[management] use peer.Status.LastSeen field to detect stale peer.Status.Connected values (#7268)

management/internals/shared/grpc/peer_update_handler.go

@@ -2,17 +2,20 @@ package grpc
 
 import (
 	"context"
+	"sync/atomic"
 	"time"
 
-	"github.com/netbirdio/netbird/encryption"
-	"github.com/netbirdio/netbird/management/internals/controllers/network_map"
-	"github.com/netbirdio/netbird/management/server/telemetry"
-	"github.com/netbirdio/netbird/shared/management/certposture"
-	"github.com/netbirdio/netbird/shared/management/proto"
 	log "github.com/sirupsen/logrus"
 	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/status"
+
+	"github.com/netbirdio/netbird/encryption"
+	"github.com/netbirdio/netbird/management/internals/controllers/network_map"
+	nbpeer "github.com/netbirdio/netbird/management/server/peer"
+	"github.com/netbirdio/netbird/management/server/telemetry"
+	"github.com/netbirdio/netbird/shared/management/certposture"
+	"github.com/netbirdio/netbird/shared/management/proto"
 )
 
 func PeerUpdateHandlerFactory(
@@ -21,6 +24,7 @@ func PeerUpdateHandlerFactory(
 	secretsManager SecretsManager,
 	challenger *certposture.Challenger,
 	onChallengeStamped func(),
+	heartbeat func(),
 	srv proto.ManagementService_SyncServer,
 	cleanupfunc func()) *PeerUpdateHandler {
 	return &PeerUpdateHandler{
@@ -29,6 +33,7 @@ func PeerUpdateHandlerFactory(
 		secretsManager:     secretsManager,
 		challenger:         challenger,
 		onChallengeStamped: onChallengeStamped,
+		heartbeat:          heartbeat,
 		srv:                srv,
 		encrypter:          encryption.DefaultEncrypter{},
 		debouncer:          NewUpdateDebouncer(1000 * time.Millisecond),
@@ -54,6 +59,7 @@ type PeerUpdateHandler struct {
 	encrypter          encryption.Encrypter
 	debouncer          Debouncer
 	cleanupFunc        func()
+	heartbeat          func()
 }
 
 func (pu *PeerUpdateHandler) WithMetrics(appMetrics telemetry.AppMetrics) *PeerUpdateHandler {
@@ -72,6 +78,14 @@ func (pu *PeerUpdateHandler) HandleUpdates(ctx context.Context) error {
 
 	defer pu.debouncer.Stop()
 
+	var heartbeat <-chan time.Time
+	var heartbeatRunning atomic.Bool
+	if pu.heartbeat != nil {
+		ticker := time.NewTicker(nbpeer.HeartbeatInterval)
+		defer ticker.Stop()
+		heartbeat = ticker.C
+	}
+
 	for {
 		select {
 		// condition when there are some updates
@@ -110,6 +124,15 @@ func (pu *PeerUpdateHandler) HandleUpdates(ctx context.Context) error {
 				}
 			}
 
+		case <-heartbeat:
+			// A heartbeat that finishes after cleanup writes nothing: it is fenced on this stream's session.
+			if heartbeatRunning.CompareAndSwap(false, true) {
+				go func() {
+					defer heartbeatRunning.Store(false)
+					pu.heartbeat()
+				}()
+			}
+
 		// condition when client <-> server connection has been terminated
 		case <-pu.srv.Context().Done():
 			// happens when connection drops, e.g. client disconnects