port is locked to `/metrics` & `8080`
 
 ## Supported annotations
@@ -56,14 +58,20 @@ The following metrics is exposed by kube-router prefixed by `kube_router_`
 * controller_bgp_peers
   Number of BGP peers of the instance
 * controller_bgp_advertisements_received
-  Number of total BGP advertisements received since kube-router start
+  Total number of BGP advertisements received since kube-router started
+* controller_bgp_advertisements_sent
+  Total number of BGP advertisements sent since kube-router started
 * controller_bgp_internal_peers_sync_time
   Time it took for the BGP internal peer sync loop to complete
+* controller_routes_sync_time
+  Time it took for controller to sync routes
 
 ### run-firewall=true
 
 * controller_iptables_sync_time
   Time it took for the iptables sync loop to complete
+* controller_policy_chains_sync_time
+  Time it took for controller to sync policy chains
 
 ### run-service-proxy = true
 
@@ -95,7 +103,7 @@ The following metrics is exposed by kube-router prefixed by `kube_router_`
   Outgoing bytes per second
 
 To get a grouped list of CPS for each service a Prometheus query could look like this e.g: 
-`sum(kube_router_service_cps) by (namespace, service_name)`
+`sum(kube_router_service_cps) by (svc_namespace, service_name)`
 
 ## Grafana Dashboard
 

pkg/controllers/netpol/network_policy_controller.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"net"
+	"regexp"
 	"strconv"
 	"strings"
 	"sync"
@@ -26,7 +27,6 @@ import (
 	"k8s.io/client-go/kubernetes"
 	listers "k8s.io/client-go/listers/core/v1"
 	"k8s.io/client-go/tools/cache"
-	"regexp"
 )
 
 const (
@@ -215,7 +215,7 @@ func (npc *NetworkPolicyController) Sync() error {
 	defer func() {
 		endTime := time.Since(start)
 		if npc.MetricsEnabled {
-			metrics.ControllerIptablesSyncTime.WithLabelValues().Set(float64(endTime.Seconds()))
+			metrics.ControllerIptablesSyncTime.Observe(endTime.Seconds())
 		}
 		glog.V(1).Infof("sync iptables took %v", endTime)
 	}()
@@ -258,7 +258,12 @@ func (npc *NetworkPolicyController) Sync() error {
 // policyspec is evaluated to set of matching pods, which are grouped in to a
 // ipset used for source ip addr matching.
 func (npc *NetworkPolicyController) syncNetworkPolicyChains(version string) (map[string]bool, map[string]bool, error) {
-
+	start := time.Now()
+	defer func() {
+		endTime := time.Since(start)
+		metrics.ControllerPolicyChainsSyncTime.Observe(endTime.Seconds())
+		glog.V(2).Infof("Syncing network policy chains took %v", endTime)
+	}()
 	activePolicyChains := make(map[string]bool)
 	activePolicyIpSets := make(map[string]bool)
 
@@ -1536,6 +1541,7 @@ func NewNetworkPolicyController(clientset kubernetes.Interface,
 	if config.MetricsEnabled {
 		//Register the metrics for this controller
 		prometheus.MustRegister(metrics.ControllerIptablesSyncTime)
+		prometheus.MustRegister(metrics.ControllerPolicyChainsSyncTime)
 		npc.MetricsEnabled = true
 	}
 

pkg/controllers/proxy/network_services_controller.go

@@ -373,7 +373,9 @@ func (nsc *NetworkServicesController) publishMetrics(serviceInfoMap serviceInfoM
 	defer func() {
 		endTime := time.Since(start)
 		glog.V(2).Infof("Publishing IPVS metrics took %v", endTime)
-		metrics.ControllerIpvsMetricsExportTime.WithLabelValues().Set(float64(endTime.Seconds()))
+		if nsc.MetricsEnabled {
+			metrics.ControllerIpvsMetricsExportTime.Observe(float64(endTime.Seconds()))
+		}
 	}()
 
 	ipvsSvcs, err := nsc.ln.ipvsGetServices()
@@ -429,7 +431,7 @@ func (nsc *NetworkServicesController) publishMetrics(serviceInfoMap serviceInfoM
 				metrics.ServicePpsIn.WithLabelValues(svc.namespace, svc.name, svcVip, svc.protocol, strconv.Itoa(svc.port)).Set(float64(ipvsSvc.Stats.PPSIn))
 				metrics.ServicePpsOut.WithLabelValues(svc.namespace, svc.name, svcVip, svc.protocol, strconv.Itoa(svc.port)).Set(float64(ipvsSvc.Stats.PPSOut))
 				metrics.ServiceTotalConn.WithLabelValues(svc.namespace, svc.name, svcVip, svc.protocol, strconv.Itoa(svc.port)).Set(float64(ipvsSvc.Stats.Connections))
-				metrics.ControllerIpvsServices.WithLabelValues().Set(float64(len(ipvsSvcs)))
+				metrics.ControllerIpvsServices.Set(float64(len(ipvsSvcs)))
 			}
 		}
 	}
@@ -528,7 +530,7 @@ func (nsc *NetworkServicesController) syncIpvsServices(serviceInfoMap serviceInf
 	defer func() {
 		endTime := time.Since(start)
 		if nsc.MetricsEnabled {
-			metrics.ControllerIpvsServicesSyncTime.WithLabelValues().Set(float64(endTime.Seconds()))
+			metrics.ControllerIpvsServicesSyncTime.Observe(endTime.Seconds())
 		}
 		glog.V(1).Infof("sync ipvs services took %v", endTime)
 	}()

pkg/controllers/routing/bgp_peers.go

@@ -30,7 +30,9 @@ func (nrc *NetworkRoutingController) syncInternalPeers() {
 	start := time.Now()
 	defer func() {
 		endTime := time.Since(start)
-		metrics.ControllerBGPInternalPeersSyncTime.WithLabelValues().Set(float64(endTime.Seconds()))
+		if nrc.MetricsEnabled {
+			metrics.ControllerBGPInternalPeersSyncTime.Observe(endTime.Seconds())
+		}
 		glog.V(2).Infof("Syncing BGP peers for the node took %v", endTime)
 	}()
 
@@ -40,8 +42,9 @@ func (nrc *NetworkRoutingController) syncInternalPeers() {
 		glog.Errorf("Failed to list nodes from API server due to: %s. Can not perform BGP peer sync", err.Error())
 		return
 	}
-
-	metrics.ControllerBPGpeers.WithLabelValues().Set(float64(len(nodes.Items)))
+	if nrc.MetricsEnabled {
+		metrics.ControllerBPGpeers.Set(float64(len(nodes.Items)))
+	}
 	// establish peer and add Pod CIDRs with current set of nodes
 	currentNodes := make([]string, 0)
 	for _, node := range nodes.Items {

pkg/controllers/routing/network_routes_controller.go

@@ -325,7 +325,7 @@ func (nrc *NetworkRoutingController) watchBgpUpdates() {
 			case *gobgp.WatchEventBestPath:
 				glog.V(3).Info("Processing bgp route advertisement from peer")
 				if nrc.MetricsEnabled {
-					metrics.ControllerBGPadvertisementsReceived.WithLabelValues().Add(float64(1))
+					metrics.ControllerBGPadvertisementsReceived.Inc()
 				}
 				for _, path := range msg.PathList {
 					if path.IsLocal() {
@@ -342,6 +342,9 @@ func (nrc *NetworkRoutingController) watchBgpUpdates() {
 }
 
 func (nrc *NetworkRoutingController) advertisePodRoute() error {
+	if nrc.MetricsEnabled {
+		metrics.ControllerBGPadvertisementsSent.Inc()
+	}
 	cidr, err := utils.GetPodCidrFromNodeSpec(nrc.clientset, nrc.hostnameOverride)
 	if err != nil {
 		return err
@@ -486,6 +489,12 @@ func (nrc *NetworkRoutingController) Cleanup() {
 }
 
 func (nrc *NetworkRoutingController) syncNodeIPSets() error {
+	start := time.Now()
+	defer func() {
+		if nrc.MetricsEnabled {
+			metrics.ControllerRoutesSyncTime.Observe(time.Since(start).Seconds())
+		}
+	}()
 	// Get the current list of the nodes from API server
 	nodes, err := nrc.clientset.CoreV1().Nodes().List(metav1.ListOptions{})
 	if err != nil {
@@ -786,6 +795,7 @@ func NewNetworkRoutingController(clientset kubernetes.Interface,
 		prometheus.MustRegister(metrics.ControllerBGPadvertisementsReceived)
 		prometheus.MustRegister(metrics.ControllerBGPInternalPeersSyncTime)
 		prometheus.MustRegister(metrics.ControllerBPGpeers)
+		prometheus.MustRegister(metrics.ControllerRoutesSyncTime)
 		nrc.MetricsEnabled = true
 	}
 

pkg/metrics/metrics_controller.go

@@ -21,108 +21,138 @@ const (
 )
 
 var (
+	// ServiceTotalConn Total incoming connections made
 	ServiceTotalConn = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_total_connections",
 		Help:      "Total incoming connections made",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServicePacketsIn Total incoming packets
 	ServicePacketsIn = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_packets_in",
 		Help:      "Total incoming packets",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServicePacketsOut Total outgoing packets
 	ServicePacketsOut = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_packets_out",
 		Help:      "Total outgoing packets",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServiceBytesIn Total incoming bytes
 	ServiceBytesIn = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_bytes_in",
 		Help:      "Total incoming bytes",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServiceBytesOut Total outgoing bytes
 	ServiceBytesOut = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_bytes_out",
 		Help:      "Total outgoing bytes",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServicePpsIn Incoming packets per second
 	ServicePpsIn = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_pps_in",
 		Help:      "Incoming packets per second",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServicePpsOut Outgoing packets per second
 	ServicePpsOut = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_pps_out",
 		Help:      "Outgoing packets per second",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServiceCPS Service connections per second
 	ServiceCPS = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_cps",
 		Help:      "Service connections per second",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServiceBpsIn Incoming bytes per second
 	ServiceBpsIn = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_bps_in",
 		Help:      "Incoming bytes per second",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ServiceBpsOut Outgoing bytes per second
 	ServiceBpsOut = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "service_bps_out",
 		Help:      "Outgoing bytes per second",
-	}, []string{"namespace", "service_name", "service_vip", "protocol", "port"})
-	ControllerIpvsServices = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	}, []string{"svc_namespace", "service_name", "service_vip", "protocol", "port"})
+	// ControllerIpvsServices Number of ipvs services in the instance
+	ControllerIpvsServices = prometheus.NewGauge(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "controller_ipvs_services",
 		Help:      "Number of ipvs services in the instance",
-	}, []string{})
-	ControllerIptablesSyncTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	})
+	// ControllerIptablesSyncTime Time it took for controller to sync iptables
+	ControllerIptablesSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: namespace,
 		Name:      "controller_iptables_sync_time",
 		Help:      "Time it took for controller to sync iptables",
-	}, []string{})
-	ControllerPublishMetricsTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
-		Namespace: namespace,
-		Name:      "controller_publish_metrics_time",
-		Help:      "Time it took to publish metrics",
-	}, []string{})
-	ControllerIpvsServicesSyncTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	})
+	// ControllerIpvsServicesSyncTime Time it took for controller to sync ipvs services
+	ControllerIpvsServicesSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: namespace,
 		Name:      "controller_ipvs_services_sync_time",
 		Help:      "Time it took for controller to sync ipvs services",
-	}, []string{})
-	ControllerBPGpeers = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	})
+	// ControllerRoutesSyncTime Time it took for controller to sync ipvs services
+	ControllerRoutesSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
+		Namespace: namespace,
+		Name:      "controller_routes_sync_time",
+		Help:      "Time it took for controller to sync routes",
+	})
+	// ControllerBPGpeers BGP peers in the runtime configuration
+	ControllerBPGpeers = prometheus.NewGauge(prometheus.GaugeOpts{
 		Namespace: namespace,
 		Name:      "controller_bgp_peers",
 		Help:      "BGP peers in the runtime configuration",
-	}, []string{})
-	ControllerBGPInternalPeersSyncTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	})
+	// ControllerBGPInternalPeersSyncTime Time it took to sync internal bgp peers
+	ControllerBGPInternalPeersSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: namespace,
 		Name:      "controller_bgp_internal_peers_sync_time",
 		Help:      "Time it took to sync internal bgp peers",
-	}, []string{})
-	ControllerBGPadvertisementsReceived = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+	})
+	// ControllerBGPadvertisementsReceived Time it took to sync internal bgp peers
+	ControllerBGPadvertisementsReceived = prometheus.NewCounter(prometheus.CounterOpts{
 		Namespace: namespace,
 		Name:      "controller_bgp_advertisements_received",
-		Help:      "Time it took to sync internal bgp peers",
-	}, []string{})
-	ControllerIpvsMetricsExportTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Help:      "BGP advertisements received",
+	})
+	// ControllerBGPadvertisementsSent Time it took to sync internal bgp peers
+	ControllerBGPadvertisementsSent = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: namespace,
+		Name:      "controller_bgp_advertisements_sent",
+		Help:      "BGP advertisements sent",
+	})
+	// ControllerIpvsMetricsExportTime Time it took to export metrics
+	ControllerIpvsMetricsExportTime = prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: namespace,
 		Name:      "controller_ipvs_metrics_export_time",
 		Help:      "Time it took to export metrics",
-	}, []string{})
+	})
+	// ControllerPolicyChainsSyncTime Time it took for controller to sync policys
+	ControllerPolicyChainsSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
+		Namespace: namespace,
+		Name:      "controller_policy_chains_sync_time",
+		Help:      "Time it took for controller to sync policy chains",
+	})
 )
 
-// MetricsController Holds settings for the metrics controller
-type MetricsController struct {
+// Controller Holds settings for the metrics controller
+type Controller struct {
 	MetricsPath string
 	MetricsPort uint16
 	mu          sync.Mutex
 	nodeIP      net.IP
 }
 
 // Run prometheus metrics controller
-func (mc *MetricsController) Run(healthChan chan<- *healthcheck.ControllerHeartbeat, stopCh <-chan struct{}, wg *sync.WaitGroup) error {
+func (mc *Controller) Run(healthChan chan<- *healthcheck.ControllerHeartbeat, stopCh <-chan struct{}, wg *sync.WaitGroup) error {
 	t := time.NewTicker(3 * time.Second)
 	defer wg.Done()
 	glog.Info("Starting metrics controller")
@@ -157,8 +187,8 @@ func (mc *MetricsController) Run(healthChan chan<- *healthcheck.ControllerHeartb
 }
 
 // NewMetricsController returns new MetricController object
-func NewMetricsController(clientset kubernetes.Interface, config *options.KubeRouterConfig) (*MetricsController, error) {
-	mc := MetricsController{}
+func NewMetricsController(clientset kubernetes.Interface, config *options.KubeRouterConfig) (*Controller, error) {
+	mc := Controller{}
 	mc.MetricsPath = config.MetricsPath
 	mc.MetricsPort = config.MetricsPort
 	return &mc, nil