feat(gobgp): add kube_router_bgp_peer_info metric

Replace the misleading kube_router_controller_bgp_peers gauge
  which only counts 'cluster nodes'. with a new per peer metric
  kube_router_bgp_peer_info with 'GaugeVec' that exposes actual
  BGP session state from gobgp. labels include peer address, asn,
  and session state. metric value is 1 if established and 0 otherwise.

Closes: #848
Signed-off-by: Roman Kuzmitskii <roman@damex.org>

Author: damex

pkg/controllers/routing/bgp_peers.go

@@ -88,9 +88,6 @@ func (nrc *NetworkRoutingController) syncInternalPeers() {
 	// get the current list of the nodes from API server
 	nodes := nrc.nodeLister.List()
 
-	if nrc.MetricsEnabled {
-		metrics.ControllerBPGpeers.Set(float64(len(nodes)))
-	}
 	// establish peer and add Pod CIDRs with current set of nodes
 	currentNodes := make([]string, 0)
 	for _, obj := range nodes {

pkg/controllers/routing/network_routes_controller.go

@@ -436,6 +436,10 @@ func (nrc *NetworkRoutingController) Run(
 			nrc.syncInternalPeers()
 		}
 
+		if nrc.MetricsEnabled {
+			nrc.updateBGPPeerMetrics()
+		}
+
 		if err == nil {
 			healthcheck.SendHeartBeat(healthChan, healthcheck.NetworkRoutesController)
 		} else {
@@ -729,6 +733,47 @@ func (nrc *NetworkRoutingController) injectRoute(path *apiutil.Path) error {
 	return nrc.routeSyncer.SyncLocalRouteTable()
 }
 
+func bgpSessionStateString(state gobgpapi.PeerState_SessionState) string {
+	switch state {
+	case gobgpapi.PeerState_SESSION_STATE_UNSPECIFIED:
+		return "unspecified"
+	case gobgpapi.PeerState_SESSION_STATE_IDLE:
+		return "idle"
+	case gobgpapi.PeerState_SESSION_STATE_CONNECT:
+		return "connect"
+	case gobgpapi.PeerState_SESSION_STATE_ACTIVE:
+		return "active"
+	case gobgpapi.PeerState_SESSION_STATE_OPENSENT:
+		return "opensent"
+	case gobgpapi.PeerState_SESSION_STATE_OPENCONFIRM:
+		return "openconfirm"
+	case gobgpapi.PeerState_SESSION_STATE_ESTABLISHED:
+		return "established"
+	default:
+		return "unknown"
+	}
+}
+
+func (nrc *NetworkRoutingController) updateBGPPeerMetrics() {
+	metrics.ControllerBGPPeerInfo.Reset()
+	err := nrc.bgpServer.ListPeer(context.Background(), &gobgpapi.ListPeerRequest{}, func(peer *gobgpapi.Peer) {
+		if peer.Conf == nil || peer.State == nil {
+			return
+		}
+		addr := peer.Conf.NeighborAddress
+		asn := strconv.FormatUint(uint64(peer.Conf.PeerAsn), 10)
+		state := bgpSessionStateString(peer.State.SessionState)
+		var value float64
+		if peer.State.SessionState == gobgpapi.PeerState_SESSION_STATE_ESTABLISHED {
+			value = 1
+		}
+		metrics.ControllerBGPPeerInfo.WithLabelValues(addr, asn, state).Set(value)
+	})
+	if err != nil {
+		klog.Errorf("error updating BGP peer metrics: %v", err)
+	}
+}
+
 func (nrc *NetworkRoutingController) isPeerEstablished(peerIP string) (bool, error) {
 	var peerConnected bool
 	peerFunc := func(peer *gobgpapi.Peer) {
@@ -1227,7 +1272,7 @@ func NewNetworkRoutingController(clientset kubernetes.Interface,
 		metrics.DefaultRegisterer.MustRegister(metrics.ControllerBGPadvertisementsReceived)
 		metrics.DefaultRegisterer.MustRegister(metrics.ControllerBGPadvertisementsSent)
 		metrics.DefaultRegisterer.MustRegister(metrics.ControllerBGPInternalPeersSyncTime)
-		metrics.DefaultRegisterer.MustRegister(metrics.ControllerBPGpeers)
+		metrics.DefaultRegisterer.MustRegister(metrics.ControllerBGPPeerInfo)
 		metrics.DefaultRegisterer.MustRegister(metrics.ControllerRoutesSyncTime)
 		nrc.MetricsEnabled = true
 	}

pkg/metrics/metrics_controller.go

@@ -158,12 +158,12 @@ var (
 		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{
+	// ControllerBGPPeerInfo BGP peer information
+	ControllerBGPPeerInfo = prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: namespace,
-		Name:      "controller_bgp_peers",
-		Help:      "BGP peers in the runtime configuration",
-	})
+		Name:      "bgp_peer_info",
+		Help:      "BGP peer information",
+	}, []string{"peer", "asn", "state"})
 	// ControllerBGPInternalPeersSyncTime Time it took to sync internal bgp peers
 	ControllerBGPInternalPeersSyncTime = prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: namespace,