extern/netbird
1
0
Fork 0
mirror of https://github.com/netbirdio/netbird.git synced 2025-06-25 12:12:12 +02:00
Code Issues Packages Projects Releases Wiki Activity

[client] Handle lazy routing peers that are part of HA groups (#3943)

Browse Source 
* Activate new lazy routing peers if the HA group is active
* Prevent lazy peers going to idle if HA group members are active (#3948)
This commit is contained in:
Viktor Liu 2025-06-20 18:07:19 +02:00 committed by GitHub
parent 83457f8b99
commit 2a51609436
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 137 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
client/internal/conn_mgr.go
View File

@ -175,7 +175,7 @@ func (e *ConnMgr) AddPeerConn(ctx context.Context, peerKey string, conn *peer.Co
PeerConnID: conn.ConnID(),
Log: conn.Log,
}
excluded, err := e.lazyConnMgr.AddPeer(lazyPeerCfg)
excluded, err := e.lazyConnMgr.AddPeer(e.lazyCtx, lazyPeerCfg)
if err != nil {
conn.Log.Errorf("failed to add peer to lazyconn manager: %v", err)
if err := conn.Open(ctx); err != nil {

5
client/internal/lazyconn/inactivity/inactivity.go
View File

@ -68,3 +68,8 @@ func (i *Monitor) PauseTimer() {
func (i *Monitor) ResetTimer() {
i.timer.Reset(i.inactivityThreshold)
}
func (i *Monitor) ResetMonitor(ctx context.Context, timeoutChan chan peer.ConnID) {
i.Stop()
go i.Start(ctx, timeoutChan)
}

137
client/internal/lazyconn/manager/manager.go
View File

@ -58,7 +58,7 @@ type Manager struct {
// Route HA group management
peerToHAGroups map[string][]route.HAUniqueID // peer ID -> HA groups they belong to
haGroupToPeers map[route.HAUniqueID][]string // HA group -> peer IDs in the group
routesMu sync.RWMutex // protects route mappings
routesMu sync.RWMutex
onInactive chan peerid.ConnID
}
@ -146,7 +146,7 @@ func (m *Manager) Start(ctx context.Context) {
case peerConnID := <-m.activityManager.OnActivityChan:
m.onPeerActivity(ctx, peerConnID)
case peerConnID := <-m.onInactive:
m.onPeerInactivityTimedOut(peerConnID)
m.onPeerInactivityTimedOut(ctx, peerConnID)
}
}
}
@ -197,7 +197,7 @@ func (m *Manager) ExcludePeer(ctx context.Context, peerConfigs []lazyconn.PeerCo
return added
}
func (m *Manager) AddPeer(peerCfg lazyconn.PeerConfig) (bool, error) {
func (m *Manager) AddPeer(ctx context.Context, peerCfg lazyconn.PeerConfig) (bool, error) {
m.managedPeersMu.Lock()
defer m.managedPeersMu.Unlock()
@ -225,6 +225,13 @@ func (m *Manager) AddPeer(peerCfg lazyconn.PeerConfig) (bool, error) {
peerCfg: &peerCfg,
expectedWatcher: watcherActivity,
}
// Check if this peer should be activated because its HA group peers are active
if group, ok := m.shouldActivateNewPeer(peerCfg.PublicKey); ok {
peerCfg.Log.Debugf("peer belongs to active HA group %s, will activate immediately", group)
m.activateNewPeerInActiveGroup(ctx, peerCfg)
}
return false, nil
}
@ -315,26 +322,29 @@ func (m *Manager) activateSinglePeer(ctx context.Context, cfg *lazyconn.PeerConf
// activateHAGroupPeers activates all peers in HA groups that the given peer belongs to
func (m *Manager) activateHAGroupPeers(ctx context.Context, triggerPeerID string) {
var peersToActivate []string
m.routesMu.RLock()
haGroups := m.peerToHAGroups[triggerPeerID]
m.routesMu.RUnlock()
if len(haGroups) == 0 {
m.routesMu.RUnlock()
log.Debugf("peer %s is not part of any HA groups", triggerPeerID)
return
}
activatedCount := 0
for _, haGroup := range haGroups {
m.routesMu.RLock()
peers := m.haGroupToPeers[haGroup]
for _, peerID := range peers {
if peerID != triggerPeerID {
peersToActivate = append(peersToActivate, peerID)
}
}
}
m.routesMu.RUnlock()
for _, peerID := range peers {
if peerID == triggerPeerID {
continue
}
activatedCount := 0
for _, peerID := range peersToActivate {
cfg, mp := m.getPeerForActivation(peerID)
if cfg == nil {
continue
@ -342,11 +352,10 @@ func (m *Manager) activateHAGroupPeers(ctx context.Context, triggerPeerID string
if m.activateSinglePeer(ctx, cfg, mp) {
activatedCount++
cfg.Log.Infof("activated peer as part of HA group %s (triggered by %s)", haGroup, triggerPeerID)
cfg.Log.Infof("activated peer as part of HA group (triggered by %s)", triggerPeerID)
m.peerStore.PeerConnOpen(m.engineCtx, cfg.PublicKey)
}
}
}
if activatedCount > 0 {
log.Infof("activated %d additional peers in HA groups for peer %s (groups: %v)",
@ -354,6 +363,51 @@ func (m *Manager) activateHAGroupPeers(ctx context.Context, triggerPeerID string
}
}
// shouldActivateNewPeer checks if a newly added peer should be activated
// because other peers in its HA groups are already active
func (m *Manager) shouldActivateNewPeer(peerID string) (route.HAUniqueID, bool) {
m.routesMu.RLock()
defer m.routesMu.RUnlock()
haGroups := m.peerToHAGroups[peerID]
if len(haGroups) == 0 {
return "", false
}
for _, haGroup := range haGroups {
peers := m.haGroupToPeers[haGroup]
for _, groupPeerID := range peers {
if groupPeerID == peerID {
continue
}
cfg, ok := m.managedPeers[groupPeerID]
if !ok {
continue
}
if mp, ok := m.managedPeersByConnID[cfg.PeerConnID]; ok && mp.expectedWatcher == watcherInactivity {
return haGroup, true
}
}
}
return "", false
}
// activateNewPeerInActiveGroup activates a newly added peer that should be active due to HA group
func (m *Manager) activateNewPeerInActiveGroup(ctx context.Context, peerCfg lazyconn.PeerConfig) {
mp, ok := m.managedPeersByConnID[peerCfg.PeerConnID]
if !ok {
return
}
if !m.activateSinglePeer(ctx, &peerCfg, mp) {
return
}
peerCfg.Log.Infof("activated newly added peer due to active HA group peers")
m.peerStore.PeerConnOpen(m.engineCtx, peerCfg.PublicKey)
}
func (m *Manager) addActivePeer(ctx context.Context, peerCfg lazyconn.PeerConfig) error {
if _, ok := m.managedPeers[peerCfg.PublicKey]; ok {
peerCfg.Log.Warnf("peer already managed")
@ -415,6 +469,48 @@ func (m *Manager) close() {
log.Infof("lazy connection manager closed")
}
// shouldDeferIdleForHA checks if peer should stay connected due to HA group requirements
func (m *Manager) shouldDeferIdleForHA(peerID string) bool {
m.routesMu.RLock()
defer m.routesMu.RUnlock()
haGroups := m.peerToHAGroups[peerID]
if len(haGroups) == 0 {
return false
}
for _, haGroup := range haGroups {
groupPeers := m.haGroupToPeers[haGroup]
for _, groupPeerID := range groupPeers {
if groupPeerID == peerID {
continue
}
cfg, ok := m.managedPeers[groupPeerID]
if !ok {
continue
}
groupMp, ok := m.managedPeersByConnID[cfg.PeerConnID]
if !ok {
continue
}
if groupMp.expectedWatcher != watcherInactivity {
continue
}
// Other member is still connected, defer idle
if peer, ok := m.peerStore.PeerConn(groupPeerID); ok && peer.IsConnected() {
return true
}
}
}
return false
}
func (m *Manager) onPeerActivity(ctx context.Context, peerConnID peerid.ConnID) {
m.managedPeersMu.Lock()
defer m.managedPeersMu.Unlock()
@ -441,7 +537,7 @@ func (m *Manager) onPeerActivity(ctx context.Context, peerConnID peerid.ConnID)
m.peerStore.PeerConnOpen(m.engineCtx, mp.peerCfg.PublicKey)
}
func (m *Manager) onPeerInactivityTimedOut(peerConnID peerid.ConnID) {
func (m *Manager) onPeerInactivityTimedOut(ctx context.Context, peerConnID peerid.ConnID) {
m.managedPeersMu.Lock()
defer m.managedPeersMu.Unlock()
@ -456,6 +552,17 @@ func (m *Manager) onPeerInactivityTimedOut(peerConnID peerid.ConnID) {
return
}
if m.shouldDeferIdleForHA(mp.peerCfg.PublicKey) {
iw, ok := m.inactivityMonitors[peerConnID]
if ok {
mp.peerCfg.Log.Debugf("resetting inactivity timer due to HA group requirements")
iw.ResetMonitor(ctx, m.onInactive)
} else {
mp.peerCfg.Log.Errorf("inactivity monitor not found for HA defer reset")
}
return
}
mp.peerCfg.Log.Infof("connection timed out")
// this is blocking operation, potentially can be optimized
@ -489,7 +596,7 @@ func (m *Manager) onPeerConnected(peerConnID peerid.ConnID) {
iw, ok := m.inactivityMonitors[mp.peerCfg.PeerConnID]
if !ok {
mp.peerCfg.Log.Errorf("inactivity monitor not found for peer")
mp.peerCfg.Log.Warnf("inactivity monitor not found for peer")
return
}

6
client/internal/peer/conn.go
View File

@ -317,12 +317,12 @@ func (conn *Conn) WgConfig() WgConfig {
return conn.config.WgConfig
}
// IsConnected unit tests only
// refactor unit test to use status recorder use refactor status recorded to manage connection status in peer.Conn
// IsConnected returns true if the peer is connected
func (conn *Conn) IsConnected() bool {
conn.mu.Lock()
defer conn.mu.Unlock()
return conn.currentConnPriority != conntype.None
return conn.evalStatus() == StatusConnected
}
func (conn *Conn) GetKey() string {