sendnrw/netbird
2
0
Fork 0
mirror of https://github.com/netbirdio/netbird.git synced 2026-09-30 18:49:08 +02:00
Code Issues Packages Projects Releases Wiki Activity

Log signal stall, ICE pair selection, restart cadence, sync content, and receive backpressure to attribute the regression

Browse Source
This commit is contained in:
Viktor Liu
2026-06-24 15:04:22 +02:00
parent 4cb2c62f2a
commit 12f2e69af2
4 changed files with 109 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
client/internal/engine.go
+24 -1
View File
@@ -14,6 +14,7 @@ import (
"sort"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/hashicorp/go-multierror"
@@ -88,6 +89,13 @@ var ErrResetConnection = fmt.Errorf("reset connection")
var ErrEngineAlreadyStarted = errors.New("engine already started")
// engineRestartCount and engineLastRestart track client-restart cadence across
// engine recreations so a restart loop is distinguishable from rare restarts.
var (
engineRestartCount atomic.Int64
engineLastRestart atomic.Int64
)
type EngineConfig struct {
WgPort int
WgIfaceName string
@@ -910,6 +918,14 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
return e.ctx.Err()
}
if nm := update.GetNetworkMap(); nm != nil {
log.Infof("sync update: serial=%d remotePeers=%d offlinePeers=%d routes=%d firewallRules=%d checks=%d configPresent=%v remotePeersEmpty=%v",
nm.GetSerial(), len(nm.GetRemotePeers()), len(nm.GetOfflinePeers()), len(nm.GetRoutes()),
len(nm.GetFirewallRules()), len(update.GetChecks()), update.GetNetbirdConfig() != nil, nm.GetRemotePeersIsEmpty())
} else {
log.Infof("sync update: config-only (no network map), configPresent=%v", update.GetNetbirdConfig() != nil)
}
if update.NetworkMap != nil && update.NetworkMap.PeerConfig != nil {
e.handleAutoUpdateVersion(update.NetworkMap.PeerConfig.AutoUpdate)
}
@@ -2171,7 +2187,14 @@ func (e *Engine) triggerClientRestart() {
return
}
log.Info("restarting engine")
// Cadence survives engine recreation (package-level), so a restart loop shows
// as a fast-climbing count with a short gap, distinct from rare intentional restarts.
n := engineRestartCount.Add(1)
var sinceLast time.Duration
if prev := engineLastRestart.Swap(time.Now().UnixNano()); prev != 0 {
sinceLast = time.Since(time.Unix(0, prev))
}
log.Infof("restarting engine (restart #%d, %s since previous)", n, sinceLast.Round(time.Second))
CtxGetState(e.ctx).Set(StatusConnecting)
_ = CtxGetState(e.ctx).Wrap(ErrResetConnection)
log.Infof("cancelling client context, engine will be recreated")
client/internal/peer/worker_ice.go
+1 -1
View File
@@ -461,7 +461,7 @@ func (w *WorkerICE) createForwardedCandidate(srflxCandidate ice.Candidate, mappi
}
func (w *WorkerICE) onICESelectedCandidatePair(agent *icemaker.ThreadSafeAgent, c1, c2 ice.Candidate) {
w.log.Debugf("selected candidate pair [local <-> remote] -> [%s <-> %s], peer %s", c1.String(), c2.String(),
w.log.Infof("selected candidate pair [local <-> remote] -> [%s <-> %s], peer %s", c1.String(), c2.String(),
w.config.Key)
pairStat, ok := agent.GetSelectedCandidatePairStats()