Merged affected-peers branch

management/internals/controllers/network_map/controller/controller.go

@@ -44,7 +44,7 @@ type Controller struct {
 	EphemeralPeersManager ephemeral.Manager
 
 	accountUpdateLocks               sync.Map
-	sendAccountUpdateLocks           sync.Map
+	affectedPeerUpdateLocks          sync.Map
 	updateAccountPeersBufferInterval atomic.Int64
 	// dnsDomain is used for peer resolution. This is appended to the peer's name
 	dnsDomain string
@@ -72,6 +72,13 @@ type bufferUpdate struct {
 	update atomic.Bool
 }
 
+type bufferAffectedUpdate struct {
+	sendMu  sync.Mutex
+	dataMu  sync.Mutex
+	next    *time.Timer
+	peerIDs map[string]struct{}
+}
+
 var _ network_map.Controller = (*Controller)(nil)
 
 func NewController(ctx context.Context, store store.Store, metrics telemetry.AppMetrics, peersUpdateManager network_map.PeersUpdateManager, requestBuffer account.RequestBuffer, integratedPeerValidator integrated_validator.IntegratedValidator, settingsManager settings.Manager, dnsDomain string, proxyController port_forwarding.Controller, ephemeralPeersManager ephemeral.Manager, config *config.Config) *Controller {
@@ -256,44 +263,6 @@ func (c *Controller) sendUpdateAccountPeers(ctx context.Context, accountID strin
 	return nil
 }
 
-func (c *Controller) bufferSendUpdateAccountPeers(ctx context.Context, accountID string, reason types.UpdateReason) error {
-	log.WithContext(ctx).Tracef("buffer sending update peers for account %s from %s", accountID, util.GetCallerName())
-
-	if c.accountManagerMetrics != nil {
-		c.accountManagerMetrics.CountUpdateAccountPeersTriggered(string(reason.Resource), string(reason.Operation))
-	}
-
-	bufUpd, _ := c.sendAccountUpdateLocks.LoadOrStore(accountID, &bufferUpdate{})
-	b := bufUpd.(*bufferUpdate)
-
-	if !b.mu.TryLock() {
-		b.update.Store(true)
-		return nil
-	}
-
-	if b.next != nil {
-		b.next.Stop()
-	}
-
-	go func() {
-		defer b.mu.Unlock()
-		_ = c.sendUpdateAccountPeers(ctx, accountID)
-		if !b.update.Load() {
-			return
-		}
-		b.update.Store(false)
-		if b.next == nil {
-			b.next = time.AfterFunc(time.Duration(c.updateAccountPeersBufferInterval.Load()), func() {
-				_ = c.sendUpdateAccountPeers(ctx, accountID)
-			})
-			return
-		}
-		b.next.Reset(time.Duration(c.updateAccountPeersBufferInterval.Load()))
-	}()
-
-	return nil
-}
-
 // UpdatePeers updates all peers that belong to an account.
 // Should be called when changes have to be synced to peers.
 func (c *Controller) UpdateAccountPeers(ctx context.Context, accountID string, reason types.UpdateReason) error {
@@ -303,6 +272,148 @@ func (c *Controller) UpdateAccountPeers(ctx context.Context, accountID string, r
 	return c.sendUpdateAccountPeers(ctx, accountID)
 }
 
+// UpdateAffectedPeers updates only the specified peers that belong to an account.
+func (c *Controller) UpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string) error {
+	if len(peerIDs) == 0 {
+		return nil
+	}
+	return c.sendUpdateForAffectedPeers(ctx, accountID, peerIDs)
+}
+
+func (c *Controller) sendUpdateForAffectedPeers(ctx context.Context, accountID string, peerIDs []string) error {
+	log.WithContext(ctx).Tracef("sendUpdateForAffectedPeers: account %s, %d affected peers: %v (caller: %s)", accountID, len(peerIDs), peerIDs, util.GetCallerName())
+
+	if !c.hasConnectedPeers(peerIDs) {
+		log.WithContext(ctx).Tracef("sendUpdateForAffectedPeers: no connected peers among %v, skipping", peerIDs)
+		return nil
+	}
+
+	account, err := c.requestBuffer.GetAccountWithBackpressure(ctx, accountID)
+	if err != nil {
+		return fmt.Errorf("failed to get account: %v", err)
+	}
+
+	globalStart := time.Now()
+
+	peersToUpdate := c.filterConnectedAffectedPeers(account, peerIDs)
+	if len(peersToUpdate) == 0 {
+		log.WithContext(ctx).Tracef("sendUpdateForAffectedPeers: no peers to update (affected peers not found in account or no channels)")
+		return nil
+	}
+
+	log.WithContext(ctx).Tracef("sendUpdateForAffectedPeers: sending network map to %d connected peers", len(peersToUpdate))
+
+	approvedPeersMap, err := c.integratedPeerValidator.GetValidatedPeers(ctx, account.Id, maps.Values(account.Groups), maps.Values(account.Peers), account.Settings.Extra)
+	if err != nil {
+		return fmt.Errorf("failed to get validate peers: %v", err)
+	}
+
+	var wg sync.WaitGroup
+	semaphore := make(chan struct{}, 10)
+
+	account.InjectProxyPolicies(ctx)
+	dnsCache := &cache.DNSConfigCache{}
+	dnsDomain := c.GetDNSDomain(account.Settings)
+	peersCustomZone := account.GetPeersCustomZone(ctx, dnsDomain)
+	resourcePolicies := account.GetResourcePoliciesMap()
+	routers := account.GetResourceRoutersMap()
+	groupIDToUserIDs := account.GetActiveGroupUsers()
+
+	proxyNetworkMaps, err := c.proxyController.GetProxyNetworkMapsAll(ctx, accountID, account.Peers)
+	if err != nil {
+		log.WithContext(ctx).Errorf("failed to get proxy network maps: %v", err)
+		return fmt.Errorf("failed to get proxy network maps: %v", err)
+	}
+
+	extraSetting, err := c.settingsManager.GetExtraSettings(ctx, accountID)
+	if err != nil {
+		return fmt.Errorf("failed to get flow enabled status: %v", err)
+	}
+
+	dnsFwdPort := computeForwarderPort(maps.Values(account.Peers), network_map.DnsForwarderPortMinVersion)
+
+	accountZones, err := c.repo.GetAccountZones(ctx, account.Id)
+	if err != nil {
+		log.WithContext(ctx).Errorf("failed to get account zones: %v", err)
+		return fmt.Errorf("failed to get account zones: %v", err)
+	}
+
+	for _, peer := range peersToUpdate {
+		wg.Add(1)
+		semaphore <- struct{}{}
+		go func(p *nbpeer.Peer) {
+			defer wg.Done()
+			defer func() { <-semaphore }()
+
+			start := time.Now()
+
+			postureChecks, err := c.getPeerPostureChecks(account, p.ID)
+			if err != nil {
+				log.WithContext(ctx).Debugf("failed to get posture checks for peer %s: %v", p.ID, err)
+				return
+			}
+
+			c.metrics.CountCalcPostureChecksDuration(time.Since(start))
+			start = time.Now()
+
+			result := account.GetPeerNetworkMapResult(ctx, p.ID, c.componentsDisabled, peersCustomZone, accountZones, approvedPeersMap, resourcePolicies, routers, c.accountManagerMetrics, groupIDToUserIDs)
+
+			c.metrics.CountCalcPeerNetworkMapDuration(time.Since(start))
+
+			proxyNetworkMap := proxyNetworkMaps[p.ID]
+			if result.NetworkMap != nil && proxyNetworkMap != nil {
+				result.NetworkMap.Merge(proxyNetworkMap)
+			}
+
+			peerGroups := account.GetPeerGroups(p.ID)
+			start = time.Now()
+			var update *proto.SyncResponse
+			if result.IsComponents() {
+				update = grpc.ToComponentSyncResponse(ctx, nil, c.config.HttpConfig, c.config.DeviceAuthorizationFlow, p, nil, nil, result.Components, proxyNetworkMap, dnsDomain, postureChecks, account.Settings, extraSetting, maps.Keys(peerGroups), dnsFwdPort)
+			} else {
+				update = grpc.ToSyncResponse(ctx, nil, c.config.HttpConfig, c.config.DeviceAuthorizationFlow, p, nil, nil, result.NetworkMap, dnsDomain, postureChecks, dnsCache, account.Settings, extraSetting, maps.Keys(peerGroups), dnsFwdPort)
+			}
+			c.metrics.CountToSyncResponseDuration(time.Since(start))
+
+			c.peersUpdateManager.SendUpdate(ctx, p.ID, &network_map.UpdateMessage{
+				Update:      update,
+				MessageType: network_map.MessageTypeNetworkMap,
+			})
+		}(peer)
+	}
+
+	wg.Wait()
+	if c.accountManagerMetrics != nil {
+		c.accountManagerMetrics.CountUpdateAccountPeersDuration(time.Since(globalStart))
+	}
+
+	return nil
+}
+
+func (c *Controller) hasConnectedPeers(peerIDs []string) bool {
+	for _, id := range peerIDs {
+		if c.peersUpdateManager.HasChannel(id) {
+			return true
+		}
+	}
+	return false
+}
+
+func (c *Controller) filterConnectedAffectedPeers(account *types.Account, peerIDs []string) []*nbpeer.Peer {
+	affected := make(map[string]struct{}, len(peerIDs))
+	for _, id := range peerIDs {
+		affected[id] = struct{}{}
+	}
+
+	var result []*nbpeer.Peer
+	for _, peer := range account.Peers {
+		if _, ok := affected[peer.ID]; ok && c.peersUpdateManager.HasChannel(peer.ID) {
+			result = append(result, peer)
+		}
+	}
+	return result
+}
+
 func (c *Controller) UpdateAccountPeer(ctx context.Context, accountId string, peerId string) error {
 	if !c.peersUpdateManager.HasChannel(peerId) {
 		return fmt.Errorf("peer %s doesn't have a channel, skipping network map update", peerId)
@@ -477,6 +588,100 @@ func (c *Controller) GetValidatedPeerWithComponents(ctx context.Context, isRequi
 	return peer, components, proxyNetworkMaps[peer.ID], postureChecks, dnsFwdPort, nil
 }
 
+// BufferUpdateAffectedPeers accumulates peer IDs and flushes them after the buffer interval.
+func (c *Controller) BufferUpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string, reason types.UpdateReason) error {
+	if len(peerIDs) == 0 {
+		return nil
+	}
+
+	if c.accountManagerMetrics != nil {
+		c.accountManagerMetrics.CountUpdateAccountPeersTriggered(string(reason.Resource), string(reason.Operation))
+	}
+
+	log.WithContext(ctx).Tracef("buffer updating %d affected peers for account %s from %s", len(peerIDs), accountID, util.GetCallerName())
+
+	bufUpd, _ := c.affectedPeerUpdateLocks.LoadOrStore(accountID, &bufferAffectedUpdate{
+		peerIDs: make(map[string]struct{}),
+	})
+	b := bufUpd.(*bufferAffectedUpdate)
+
+	b.addPeerIDs(peerIDs)
+
+	if !b.sendMu.TryLock() {
+		// Another goroutine is already sending; it will pick up our IDs on its next drain.
+		return nil
+	}
+
+	b.stopTimer()
+
+	collected := b.drainPeerIDs()
+	go func() {
+		defer b.sendMu.Unlock()
+		_ = c.sendUpdateForAffectedPeers(ctx, accountID, collected)
+
+		// Check if more peer IDs accumulated while we were sending.
+		if !b.hasPending() {
+			return
+		}
+
+		// Schedule a debounced flush for the newly accumulated IDs.
+		b.setTimer(time.Duration(c.updateAccountPeersBufferInterval.Load()), func() {
+			ids := b.drainPeerIDs()
+			if len(ids) > 0 {
+				_ = c.sendUpdateForAffectedPeers(ctx, accountID, ids)
+			}
+		})
+	}()
+
+	return nil
+}
+
+func (b *bufferAffectedUpdate) addPeerIDs(ids []string) {
+	b.dataMu.Lock()
+	for _, id := range ids {
+		b.peerIDs[id] = struct{}{}
+	}
+	b.dataMu.Unlock()
+}
+
+func (b *bufferAffectedUpdate) drainPeerIDs() []string {
+	b.dataMu.Lock()
+	defer b.dataMu.Unlock()
+	if len(b.peerIDs) == 0 {
+		return nil
+	}
+	ids := make([]string, 0, len(b.peerIDs))
+	for id := range b.peerIDs {
+		ids = append(ids, id)
+	}
+	b.peerIDs = make(map[string]struct{})
+	return ids
+}
+
+func (b *bufferAffectedUpdate) hasPending() bool {
+	b.dataMu.Lock()
+	defer b.dataMu.Unlock()
+	return len(b.peerIDs) > 0
+}
+
+func (b *bufferAffectedUpdate) stopTimer() {
+	b.dataMu.Lock()
+	defer b.dataMu.Unlock()
+	if b.next != nil {
+		b.next.Stop()
+	}
+}
+
+func (b *bufferAffectedUpdate) setTimer(d time.Duration, f func()) {
+	b.dataMu.Lock()
+	defer b.dataMu.Unlock()
+	if b.next == nil {
+		b.next = time.AfterFunc(d, f)
+		return
+	}
+	b.next.Reset(d)
+}
+
 func (c *Controller) GetValidatedPeerWithMap(ctx context.Context, isRequiresApproval bool, accountID string, peer *nbpeer.Peer) (*nbpeer.Peer, *types.NetworkMap, []*posture.Checks, int64, error) {
 	if isRequiresApproval {
 		network, err := c.repo.GetAccountNetwork(ctx, accountID)
@@ -674,21 +879,24 @@ func isPeerInPolicySourceGroups(account *types.Account, peerID string, policy *t
 	return false, nil
 }
 
-func (c *Controller) OnPeersUpdated(ctx context.Context, accountID string, peerIDs []string) error {
-	err := c.bufferSendUpdateAccountPeers(ctx, accountID, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationUpdate})
-	if err != nil {
-		log.WithContext(ctx).Errorf("failed to buffer update account peers for peer update in account %s: %v", accountID, err)
+func (c *Controller) OnPeersUpdated(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error {
+	if len(affectedPeerIDs) == 0 {
+		log.WithContext(ctx).Tracef("no affected peers for peer update in account %s, skipping", accountID)
+		return nil
 	}
-
-	return nil
+	return c.BufferUpdateAffectedPeers(ctx, accountID, affectedPeerIDs, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationUpdate})
 }
 
-func (c *Controller) OnPeersAdded(ctx context.Context, accountID string, peerIDs []string) error {
+func (c *Controller) OnPeersAdded(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error {
 	log.WithContext(ctx).Debugf("OnPeersAdded call to add peers: %v", peerIDs)
-	return c.bufferSendUpdateAccountPeers(ctx, accountID, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationCreate})
+	if len(affectedPeerIDs) == 0 {
+		log.WithContext(ctx).Tracef("no affected peers for peer add in account %s, skipping", accountID)
+		return nil
+	}
+	return c.BufferUpdateAffectedPeers(ctx, accountID, affectedPeerIDs, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationCreate})
 }
 
-func (c *Controller) OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string) error {
+func (c *Controller) OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error {
 	network, err := c.repo.GetAccountNetwork(ctx, accountID)
 	if err != nil {
 		return err
@@ -721,7 +929,11 @@ func (c *Controller) OnPeersDeleted(ctx context.Context, accountID string, peerI
 		c.peersUpdateManager.CloseChannel(ctx, peerID)
 	}
 
-	return c.bufferSendUpdateAccountPeers(ctx, accountID, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationDelete})
+	if len(affectedPeerIDs) == 0 {
+		log.WithContext(ctx).Tracef("no affected peers for peer delete in account %s, skipping network map update", accountID)
+		return nil
+	}
+	return c.BufferUpdateAffectedPeers(ctx, accountID, affectedPeerIDs, types.UpdateReason{Resource: types.UpdateResourcePeer, Operation: types.UpdateOperationDelete})
 }
 
 // GetNetworkMap returns Network map for a given peer (omits original peer from the Peers result)

management/internals/controllers/network_map/interface.go

@@ -19,6 +19,8 @@ const (
 
 type Controller interface {
 	UpdateAccountPeers(ctx context.Context, accountID string, reason types.UpdateReason) error
+	UpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string) error
+	BufferUpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string, reason types.UpdateReason) error
 	UpdateAccountPeer(ctx context.Context, accountId string, peerId string) error
 	BufferUpdateAccountPeers(ctx context.Context, accountID string, reason types.UpdateReason) error
 	GetValidatedPeerWithMap(ctx context.Context, isRequiresApproval bool, accountID string, p *nbpeer.Peer) (*nbpeer.Peer, *types.NetworkMap, []*posture.Checks, int64, error)
@@ -31,9 +33,9 @@ type Controller interface {
 	GetNetworkMap(ctx context.Context, peerID string) (*types.NetworkMap, error)
 	CountStreams() int
 
-	OnPeersUpdated(ctx context.Context, accountId string, peerIDs []string) error
-	OnPeersAdded(ctx context.Context, accountID string, peerIDs []string) error
-	OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string) error
+	OnPeersUpdated(ctx context.Context, accountId string, peerIDs []string, affectedPeerIDs []string) error
+	OnPeersAdded(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error
+	OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error
 	DisconnectPeers(ctx context.Context, accountId string, peerIDs []string)
 	OnPeerConnected(ctx context.Context, accountID string, peerID string) (chan *UpdateMessage, error)
 	OnPeerDisconnected(ctx context.Context, accountID string, peerID string)

management/internals/controllers/network_map/interface_mock.go

@@ -57,6 +57,20 @@ func (mr *MockControllerMockRecorder) BufferUpdateAccountPeers(ctx, accountID, r
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "BufferUpdateAccountPeers", reflect.TypeOf((*MockController)(nil).BufferUpdateAccountPeers), ctx, accountID, reason)
 }
 
+// BufferUpdateAffectedPeers mocks base method.
+func (m *MockController) BufferUpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string, reason types.UpdateReason) error {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "BufferUpdateAffectedPeers", ctx, accountID, peerIDs, reason)
+	ret0, _ := ret[0].(error)
+	return ret0
+}
+
+// BufferUpdateAffectedPeers indicates an expected call of BufferUpdateAffectedPeers.
+func (mr *MockControllerMockRecorder) BufferUpdateAffectedPeers(ctx, accountID, peerIDs, reason any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "BufferUpdateAffectedPeers", reflect.TypeOf((*MockController)(nil).BufferUpdateAffectedPeers), ctx, accountID, peerIDs, reason)
+}
+
 // CountStreams mocks base method.
 func (m *MockController) CountStreams() int {
 	m.ctrl.T.Helper()
@@ -191,45 +205,45 @@ func (mr *MockControllerMockRecorder) OnPeerDisconnected(ctx, accountID, peerID
 }
 
 // OnPeersAdded mocks base method.
-func (m *MockController) OnPeersAdded(ctx context.Context, accountID string, peerIDs []string) error {
+func (m *MockController) OnPeersAdded(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error {
 	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "OnPeersAdded", ctx, accountID, peerIDs)
+	ret := m.ctrl.Call(m, "OnPeersAdded", ctx, accountID, peerIDs, affectedPeerIDs)
 	ret0, _ := ret[0].(error)
 	return ret0
 }
 
 // OnPeersAdded indicates an expected call of OnPeersAdded.
-func (mr *MockControllerMockRecorder) OnPeersAdded(ctx, accountID, peerIDs any) *gomock.Call {
+func (mr *MockControllerMockRecorder) OnPeersAdded(ctx, accountID, peerIDs, affectedPeerIDs any) *gomock.Call {
 	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersAdded", reflect.TypeOf((*MockController)(nil).OnPeersAdded), ctx, accountID, peerIDs)
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersAdded", reflect.TypeOf((*MockController)(nil).OnPeersAdded), ctx, accountID, peerIDs, affectedPeerIDs)
 }
 
 // OnPeersDeleted mocks base method.
-func (m *MockController) OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string) error {
+func (m *MockController) OnPeersDeleted(ctx context.Context, accountID string, peerIDs []string, affectedPeerIDs []string) error {
 	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "OnPeersDeleted", ctx, accountID, peerIDs)
+	ret := m.ctrl.Call(m, "OnPeersDeleted", ctx, accountID, peerIDs, affectedPeerIDs)
 	ret0, _ := ret[0].(error)
 	return ret0
 }
 
 // OnPeersDeleted indicates an expected call of OnPeersDeleted.
-func (mr *MockControllerMockRecorder) OnPeersDeleted(ctx, accountID, peerIDs any) *gomock.Call {
+func (mr *MockControllerMockRecorder) OnPeersDeleted(ctx, accountID, peerIDs, affectedPeerIDs any) *gomock.Call {
 	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersDeleted", reflect.TypeOf((*MockController)(nil).OnPeersDeleted), ctx, accountID, peerIDs)
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersDeleted", reflect.TypeOf((*MockController)(nil).OnPeersDeleted), ctx, accountID, peerIDs, affectedPeerIDs)
 }
 
 // OnPeersUpdated mocks base method.
-func (m *MockController) OnPeersUpdated(ctx context.Context, accountId string, peerIDs []string) error {
+func (m *MockController) OnPeersUpdated(ctx context.Context, accountId string, peerIDs []string, affectedPeerIDs []string) error {
 	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "OnPeersUpdated", ctx, accountId, peerIDs)
+	ret := m.ctrl.Call(m, "OnPeersUpdated", ctx, accountId, peerIDs, affectedPeerIDs)
 	ret0, _ := ret[0].(error)
 	return ret0
 }
 
 // OnPeersUpdated indicates an expected call of OnPeersUpdated.
-func (mr *MockControllerMockRecorder) OnPeersUpdated(ctx, accountId, peerIDs any) *gomock.Call {
+func (mr *MockControllerMockRecorder) OnPeersUpdated(ctx, accountId, peerIDs, affectedPeerIDs any) *gomock.Call {
 	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersUpdated", reflect.TypeOf((*MockController)(nil).OnPeersUpdated), ctx, accountId, peerIDs)
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "OnPeersUpdated", reflect.TypeOf((*MockController)(nil).OnPeersUpdated), ctx, accountId, peerIDs, affectedPeerIDs)
 }
 
 // StartWarmup mocks base method.
@@ -283,3 +297,17 @@ func (mr *MockControllerMockRecorder) UpdateAccountPeers(ctx, accountID, reason
 	mr.mock.ctrl.T.Helper()
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "UpdateAccountPeers", reflect.TypeOf((*MockController)(nil).UpdateAccountPeers), ctx, accountID, reason)
 }
+
+// UpdateAffectedPeers mocks base method.
+func (m *MockController) UpdateAffectedPeers(ctx context.Context, accountID string, peerIDs []string) error {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "UpdateAffectedPeers", ctx, accountID, peerIDs)
+	ret0, _ := ret[0].(error)
+	return ret0
+}
+
+// UpdateAffectedPeers indicates an expected call of UpdateAffectedPeers.
+func (mr *MockControllerMockRecorder) UpdateAffectedPeers(ctx, accountID, peerIDs any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "UpdateAffectedPeers", reflect.TypeOf((*MockController)(nil).UpdateAffectedPeers), ctx, accountID, peerIDs)
+}

management/internals/modules/reverseproxy/proxy/manager/manager.go

@@ -17,7 +17,7 @@ type store interface {
 	UpdateProxyHeartbeat(ctx context.Context, p *proxy.Proxy) error
 	GetActiveProxyClusterAddresses(ctx context.Context) ([]string, error)
 	GetActiveProxyClusterAddressesForAccount(ctx context.Context, accountID string) ([]string, error)
-	GetActiveProxyClusters(ctx context.Context, accountID string) ([]proxy.Cluster, error)
+	GetProxyClusters(ctx context.Context, accountID string) ([]proxy.Cluster, error)
 	GetClusterSupportsCustomPorts(ctx context.Context, clusterAddr string) *bool
 	GetClusterRequireSubdomain(ctx context.Context, clusterAddr string) *bool
 	GetClusterSupportsCrowdSec(ctx context.Context, clusterAddr string) *bool

management/internals/modules/reverseproxy/proxy/manager/manager_test.go

@@ -57,7 +57,7 @@ func (m *mockStore) GetActiveProxyClusterAddressesForAccount(ctx context.Context
 	}
 	return nil, nil
 }
-func (m *mockStore) GetActiveProxyClusters(_ context.Context, _ string) ([]proxy.Cluster, error) {
+func (m *mockStore) GetProxyClusters(_ context.Context, _ string) ([]proxy.Cluster, error) {
 	return nil, nil
 }
 func (m *mockStore) CleanupStaleProxies(ctx context.Context, d time.Duration) error {

management/internals/modules/reverseproxy/proxy/proxy.go

@@ -42,10 +42,35 @@ func (Proxy) TableName() string {
 	return "proxies"
 }
 
+// ClusterType is the source of a proxy cluster.
+type ClusterType string
+
+const (
+	// ClusterTypeAccount is a cluster operated by the account itself (BYOP) —
+	// at least one proxy row in the cluster carries a non-NULL account_id.
+	ClusterTypeAccount ClusterType = "account"
+	// ClusterTypeShared is a cluster operated by NetBird and shared across
+	// accounts — all proxy rows in the cluster have account_id IS NULL.
+	ClusterTypeShared ClusterType = "shared"
+)
+
 // Cluster represents a group of proxy nodes serving the same address.
+//
+// Online and ConnectedProxies derive from the same 2-min active window
+// the rest of the module uses, but Cluster rows are not gated on it —
+// the cluster listing surfaces offline clusters too so operators can
+// see and clean them up. The 1-hour heartbeat reaper still bounds the
+// table eventually.
 type Cluster struct {
 	ID               string
 	Address          string
+	Type             ClusterType
+	Online           bool
 	ConnectedProxies int
-	SelfHosted       bool
+	// Capability flags. *bool because nil means "no proxy reported a
+	// capability for this cluster" — the dashboard renders these as
+	// unknown rather than false.
+	SupportsCustomPorts *bool
+	RequireSubdomain    *bool
+	SupportsCrowdSec    *bool
 }

management/internals/modules/reverseproxy/service/interface.go

@@ -9,7 +9,7 @@ import (
 )
 
 type Manager interface {
-	GetActiveClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error)
+	GetClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error)
 	DeleteAccountCluster(ctx context.Context, accountID, userID, clusterAddress string) error
 	GetAllServices(ctx context.Context, accountID, userID string) ([]*Service, error)
 	GetService(ctx context.Context, accountID, userID, serviceID string) (*Service, error)

management/internals/modules/reverseproxy/service/interface_mock.go

@@ -65,20 +65,6 @@ func (mr *MockManagerMockRecorder) CreateServiceFromPeer(ctx, accountID, peerID,
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "CreateServiceFromPeer", reflect.TypeOf((*MockManager)(nil).CreateServiceFromPeer), ctx, accountID, peerID, req)
 }
 
-// DeleteAllServices mocks base method.
-func (m *MockManager) DeleteAllServices(ctx context.Context, accountID, userID string) error {
-	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "DeleteAllServices", ctx, accountID, userID)
-	ret0, _ := ret[0].(error)
-	return ret0
-}
-
-// DeleteAllServices indicates an expected call of DeleteAllServices.
-func (mr *MockManagerMockRecorder) DeleteAllServices(ctx, accountID, userID interface{}) *gomock.Call {
-	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "DeleteAllServices", reflect.TypeOf((*MockManager)(nil).DeleteAllServices), ctx, accountID, userID)
-}
-
 // DeleteAccountCluster mocks base method.
 func (m *MockManager) DeleteAccountCluster(ctx context.Context, accountID, userID, clusterAddress string) error {
 	m.ctrl.T.Helper()
@@ -93,6 +79,20 @@ func (mr *MockManagerMockRecorder) DeleteAccountCluster(ctx, accountID, userID,
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "DeleteAccountCluster", reflect.TypeOf((*MockManager)(nil).DeleteAccountCluster), ctx, accountID, userID, clusterAddress)
 }
 
+// DeleteAllServices mocks base method.
+func (m *MockManager) DeleteAllServices(ctx context.Context, accountID, userID string) error {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "DeleteAllServices", ctx, accountID, userID)
+	ret0, _ := ret[0].(error)
+	return ret0
+}
+
+// DeleteAllServices indicates an expected call of DeleteAllServices.
+func (mr *MockManagerMockRecorder) DeleteAllServices(ctx, accountID, userID interface{}) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "DeleteAllServices", reflect.TypeOf((*MockManager)(nil).DeleteAllServices), ctx, accountID, userID)
+}
+
 // DeleteService mocks base method.
 func (m *MockManager) DeleteService(ctx context.Context, accountID, userID, serviceID string) error {
 	m.ctrl.T.Helper()
@@ -122,21 +122,6 @@ func (mr *MockManagerMockRecorder) GetAccountServices(ctx, accountID interface{}
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetAccountServices", reflect.TypeOf((*MockManager)(nil).GetAccountServices), ctx, accountID)
 }
 
-// GetActiveClusters mocks base method.
-func (m *MockManager) GetActiveClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error) {
-	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "GetActiveClusters", ctx, accountID, userID)
-	ret0, _ := ret[0].([]proxy.Cluster)
-	ret1, _ := ret[1].(error)
-	return ret0, ret1
-}
-
-// GetActiveClusters indicates an expected call of GetActiveClusters.
-func (mr *MockManagerMockRecorder) GetActiveClusters(ctx, accountID, userID interface{}) *gomock.Call {
-	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetActiveClusters", reflect.TypeOf((*MockManager)(nil).GetActiveClusters), ctx, accountID, userID)
-}
-
 // GetAllServices mocks base method.
 func (m *MockManager) GetAllServices(ctx context.Context, accountID, userID string) ([]*Service, error) {
 	m.ctrl.T.Helper()
@@ -152,19 +137,19 @@ func (mr *MockManagerMockRecorder) GetAllServices(ctx, accountID, userID interfa
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetAllServices", reflect.TypeOf((*MockManager)(nil).GetAllServices), ctx, accountID, userID)
 }
 
-// GetServiceByDomain mocks base method.
-func (m *MockManager) GetServiceByDomain(ctx context.Context, domain string) (*Service, error) {
+// GetClusters mocks base method.
+func (m *MockManager) GetClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error) {
 	m.ctrl.T.Helper()
-	ret := m.ctrl.Call(m, "GetServiceByDomain", ctx, domain)
-	ret0, _ := ret[0].(*Service)
+	ret := m.ctrl.Call(m, "GetClusters", ctx, accountID, userID)
+	ret0, _ := ret[0].([]proxy.Cluster)
 	ret1, _ := ret[1].(error)
 	return ret0, ret1
 }
 
-// GetServiceByDomain indicates an expected call of GetServiceByDomain.
-func (mr *MockManagerMockRecorder) GetServiceByDomain(ctx, domain interface{}) *gomock.Call {
+// GetClusters indicates an expected call of GetClusters.
+func (mr *MockManagerMockRecorder) GetClusters(ctx, accountID, userID interface{}) *gomock.Call {
 	mr.mock.ctrl.T.Helper()
-	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetServiceByDomain", reflect.TypeOf((*MockManager)(nil).GetServiceByDomain), ctx, domain)
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetClusters", reflect.TypeOf((*MockManager)(nil).GetClusters), ctx, accountID, userID)
 }
 
 // GetGlobalServices mocks base method.
@@ -197,6 +182,21 @@ func (mr *MockManagerMockRecorder) GetService(ctx, accountID, userID, serviceID
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetService", reflect.TypeOf((*MockManager)(nil).GetService), ctx, accountID, userID, serviceID)
 }
 
+// GetServiceByDomain mocks base method.
+func (m *MockManager) GetServiceByDomain(ctx context.Context, domain string) (*Service, error) {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "GetServiceByDomain", ctx, domain)
+	ret0, _ := ret[0].(*Service)
+	ret1, _ := ret[1].(error)
+	return ret0, ret1
+}
+
+// GetServiceByDomain indicates an expected call of GetServiceByDomain.
+func (mr *MockManagerMockRecorder) GetServiceByDomain(ctx, domain interface{}) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetServiceByDomain", reflect.TypeOf((*MockManager)(nil).GetServiceByDomain), ctx, domain)
+}
+
 // GetServiceByID mocks base method.
 func (m *MockManager) GetServiceByID(ctx context.Context, accountID, serviceID string) (*Service, error) {
 	m.ctrl.T.Helper()

management/internals/modules/reverseproxy/service/manager/api.go

@@ -187,7 +187,7 @@ func (h *handler) getClusters(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	clusters, err := h.manager.GetActiveClusters(r.Context(), userAuth.AccountId, userAuth.UserId)
+	clusters, err := h.manager.GetClusters(r.Context(), userAuth.AccountId, userAuth.UserId)
 	if err != nil {
 		util.WriteError(r.Context(), err, w)
 		return
@@ -196,10 +196,14 @@ func (h *handler) getClusters(w http.ResponseWriter, r *http.Request) {
 	apiClusters := make([]api.ProxyCluster, 0, len(clusters))
 	for _, c := range clusters {
 		apiClusters = append(apiClusters, api.ProxyCluster{
-			Id:               c.ID,
-			Address:          c.Address,
-			ConnectedProxies: c.ConnectedProxies,
-			SelfHosted:       c.SelfHosted,
+			Id:                  c.ID,
+			Address:             c.Address,
+			Type:                api.ProxyClusterType(c.Type),
+			Online:              c.Online,
+			ConnectedProxies:    c.ConnectedProxies,
+			SupportsCustomPorts: c.SupportsCustomPorts,
+			RequireSubdomain:    c.RequireSubdomain,
+			SupportsCrowdsec:    c.SupportsCrowdSec,
 		})
 	}
 

management/internals/modules/reverseproxy/service/manager/manager.go

@@ -81,6 +81,7 @@ type ClusterDeriver interface {
 type CapabilityProvider interface {
 	ClusterSupportsCustomPorts(ctx context.Context, clusterAddr string) *bool
 	ClusterRequireSubdomain(ctx context.Context, clusterAddr string) *bool
+	ClusterSupportsCrowdSec(ctx context.Context, clusterAddr string) *bool
 }
 
 type Manager struct {
@@ -112,8 +113,12 @@ func (m *Manager) StartExposeReaper(ctx context.Context) {
 	m.exposeReaper.StartExposeReaper(ctx)
 }
 
-// GetActiveClusters returns all active proxy clusters with their connected proxy count.
-func (m *Manager) GetActiveClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error) {
+// GetClusters returns every proxy cluster visible to the account
+// (shared + its own BYOP), regardless of whether any proxy in the
+// cluster is currently heartbeating. Each cluster is enriched with the
+// capability flags reported by its active proxies so the dashboard can
+// render feature support without a second round-trip.
+func (m *Manager) GetClusters(ctx context.Context, accountID, userID string) ([]proxy.Cluster, error) {
 	ok, err := m.permissionsManager.ValidateUserPermissions(ctx, accountID, userID, modules.Services, operations.Read)
 	if err != nil {
 		return nil, status.NewPermissionValidationError(err)
@@ -122,7 +127,18 @@ func (m *Manager) GetActiveClusters(ctx context.Context, accountID, userID strin
 		return nil, status.NewPermissionDeniedError()
 	}
 
-	return m.store.GetActiveProxyClusters(ctx, accountID)
+	clusters, err := m.store.GetProxyClusters(ctx, accountID)
+	if err != nil {
+		return nil, err
+	}
+
+	for i := range clusters {
+		clusters[i].SupportsCustomPorts = m.capabilities.ClusterSupportsCustomPorts(ctx, clusters[i].Address)
+		clusters[i].RequireSubdomain = m.capabilities.ClusterRequireSubdomain(ctx, clusters[i].Address)
+		clusters[i].SupportsCrowdSec = m.capabilities.ClusterSupportsCrowdSec(ctx, clusters[i].Address)
+	}
+
+	return clusters, nil
 }
 
 // DeleteAccountCluster removes all proxy registrations for the given cluster address

management/internals/shared/grpc/proxy_group_access_test.go

@@ -109,7 +109,7 @@ func (m *mockReverseProxyManager) GetServiceByDomain(_ context.Context, domain s
 	return nil, errors.New("service not found for domain: " + domain)
 }
 
-func (m *mockReverseProxyManager) GetActiveClusters(_ context.Context, _, _ string) ([]proxy.Cluster, error) {
+func (m *mockReverseProxyManager) GetClusters(_ context.Context, _, _ string) ([]proxy.Cluster, error) {
 	return nil, nil
 }
 

management/internals/shared/grpc/validate_session_test.go

@@ -322,7 +322,7 @@ func (m *testValidateSessionServiceManager) GetServiceByDomain(ctx context.Conte
 	return m.store.GetServiceByDomain(ctx, domain)
 }
 
-func (m *testValidateSessionServiceManager) GetActiveClusters(_ context.Context, _, _ string) ([]proxy.Cluster, error) {
+func (m *testValidateSessionServiceManager) GetClusters(_ context.Context, _, _ string) ([]proxy.Cluster, error) {
 	return nil, nil
 }