Merge branch 'main' into ssh-rewrite

client/internal/peer/conn.go

@@ -24,9 +24,8 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer/id"
 	"github.com/netbirdio/netbird/client/internal/peer/worker"
 	"github.com/netbirdio/netbird/client/internal/stdnet"
-	relayClient "github.com/netbirdio/netbird/relay/client"
 	"github.com/netbirdio/netbird/route"
-	nbnet "github.com/netbirdio/netbird/util/net"
+	relayClient "github.com/netbirdio/netbird/shared/relay/client"
 	semaphoregroup "github.com/netbirdio/netbird/util/semaphore-group"
 )
 
@@ -106,10 +105,6 @@ type Conn struct {
 	workerRelay *WorkerRelay
 	wgWatcherWg sync.WaitGroup
 
-	connIDRelay          nbnet.ConnectionID
-	connIDICE            nbnet.ConnectionID
-	beforeAddPeerHooks   []nbnet.AddHookFunc
-	afterRemovePeerHooks []nbnet.RemoveHookFunc
 	// used to store the remote Rosenpass key for Relayed connection in case of connection update from ice
 	rosenpassRemoteKey []byte
 
@@ -167,7 +162,7 @@ func (conn *Conn) Open(engineCtx context.Context) error {
 
 	conn.ctx, conn.ctxCancel = context.WithCancel(engineCtx)
 
-	conn.workerRelay = NewWorkerRelay(conn.Log, isController(conn.config), conn.config, conn, conn.relayManager, conn.dumpState)
+	conn.workerRelay = NewWorkerRelay(conn.ctx, conn.Log, isController(conn.config), conn.config, conn, conn.relayManager, conn.dumpState)
 
 	relayIsSupportedLocally := conn.workerRelay.RelayIsSupportedLocally()
 	workerICE, err := NewWorkerICE(conn.ctx, conn.Log, conn.config, conn, conn.signaler, conn.iFaceDiscover, conn.statusRecorder, relayIsSupportedLocally)
@@ -205,19 +200,11 @@ func (conn *Conn) Open(engineCtx context.Context) error {
 	conn.wg.Add(1)
 	go func() {
 		defer conn.wg.Done()
+
 		conn.waitInitialRandomSleepTime(conn.ctx)
 		conn.semaphore.Done(conn.ctx)
 
-		conn.dumpState.SendOffer()
-		if err := conn.handshaker.sendOffer(); err != nil {
-			conn.Log.Errorf("failed to send initial offer: %v", err)
-		}
-
-		conn.wg.Add(1)
-		go func() {
-			conn.guard.Start(conn.ctx, conn.onGuardEvent)
-			conn.wg.Done()
-		}()
+		conn.guard.Start(conn.ctx, conn.onGuardEvent)
 	}()
 	conn.opened = true
 	return nil
@@ -267,8 +254,6 @@ func (conn *Conn) Close(signalToRemote bool) {
 		conn.Log.Errorf("failed to remove wg endpoint: %v", err)
 	}
 
-	conn.freeUpConnID()
-
 	if conn.evalStatus() == StatusConnected && conn.onDisconnected != nil {
 		conn.onDisconnected(conn.config.WgConfig.RemoteKey)
 	}
@@ -281,10 +266,10 @@ func (conn *Conn) Close(signalToRemote bool) {
 
 // OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
 // doesn't block, discards the message if connection wasn't ready
-func (conn *Conn) OnRemoteAnswer(answer OfferAnswer) bool {
+func (conn *Conn) OnRemoteAnswer(answer OfferAnswer) {
 	conn.dumpState.RemoteAnswer()
 	conn.Log.Infof("OnRemoteAnswer, priority: %s, status ICE: %s, status relay: %s", conn.currentConnPriority, conn.statusICE, conn.statusRelay)
-	return conn.handshaker.OnRemoteAnswer(answer)
+	conn.handshaker.OnRemoteAnswer(answer)
 }
 
 // OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
@@ -293,13 +278,6 @@ func (conn *Conn) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMa
 	conn.workerICE.OnRemoteCandidate(candidate, haRoutes)
 }
 
-func (conn *Conn) AddBeforeAddPeerHook(hook nbnet.AddHookFunc) {
-	conn.beforeAddPeerHooks = append(conn.beforeAddPeerHooks, hook)
-}
-func (conn *Conn) AddAfterRemovePeerHook(hook nbnet.RemoveHookFunc) {
-	conn.afterRemovePeerHooks = append(conn.afterRemovePeerHooks, hook)
-}
-
 // SetOnConnected sets a handler function to be triggered by Conn when a new connection to a remote peer established
 func (conn *Conn) SetOnConnected(handler func(remoteWireGuardKey string, remoteRosenpassPubKey []byte, wireGuardIP string, remoteRosenpassAddr string)) {
 	conn.onConnected = handler
@@ -310,10 +288,10 @@ func (conn *Conn) SetOnDisconnected(handler func(remotePeer string)) {
 	conn.onDisconnected = handler
 }
 
-func (conn *Conn) OnRemoteOffer(offer OfferAnswer) bool {
+func (conn *Conn) OnRemoteOffer(offer OfferAnswer) {
 	conn.dumpState.RemoteOffer()
 	conn.Log.Infof("OnRemoteOffer, on status ICE: %s, status Relay: %s", conn.statusICE, conn.statusRelay)
-	return conn.handshaker.OnRemoteOffer(offer)
+	conn.handshaker.OnRemoteOffer(offer)
 }
 
 // WgConfig returns the WireGuard config
@@ -387,10 +365,6 @@ func (conn *Conn) onICEConnectionIsReady(priority conntype.ConnPriority, iceConn
 		ep = directEp
 	}
 
-	if err := conn.runBeforeAddPeerHooks(ep.IP); err != nil {
-		conn.Log.Errorf("Before add peer hook failed: %v", err)
-	}
-
 	conn.workerRelay.DisableWgWatcher()
 	// todo consider to run conn.wgWatcherWg.Wait() here
 
@@ -489,6 +463,8 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
 		conn.Log.Errorf("failed to add relayed net.Conn to local proxy: %v", err)
 		return
 	}
+	wgProxy.SetDisconnectListener(conn.onRelayDisconnected)
+
 	conn.dumpState.NewLocalProxy()
 
 	conn.Log.Infof("created new wgProxy for relay connection: %s", wgProxy.EndpointAddr().String())
@@ -501,10 +477,6 @@ func (conn *Conn) onRelayConnectionIsReady(rci RelayConnInfo) {
 		return
 	}
 
-	if err := conn.runBeforeAddPeerHooks(wgProxy.EndpointAddr().IP); err != nil {
-		conn.Log.Errorf("Before add peer hook failed: %v", err)
-	}
-
 	wgProxy.Work()
 	if err := conn.configureWGEndpoint(wgProxy.EndpointAddr(), rci.rosenpassPubKey); err != nil {
 		if err := wgProxy.CloseConn(); err != nil {
@@ -568,7 +540,6 @@ func (conn *Conn) onRelayDisconnected() {
 }
 
 func (conn *Conn) onGuardEvent() {
-	conn.Log.Debugf("send offer to peer")
 	conn.dumpState.SendOffer()
 	if err := conn.handshaker.SendOffer(); err != nil {
 		conn.Log.Errorf("failed to send offer: %v", err)
@@ -692,7 +663,7 @@ func (conn *Conn) isConnectedOnAllWay() (connected bool) {
 		}
 	}()
 
-	if conn.statusICE.Get() == worker.StatusDisconnected {
+	if conn.statusICE.Get() == worker.StatusDisconnected && !conn.workerICE.InProgress() {
 		return false
 	}
 
@@ -705,36 +676,6 @@ func (conn *Conn) isConnectedOnAllWay() (connected bool) {
 	return true
 }
 
-func (conn *Conn) runBeforeAddPeerHooks(ip net.IP) error {
-	conn.connIDICE = nbnet.GenerateConnID()
-	for _, hook := range conn.beforeAddPeerHooks {
-		if err := hook(conn.connIDICE, ip); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func (conn *Conn) freeUpConnID() {
-	if conn.connIDRelay != "" {
-		for _, hook := range conn.afterRemovePeerHooks {
-			if err := hook(conn.connIDRelay); err != nil {
-				conn.Log.Errorf("After remove peer hook failed: %v", err)
-			}
-		}
-		conn.connIDRelay = ""
-	}
-
-	if conn.connIDICE != "" {
-		for _, hook := range conn.afterRemovePeerHooks {
-			if err := hook(conn.connIDICE); err != nil {
-				conn.Log.Errorf("After remove peer hook failed: %v", err)
-			}
-		}
-		conn.connIDICE = ""
-	}
-}
-
 func (conn *Conn) newProxy(remoteConn net.Conn) (wgproxy.Proxy, error) {
 	conn.Log.Debugf("setup proxied WireGuard connection")
 	udpAddr := &net.UDPAddr{

client/internal/peer/conn_test.go

@@ -1,9 +1,9 @@
 package peer
 
 import (
+	"context"
 	"fmt"
 	"os"
-	"sync"
 	"testing"
 	"time"
 
@@ -31,7 +31,7 @@ var connConf = ConnConfig{
 }
 
 func TestMain(m *testing.M) {
-	_ = util.InitLog("trace", "console")
+	_ = util.InitLog("trace", util.LogConsole)
 	code := m.Run()
 	os.Exit(code)
 }
@@ -79,31 +79,30 @@ func TestConn_OnRemoteOffer(t *testing.T) {
 		return
 	}
 
-	wg := sync.WaitGroup{}
-	wg.Add(2)
-	go func() {
-		<-conn.handshaker.remoteOffersCh
-		wg.Done()
-	}()
+	onNewOffeChan := make(chan struct{})
 
-	go func() {
-		for {
-			accepted := conn.OnRemoteOffer(OfferAnswer{
-				IceCredentials: IceCredentials{
-					UFrag: "test",
-					Pwd:   "test",
-				},
-				WgListenPort: 0,
-				Version:      "",
-			})
-			if accepted {
-				wg.Done()
-				return
-			}
-		}
-	}()
+	conn.handshaker.AddOnNewOfferListener(func(remoteOfferAnswer *OfferAnswer) {
+		onNewOffeChan <- struct{}{}
+	})
 
-	wg.Wait()
+	conn.OnRemoteOffer(OfferAnswer{
+		IceCredentials: IceCredentials{
+			UFrag: "test",
+			Pwd:   "test",
+		},
+		WgListenPort: 0,
+		Version:      "",
+	})
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	select {
+	case <-onNewOffeChan:
+		// success
+	case <-ctx.Done():
+		t.Error("expected to receive a new offer notification, but timed out")
+	}
 }
 
 func TestConn_OnRemoteAnswer(t *testing.T) {
@@ -119,31 +118,29 @@ func TestConn_OnRemoteAnswer(t *testing.T) {
 		return
 	}
 
-	wg := sync.WaitGroup{}
-	wg.Add(2)
-	go func() {
-		<-conn.handshaker.remoteAnswerCh
-		wg.Done()
-	}()
+	onNewOffeChan := make(chan struct{})
 
-	go func() {
-		for {
-			accepted := conn.OnRemoteAnswer(OfferAnswer{
-				IceCredentials: IceCredentials{
-					UFrag: "test",
-					Pwd:   "test",
-				},
-				WgListenPort: 0,
-				Version:      "",
-			})
-			if accepted {
-				wg.Done()
-				return
-			}
-		}
-	}()
+	conn.handshaker.AddOnNewOfferListener(func(remoteOfferAnswer *OfferAnswer) {
+		onNewOffeChan <- struct{}{}
+	})
 
-	wg.Wait()
+	conn.OnRemoteAnswer(OfferAnswer{
+		IceCredentials: IceCredentials{
+			UFrag: "test",
+			Pwd:   "test",
+		},
+		WgListenPort: 0,
+		Version:      "",
+	})
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	select {
+	case <-onNewOffeChan:
+		// success
+	case <-ctx.Done():
+		t.Error("expected to receive a new offer notification, but timed out")
+	}
 }
 
 func TestConn_presharedKey(t *testing.T) {

client/internal/peer/guard/guard.go

@@ -19,7 +19,6 @@ type isConnectedFunc func() bool
 // - Relayed connection disconnected
 // - ICE candidate changes
 type Guard struct {
-	Reconnect               chan struct{}
 	log                     *log.Entry
 	isConnectedOnAllWay     isConnectedFunc
 	timeout                 time.Duration
@@ -30,7 +29,6 @@ type Guard struct {
 
 func NewGuard(log *log.Entry, isConnectedFn isConnectedFunc, timeout time.Duration, srWatcher *SRWatcher) *Guard {
 	return &Guard{
-		Reconnect:               make(chan struct{}, 1),
 		log:                     log,
 		isConnectedOnAllWay:     isConnectedFn,
 		timeout:                 timeout,
@@ -41,6 +39,7 @@ func NewGuard(log *log.Entry, isConnectedFn isConnectedFunc, timeout time.Durati
 }
 
 func (g *Guard) Start(ctx context.Context, eventCallback func()) {
+	g.log.Infof("starting guard for reconnection with MaxInterval: %s", g.timeout)
 	g.reconnectLoopWithRetry(ctx, eventCallback)
 }
 
@@ -61,17 +60,14 @@ func (g *Guard) SetICEConnDisconnected() {
 // reconnectLoopWithRetry periodically check the connection status.
 // Try to send offer while the P2P is not established or while the Relay is not connected if is it supported
 func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
-	waitForInitialConnectionTry(ctx)
-
 	srReconnectedChan := g.srWatcher.NewListener()
 	defer g.srWatcher.RemoveListener(srReconnectedChan)
 
-	ticker := g.prepareExponentTicker(ctx)
+	ticker := g.initialTicker(ctx)
 	defer ticker.Stop()
 
 	tickerChannel := ticker.C
 
-	g.log.Infof("start reconnect loop...")
 	for {
 		select {
 		case t := <-tickerChannel:
@@ -85,7 +81,6 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
 			if !g.isConnectedOnAllWay() {
 				callback()
 			}
-
 		case <-g.relayedConnDisconnected:
 			g.log.Debugf("Relay connection changed, reset reconnection ticker")
 			ticker.Stop()
@@ -111,6 +106,20 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
 	}
 }
 
+// initialTicker give chance to the peer to establish the initial connection.
+func (g *Guard) initialTicker(ctx context.Context) *backoff.Ticker {
+	bo := backoff.WithContext(&backoff.ExponentialBackOff{
+		InitialInterval:     3 * time.Second,
+		RandomizationFactor: 0.1,
+		Multiplier:          2,
+		MaxInterval:         g.timeout,
+		Stop:                backoff.Stop,
+		Clock:               backoff.SystemClock,
+	}, ctx)
+
+	return backoff.NewTicker(bo)
+}
+
 func (g *Guard) prepareExponentTicker(ctx context.Context) *backoff.Ticker {
 	bo := backoff.WithContext(&backoff.ExponentialBackOff{
 		InitialInterval:     800 * time.Millisecond,
@@ -126,13 +135,3 @@ func (g *Guard) prepareExponentTicker(ctx context.Context) *backoff.Ticker {
 
 	return ticker
 }
-
-// Give chance to the peer to establish the initial connection.
-// With it, we can decrease to send necessary offer
-func waitForInitialConnectionTry(ctx context.Context) {
-	select {
-	case <-ctx.Done():
-		return
-	case <-time.After(3 * time.Second):
-	}
-}

client/internal/peer/handshaker.go

@@ -39,6 +39,15 @@ type OfferAnswer struct {
 
 	// relay server address
 	RelaySrvAddress string
+	// SessionID is the unique identifier of the session, used to discard old messages
+	SessionID *ICESessionID
+}
+
+func (oa *OfferAnswer) SessionIDString() string {
+	if oa.SessionID == nil {
+		return "unknown"
+	}
+	return oa.SessionID.String()
 }
 
 type Handshaker struct {
@@ -74,21 +83,25 @@ func (h *Handshaker) AddOnNewOfferListener(offer func(remoteOfferAnswer *OfferAn
 
 func (h *Handshaker) Listen(ctx context.Context) {
 	for {
-		h.log.Info("wait for remote offer confirmation")
-		remoteOfferAnswer, err := h.waitForRemoteOfferConfirmation(ctx)
-		if err != nil {
-			var connectionClosedError *ConnectionClosedError
-			if errors.As(err, &connectionClosedError) {
-				h.log.Info("exit from handshaker")
-				return
+		select {
+		case remoteOfferAnswer := <-h.remoteOffersCh:
+			// received confirmation from the remote peer -> ready to proceed
+			if err := h.sendAnswer(); err != nil {
+				h.log.Errorf("failed to send remote offer confirmation: %s", err)
+				continue
 			}
-			h.log.Errorf("failed to received remote offer confirmation: %s", err)
-			continue
-		}
-
-		h.log.Infof("received connection confirmation, running version %s and with remote WireGuard listen port %d", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort)
-		for _, listener := range h.onNewOfferListeners {
-			go listener(remoteOfferAnswer)
+			for _, listener := range h.onNewOfferListeners {
+				listener(&remoteOfferAnswer)
+			}
+			h.log.Infof("received offer, running version %s, remote WireGuard listen port %d, session id: %s", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString())
+		case remoteOfferAnswer := <-h.remoteAnswerCh:
+			h.log.Infof("received answer, running version %s, remote WireGuard listen port %d, session id: %s", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString())
+			for _, listener := range h.onNewOfferListeners {
+				listener(&remoteOfferAnswer)
+			}
+		case <-ctx.Done():
+			h.log.Infof("stop listening for remote offers and answers")
+			return
 		}
 	}
 }
@@ -101,43 +114,27 @@ func (h *Handshaker) SendOffer() error {
 
 // OnRemoteOffer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
 // doesn't block, discards the message if connection wasn't ready
-func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) bool {
+func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) {
 	select {
 	case h.remoteOffersCh <- offer:
-		return true
+		return
 	default:
-		h.log.Warnf("OnRemoteOffer skipping message because is not ready")
+		h.log.Warnf("skipping remote offer message because receiver not ready")
 		// connection might not be ready yet to receive so we ignore the message
-		return false
+		return
 	}
 }
 
 // OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
 // doesn't block, discards the message if connection wasn't ready
-func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) bool {
+func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) {
 	select {
 	case h.remoteAnswerCh <- answer:
-		return true
+		return
 	default:
 		// connection might not be ready yet to receive so we ignore the message
-		h.log.Debugf("OnRemoteAnswer skipping message because is not ready")
-		return false
-	}
-}
-
-func (h *Handshaker) waitForRemoteOfferConfirmation(ctx context.Context) (*OfferAnswer, error) {
-	select {
-	case remoteOfferAnswer := <-h.remoteOffersCh:
-		// received confirmation from the remote peer -> ready to proceed
-		if err := h.sendAnswer(); err != nil {
-			return nil, err
-		}
-		return &remoteOfferAnswer, nil
-	case remoteOfferAnswer := <-h.remoteAnswerCh:
-		return &remoteOfferAnswer, nil
-	case <-ctx.Done():
-		// closed externally
-		return nil, NewConnectionClosedError(h.config.Key)
+		h.log.Warnf("skipping remote answer message because receiver not ready")
+		return
 	}
 }
 
@@ -147,43 +144,34 @@ func (h *Handshaker) sendOffer() error {
 		return ErrSignalIsNotReady
 	}
 
-	iceUFrag, icePwd := h.ice.GetLocalUserCredentials()
-	offer := OfferAnswer{
-		IceCredentials:  IceCredentials{iceUFrag, icePwd},
-		WgListenPort:    h.config.LocalWgPort,
-		Version:         version.NetbirdVersion(),
-		RosenpassPubKey: h.config.RosenpassConfig.PubKey,
-		RosenpassAddr:   h.config.RosenpassConfig.Addr,
-	}
-
-	addr, err := h.relay.RelayInstanceAddress()
-	if err == nil {
-		offer.RelaySrvAddress = addr
-	}
+	offer := h.buildOfferAnswer()
+	h.log.Infof("sending offer with serial: %s", offer.SessionIDString())
 
 	return h.signaler.SignalOffer(offer, h.config.Key)
 }
 
 func (h *Handshaker) sendAnswer() error {
-	h.log.Infof("sending answer")
-	uFrag, pwd := h.ice.GetLocalUserCredentials()
+	answer := h.buildOfferAnswer()
+	h.log.Infof("sending answer with serial: %s", answer.SessionIDString())
 
+	return h.signaler.SignalAnswer(answer, h.config.Key)
+}
+
+func (h *Handshaker) buildOfferAnswer() OfferAnswer {
+	uFrag, pwd := h.ice.GetLocalUserCredentials()
+	sid := h.ice.SessionID()
 	answer := OfferAnswer{
 		IceCredentials:  IceCredentials{uFrag, pwd},
 		WgListenPort:    h.config.LocalWgPort,
 		Version:         version.NetbirdVersion(),
 		RosenpassPubKey: h.config.RosenpassConfig.PubKey,
 		RosenpassAddr:   h.config.RosenpassConfig.Addr,
+		SessionID:       &sid,
 	}
-	addr, err := h.relay.RelayInstanceAddress()
-	if err == nil {
+
+	if addr, err := h.relay.RelayInstanceAddress(); err == nil {
 		answer.RelaySrvAddress = addr
 	}
 
-	err = h.signaler.SignalAnswer(answer, h.config.Key)
-	if err != nil {
-		return err
-	}
-
-	return nil
+	return answer
 }

client/internal/peer/ice/agent.go

@@ -1,6 +1,7 @@
 package ice
 
 import (
+	"sync"
 	"time"
 
 	"github.com/pion/ice/v3"
@@ -18,17 +19,28 @@ const (
 
 	iceKeepAliveDefault           = 4 * time.Second
 	iceDisconnectedTimeoutDefault = 6 * time.Second
+	iceFailedTimeoutDefault       = 6 * time.Second
 	// iceRelayAcceptanceMinWaitDefault is the same as in the Pion ICE package
 	iceRelayAcceptanceMinWaitDefault = 2 * time.Second
 )
 
-var (
-	failedTimeout = 6 * time.Second
-)
+type ThreadSafeAgent struct {
+	*ice.Agent
+	once sync.Once
+}
 
-func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candidateTypes []ice.CandidateType, ufrag string, pwd string) (*ice.Agent, error) {
+func (a *ThreadSafeAgent) Close() error {
+	var err error
+	a.once.Do(func() {
+		err = a.Agent.Close()
+	})
+	return err
+}
+
+func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candidateTypes []ice.CandidateType, ufrag string, pwd string) (*ThreadSafeAgent, error) {
 	iceKeepAlive := iceKeepAlive()
 	iceDisconnectedTimeout := iceDisconnectedTimeout()
+	iceFailedTimeout := iceFailedTimeout()
 	iceRelayAcceptanceMinWait := iceRelayAcceptanceMinWait()
 
 	transportNet, err := newStdNet(iFaceDiscover, config.InterfaceBlackList)
@@ -50,7 +62,7 @@ func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candida
 		UDPMuxSrflx:            config.UDPMuxSrflx,
 		NAT1To1IPs:             config.NATExternalIPs,
 		Net:                    transportNet,
-		FailedTimeout:          &failedTimeout,
+		FailedTimeout:          &iceFailedTimeout,
 		DisconnectedTimeout:    &iceDisconnectedTimeout,
 		KeepaliveInterval:      &iceKeepAlive,
 		RelayAcceptanceMinWait: &iceRelayAcceptanceMinWait,
@@ -63,7 +75,12 @@ func NewAgent(iFaceDiscover stdnet.ExternalIFaceDiscover, config Config, candida
 		agentConfig.NetworkTypes = []ice.NetworkType{ice.NetworkTypeUDP4}
 	}
 
-	return ice.NewAgent(agentConfig)
+	agent, err := ice.NewAgent(agentConfig)
+	if err != nil {
+		return nil, err
+	}
+
+	return &ThreadSafeAgent{Agent: agent}, nil
 }
 
 func GenerateICECredentials() (string, string, error) {

client/internal/peer/ice/env.go

@@ -13,6 +13,7 @@ const (
 	envICEForceRelayConn            = "NB_ICE_FORCE_RELAY_CONN"
 	envICEKeepAliveIntervalSec      = "NB_ICE_KEEP_ALIVE_INTERVAL_SEC"
 	envICEDisconnectedTimeoutSec    = "NB_ICE_DISCONNECTED_TIMEOUT_SEC"
+	envICEFailedTimeoutSec          = "NB_ICE_FAILED_TIMEOUT_SEC"
 	envICERelayAcceptanceMinWaitSec = "NB_ICE_RELAY_ACCEPTANCE_MIN_WAIT_SEC"
 
 	msgWarnInvalidValue = "invalid value %s set for %s, using default %v"
@@ -55,6 +56,22 @@ func iceDisconnectedTimeout() time.Duration {
 	return time.Duration(disconnectedTimeoutSec) * time.Second
 }
 
+func iceFailedTimeout() time.Duration {
+	failedTimeoutEnv := os.Getenv(envICEFailedTimeoutSec)
+	if failedTimeoutEnv == "" {
+		return iceFailedTimeoutDefault
+	}
+
+	log.Infof("setting ICE failed timeout to %s seconds", failedTimeoutEnv)
+	failedTimeoutSec, err := strconv.Atoi(failedTimeoutEnv)
+	if err != nil {
+		log.Warnf(msgWarnInvalidValue, failedTimeoutEnv, envICEFailedTimeoutSec, iceFailedTimeoutDefault)
+		return iceFailedTimeoutDefault
+	}
+
+	return time.Duration(failedTimeoutSec) * time.Second
+}
+
 func iceRelayAcceptanceMinWait() time.Duration {
 	iceRelayAcceptanceMinWaitEnv := os.Getenv(envICERelayAcceptanceMinWaitSec)
 	if iceRelayAcceptanceMinWaitEnv == "" {

client/internal/peer/session_id.go

@@ -0,0 +1,47 @@
+package peer
+
+import (
+	"crypto/rand"
+	"encoding/hex"
+	"fmt"
+	"io"
+)
+
+const sessionIDSize = 5
+
+type ICESessionID string
+
+// NewICESessionID generates a new session ID for distinguishing sessions
+func NewICESessionID() (ICESessionID, error) {
+	b := make([]byte, sessionIDSize)
+	if _, err := io.ReadFull(rand.Reader, b); err != nil {
+		return "", fmt.Errorf("failed to generate session ID: %w", err)
+	}
+	return ICESessionID(hex.EncodeToString(b)), nil
+}
+
+func ICESessionIDFromBytes(b []byte) (ICESessionID, error) {
+	if len(b) != sessionIDSize {
+		return "", fmt.Errorf("invalid session ID length: %d", len(b))
+	}
+	return ICESessionID(hex.EncodeToString(b)), nil
+}
+
+// Bytes returns the raw bytes of the session ID for protobuf serialization
+func (id ICESessionID) Bytes() ([]byte, error) {
+	if len(id) == 0 {
+		return nil, fmt.Errorf("ICE session ID is empty")
+	}
+	b, err := hex.DecodeString(string(id))
+	if err != nil {
+		return nil, fmt.Errorf("invalid ICE session ID encoding: %w", err)
+	}
+	if len(b) != sessionIDSize {
+		return nil, fmt.Errorf("invalid ICE session ID length: expected %d bytes, got %d", sessionIDSize, len(b))
+	}
+	return b, nil
+}
+
+func (id ICESessionID) String() string {
+	return string(id)
+}

client/internal/peer/signaler.go

@@ -2,10 +2,11 @@ package peer
 
 import (
 	"github.com/pion/ice/v3"
+	log "github.com/sirupsen/logrus"
 	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
 
-	signal "github.com/netbirdio/netbird/signal/client"
-	sProto "github.com/netbirdio/netbird/signal/proto"
+	signal "github.com/netbirdio/netbird/shared/signal/client"
+	sProto "github.com/netbirdio/netbird/shared/signal/proto"
 )
 
 type Signaler struct {
@@ -45,6 +46,10 @@ func (s *Signaler) Ready() bool {
 
 // SignalOfferAnswer signals either an offer or an answer to remote peer
 func (s *Signaler) signalOfferAnswer(offerAnswer OfferAnswer, remoteKey string, bodyType sProto.Body_Type) error {
+	sessionIDBytes, err := offerAnswer.SessionID.Bytes()
+	if err != nil {
+		log.Warnf("failed to get session ID bytes: %v", err)
+	}
 	msg, err := signal.MarshalCredential(
 		s.wgPrivateKey,
 		offerAnswer.WgListenPort,
@@ -56,13 +61,13 @@ func (s *Signaler) signalOfferAnswer(offerAnswer OfferAnswer, remoteKey string,
 		bodyType,
 		offerAnswer.RosenpassPubKey,
 		offerAnswer.RosenpassAddr,
-		offerAnswer.RelaySrvAddress)
+		offerAnswer.RelaySrvAddress,
+		sessionIDBytes)
 	if err != nil {
 		return err
 	}
 
-	err = s.signal.Send(msg)
-	if err != nil {
+	if err = s.signal.Send(msg); err != nil {
 		return err
 	}
 

client/internal/peer/status.go

@@ -21,8 +21,8 @@ import (
 	"github.com/netbirdio/netbird/client/internal/ingressgw"
 	"github.com/netbirdio/netbird/client/internal/relay"
 	"github.com/netbirdio/netbird/client/proto"
-	"github.com/netbirdio/netbird/management/domain"
-	relayClient "github.com/netbirdio/netbird/relay/client"
+	"github.com/netbirdio/netbird/shared/management/domain"
+	relayClient "github.com/netbirdio/netbird/shared/relay/client"
 	"github.com/netbirdio/netbird/route"
 )
 
@@ -141,7 +141,7 @@ type RosenpassState struct {
 // whether it's enabled, and the last error message encountered during probing.
 type NSGroupState struct {
 	ID      string
-	Servers []string
+	Servers []netip.AddrPort
 	Domains []string
 	Enabled bool
 	Error   error

client/internal/peer/worker_ice.go

@@ -42,8 +42,18 @@ type WorkerICE struct {
 	statusRecorder    *Status
 	hasRelayOnLocally bool
 
-	agent    *ice.Agent
-	muxAgent sync.Mutex
+	agent             *icemaker.ThreadSafeAgent
+	agentDialerCancel context.CancelFunc
+	agentConnecting   bool      // while it is true, drop all incoming offers
+	lastSuccess       time.Time // with this avoid the too frequent ICE agent recreation
+	// remoteSessionID represents the peer's session identifier from the latest remote offer.
+	remoteSessionID ICESessionID
+	// sessionID is used to track the current session ID of the ICE agent
+	// increase by one when disconnecting the agent
+	// with it the remote peer can discard the already deprecated offer/answer
+	// Without it the remote peer may recreate a workable ICE connection
+	sessionID ICESessionID
+	muxAgent  sync.Mutex
 
 	StunTurn []*stun.URI
 
@@ -57,6 +67,11 @@ type WorkerICE struct {
 }
 
 func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *Conn, signaler *Signaler, ifaceDiscover stdnet.ExternalIFaceDiscover, statusRecorder *Status, hasRelayOnLocally bool) (*WorkerICE, error) {
+	sessionID, err := NewICESessionID()
+	if err != nil {
+		return nil, err
+	}
+
 	w := &WorkerICE{
 		ctx:               ctx,
 		log:               log,
@@ -67,6 +82,7 @@ func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *
 		statusRecorder:    statusRecorder,
 		hasRelayOnLocally: hasRelayOnLocally,
 		lastKnownState:    ice.ConnectionStateDisconnected,
+		sessionID:         sessionID,
 	}
 
 	localUfrag, localPwd, err := icemaker.GenerateICECredentials()
@@ -79,15 +95,36 @@ func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *
 }
 
 func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
-	w.log.Debugf("OnNewOffer for ICE")
+	w.log.Debugf("OnNewOffer for ICE, serial: %s", remoteOfferAnswer.SessionIDString())
 	w.muxAgent.Lock()
 
-	if w.agent != nil {
-		w.log.Debugf("agent already exists, skipping the offer")
+	if w.agentConnecting {
+		w.log.Debugf("agent connection is in progress, skipping the offer")
 		w.muxAgent.Unlock()
 		return
 	}
 
+	if w.agent != nil {
+		// backward compatibility with old clients that do not send session ID
+		if remoteOfferAnswer.SessionID == nil {
+			w.log.Debugf("agent already exists, skipping the offer")
+			w.muxAgent.Unlock()
+			return
+		}
+		if w.remoteSessionID == *remoteOfferAnswer.SessionID {
+			w.log.Debugf("agent already exists and session ID matches, skipping the offer: %s", remoteOfferAnswer.SessionIDString())
+			w.muxAgent.Unlock()
+			return
+		}
+		w.log.Debugf("agent already exists, recreate the connection")
+		w.agentDialerCancel()
+		if err := w.agent.Close(); err != nil {
+			w.log.Warnf("failed to close ICE agent: %s", err)
+		}
+		w.agent = nil
+		// todo consider to switch to Relay connection while establishing a new ICE connection
+	}
+
 	var preferredCandidateTypes []ice.CandidateType
 	if w.hasRelayOnLocally && remoteOfferAnswer.RelaySrvAddress != "" {
 		preferredCandidateTypes = icemaker.CandidateTypesP2P()
@@ -96,36 +133,124 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 	}
 
 	w.log.Debugf("recreate ICE agent")
-	agentCtx, agentCancel := context.WithCancel(w.ctx)
-	agent, err := w.reCreateAgent(agentCancel, preferredCandidateTypes)
+	dialerCtx, dialerCancel := context.WithCancel(w.ctx)
+	agent, err := w.reCreateAgent(dialerCancel, preferredCandidateTypes)
 	if err != nil {
 		w.log.Errorf("failed to recreate ICE Agent: %s", err)
 		w.muxAgent.Unlock()
 		return
 	}
+	w.sentExtraSrflx = false
 	w.agent = agent
+	w.agentDialerCancel = dialerCancel
+	w.agentConnecting = true
 	w.muxAgent.Unlock()
 
+	go w.connect(dialerCtx, agent, remoteOfferAnswer)
+}
+
+// OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
+func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
+	w.muxAgent.Lock()
+	defer w.muxAgent.Unlock()
+	w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.config.Key, candidate.String())
+	if w.agent == nil {
+		w.log.Warnf("ICE Agent is not initialized yet")
+		return
+	}
+
+	if candidateViaRoutes(candidate, haRoutes) {
+		return
+	}
+
+	if err := w.agent.AddRemoteCandidate(candidate); err != nil {
+		w.log.Errorf("error while handling remote candidate")
+		return
+	}
+}
+
+func (w *WorkerICE) GetLocalUserCredentials() (frag string, pwd string) {
+	return w.localUfrag, w.localPwd
+}
+
+func (w *WorkerICE) InProgress() bool {
+	w.muxAgent.Lock()
+	defer w.muxAgent.Unlock()
+
+	return w.agentConnecting
+}
+
+func (w *WorkerICE) Close() {
+	w.muxAgent.Lock()
+	defer w.muxAgent.Unlock()
+
+	if w.agent == nil {
+		return
+	}
+
+	w.agentDialerCancel()
+	if err := w.agent.Close(); err != nil {
+		w.log.Warnf("failed to close ICE agent: %s", err)
+	}
+
+	w.agent = nil
+}
+
+func (w *WorkerICE) reCreateAgent(dialerCancel context.CancelFunc, candidates []ice.CandidateType) (*icemaker.ThreadSafeAgent, error) {
+	agent, err := icemaker.NewAgent(w.iFaceDiscover, w.config.ICEConfig, candidates, w.localUfrag, w.localPwd)
+	if err != nil {
+		return nil, fmt.Errorf("create agent: %w", err)
+	}
+
+	if err := agent.OnCandidate(w.onICECandidate); err != nil {
+		return nil, err
+	}
+
+	if err := agent.OnConnectionStateChange(w.onConnectionStateChange(agent, dialerCancel)); err != nil {
+		return nil, err
+	}
+
+	if err := agent.OnSelectedCandidatePairChange(w.onICESelectedCandidatePair); err != nil {
+		return nil, err
+	}
+
+	if err := agent.OnSuccessfulSelectedPairBindingResponse(w.onSuccessfulSelectedPairBindingResponse); err != nil {
+		return nil, fmt.Errorf("failed setting binding response callback: %w", err)
+	}
+
+	return agent, nil
+}
+
+func (w *WorkerICE) SessionID() ICESessionID {
+	w.muxAgent.Lock()
+	defer w.muxAgent.Unlock()
+
+	return w.sessionID
+}
+
+// will block until connection succeeded
+// but it won't release if ICE Agent went into Disconnected or Failed state,
+// so we have to cancel it with the provided context once agent detected a broken connection
+func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) {
 	w.log.Debugf("gather candidates")
-	err = w.agent.GatherCandidates()
-	if err != nil {
-		w.log.Debugf("failed to gather candidates: %s", err)
+	if err := agent.GatherCandidates(); err != nil {
+		w.log.Warnf("failed to gather candidates: %s", err)
+		w.closeAgent(agent, w.agentDialerCancel)
 		return
 	}
 
-	// will block until connection succeeded
-	// but it won't release if ICE Agent went into Disconnected or Failed state,
-	// so we have to cancel it with the provided context once agent detected a broken connection
 	w.log.Debugf("turn agent dial")
-	remoteConn, err := w.turnAgentDial(agentCtx, remoteOfferAnswer)
+	remoteConn, err := w.turnAgentDial(ctx, agent, remoteOfferAnswer)
 	if err != nil {
 		w.log.Debugf("failed to dial the remote peer: %s", err)
+		w.closeAgent(agent, w.agentDialerCancel)
 		return
 	}
 	w.log.Debugf("agent dial succeeded")
 
-	pair, err := w.agent.GetSelectedCandidatePair()
+	pair, err := agent.GetSelectedCandidatePair()
 	if err != nil {
+		w.closeAgent(agent, w.agentDialerCancel)
 		return
 	}
 
@@ -152,114 +277,39 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 		RelayedOnLocal:             isRelayCandidate(pair.Local),
 	}
 	w.log.Debugf("on ICE conn is ready to use")
-	go w.conn.onICEConnectionIsReady(selectedPriority(pair), ci)
-}
 
-// OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
-func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
+	w.log.Infof("connection succeeded with offer session: %s", remoteOfferAnswer.SessionIDString())
 	w.muxAgent.Lock()
-	defer w.muxAgent.Unlock()
-	w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.config.Key, candidate.String())
-	if w.agent == nil {
-		w.log.Warnf("ICE Agent is not initialized yet")
-		return
+	w.agentConnecting = false
+	w.lastSuccess = time.Now()
+	if remoteOfferAnswer.SessionID != nil {
+		w.remoteSessionID = *remoteOfferAnswer.SessionID
 	}
+	w.muxAgent.Unlock()
 
-	if candidateViaRoutes(candidate, haRoutes) {
-		return
-	}
-
-	err := w.agent.AddRemoteCandidate(candidate)
-	if err != nil {
-		w.log.Errorf("error while handling remote candidate")
-		return
-	}
+	// todo: the potential problem is a race between the onConnectionStateChange
+	w.conn.onICEConnectionIsReady(selectedPriority(pair), ci)
 }
 
-func (w *WorkerICE) GetLocalUserCredentials() (frag string, pwd string) {
-	w.muxAgent.Lock()
-	defer w.muxAgent.Unlock()
-	return w.localUfrag, w.localPwd
-}
-
-func (w *WorkerICE) Close() {
-	w.muxAgent.Lock()
-	defer w.muxAgent.Unlock()
-
-	if w.agent == nil {
-		return
-	}
-
-	if err := w.agent.Close(); err != nil {
-		w.log.Warnf("failed to close ICE agent: %s", err)
-	}
-}
-
-func (w *WorkerICE) reCreateAgent(agentCancel context.CancelFunc, candidates []ice.CandidateType) (*ice.Agent, error) {
-	w.sentExtraSrflx = false
-
-	agent, err := icemaker.NewAgent(w.iFaceDiscover, w.config.ICEConfig, candidates, w.localUfrag, w.localPwd)
-	if err != nil {
-		return nil, fmt.Errorf("create agent: %w", err)
-	}
-
-	err = agent.OnCandidate(w.onICECandidate)
-	if err != nil {
-		return nil, err
-	}
-
-	err = agent.OnConnectionStateChange(func(state ice.ConnectionState) {
-		w.log.Debugf("ICE ConnectionState has changed to %s", state.String())
-		switch state {
-		case ice.ConnectionStateConnected:
-			w.lastKnownState = ice.ConnectionStateConnected
-			return
-		case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected:
-			if w.lastKnownState == ice.ConnectionStateConnected {
-				w.lastKnownState = ice.ConnectionStateDisconnected
-				w.conn.onICEStateDisconnected()
-			}
-			w.closeAgent(agentCancel)
-		default:
-			return
-		}
-	})
-	if err != nil {
-		return nil, err
-	}
-
-	err = agent.OnSelectedCandidatePairChange(w.onICESelectedCandidatePair)
-	if err != nil {
-		return nil, err
-	}
-
-	err = agent.OnSuccessfulSelectedPairBindingResponse(func(p *ice.CandidatePair) {
-		err := w.statusRecorder.UpdateLatency(w.config.Key, p.Latency())
-		if err != nil {
-			w.log.Debugf("failed to update latency for peer: %s", err)
-			return
-		}
-	})
-	if err != nil {
-		return nil, fmt.Errorf("failed setting binding response callback: %w", err)
-	}
-
-	return agent, nil
-}
-
-func (w *WorkerICE) closeAgent(cancel context.CancelFunc) {
-	w.muxAgent.Lock()
-	defer w.muxAgent.Unlock()
-
+func (w *WorkerICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.CancelFunc) {
 	cancel()
-	if w.agent == nil {
-		return
-	}
-
-	if err := w.agent.Close(); err != nil {
+	if err := agent.Close(); err != nil {
 		w.log.Warnf("failed to close ICE agent: %s", err)
 	}
-	w.agent = nil
+
+	w.muxAgent.Lock()
+	// todo review does it make sense to generate new session ID all the time when w.agent==agent
+	sessionID, err := NewICESessionID()
+	if err != nil {
+		w.log.Errorf("failed to create new session ID: %s", err)
+	}
+	w.sessionID = sessionID
+
+	if w.agent == agent {
+		w.agent = nil
+		w.agentConnecting = false
+	}
+	w.muxAgent.Unlock()
 }
 
 func (w *WorkerICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int) {
@@ -331,6 +381,32 @@ func (w *WorkerICE) onICESelectedCandidatePair(c1 ice.Candidate, c2 ice.Candidat
 		w.config.Key)
 }
 
+func (w *WorkerICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dialerCancel context.CancelFunc) func(ice.ConnectionState) {
+	return func(state ice.ConnectionState) {
+		w.log.Debugf("ICE ConnectionState has changed to %s", state.String())
+		switch state {
+		case ice.ConnectionStateConnected:
+			w.lastKnownState = ice.ConnectionStateConnected
+			return
+		case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected:
+			if w.lastKnownState == ice.ConnectionStateConnected {
+				w.lastKnownState = ice.ConnectionStateDisconnected
+				w.conn.onICEStateDisconnected()
+			}
+			w.closeAgent(agent, dialerCancel)
+		default:
+			return
+		}
+	}
+}
+
+func (w *WorkerICE) onSuccessfulSelectedPairBindingResponse(pair *ice.CandidatePair) {
+	if err := w.statusRecorder.UpdateLatency(w.config.Key, pair.Latency()); err != nil {
+		w.log.Debugf("failed to update latency for peer: %s", err)
+		return
+	}
+}
+
 func (w *WorkerICE) shouldSendExtraSrflxCandidate(candidate ice.Candidate) bool {
 	if !w.sentExtraSrflx && candidate.Type() == ice.CandidateTypeServerReflexive && candidate.Port() != candidate.RelatedAddress().Port {
 		return true
@@ -338,12 +414,12 @@ func (w *WorkerICE) shouldSendExtraSrflxCandidate(candidate ice.Candidate) bool
 	return false
 }
 
-func (w *WorkerICE) turnAgentDial(ctx context.Context, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
+func (w *WorkerICE) turnAgentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
 	isControlling := w.config.LocalKey > w.config.Key
 	if isControlling {
-		return w.agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
+		return agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
 	} else {
-		return w.agent.Accept(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
+		return agent.Accept(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
 	}
 }
 

client/internal/peer/worker_relay.go

@@ -9,7 +9,7 @@ import (
 
 	log "github.com/sirupsen/logrus"
 
-	relayClient "github.com/netbirdio/netbird/relay/client"
+	relayClient "github.com/netbirdio/netbird/shared/relay/client"
 )
 
 type RelayConnInfo struct {
@@ -19,11 +19,12 @@ type RelayConnInfo struct {
 }
 
 type WorkerRelay struct {
+	peerCtx      context.Context
 	log          *log.Entry
 	isController bool
 	config       ConnConfig
 	conn         *Conn
-	relayManager relayClient.ManagerService
+	relayManager *relayClient.Manager
 
 	relayedConn net.Conn
 	relayLock   sync.Mutex
@@ -33,8 +34,9 @@ type WorkerRelay struct {
 	wgWatcher *WGWatcher
 }
 
-func NewWorkerRelay(log *log.Entry, ctrl bool, config ConnConfig, conn *Conn, relayManager relayClient.ManagerService, stateDump *stateDump) *WorkerRelay {
+func NewWorkerRelay(ctx context.Context, log *log.Entry, ctrl bool, config ConnConfig, conn *Conn, relayManager *relayClient.Manager, stateDump *stateDump) *WorkerRelay {
 	r := &WorkerRelay{
+		peerCtx:      ctx,
 		log:          log,
 		isController: ctrl,
 		config:       config,
@@ -62,7 +64,7 @@ func (w *WorkerRelay) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 
 	srv := w.preferredRelayServer(currentRelayAddress, remoteOfferAnswer.RelaySrvAddress)
 
-	relayedConn, err := w.relayManager.OpenConn(srv, w.config.Key)
+	relayedConn, err := w.relayManager.OpenConn(w.peerCtx, srv, w.config.Key)
 	if err != nil {
 		if errors.Is(err, relayClient.ErrConnAlreadyExists) {
 			w.log.Debugf("handled offer by reusing existing relay connection")