puller(ticdc): fix stuck detect issue (#10258) (#10259)

close #10256

cdc/puller/puller.go

@@ -77,6 +77,8 @@ type pullerImpl struct {
 	cfg                   *config.ServerConfig
 	lastForwardTime       time.Time
 	lastForwardResolvedTs uint64
+	// startResolvedTs is the resolvedTs when puller is initialized
+	startResolvedTs uint64
 }
 
 // New create a new Puller fetch event start from checkpointTs and put into buf.
@@ -126,6 +128,8 @@ func New(ctx context.Context,
 		tableID:      tableID,
 		tableName:    tableName,
 		cfg:          cfg,
+
+		startResolvedTs: checkpointTs,
 	}
 	return p
 }
@@ -213,7 +217,7 @@ func (p *pullerImpl) Run(ctx context.Context) error {
 				metricOutputChanSize.Observe(float64(len(p.outputCh)))
 				metricPullerResolvedTs.Set(float64(oracle.ExtractPhysical(atomic.LoadUint64(&p.resolvedTs))))
 			case <-stuckDetectorTicker.C:
-				if err := p.detectResolvedTsStuck(initialized); err != nil {
+				if err := p.detectResolvedTsStuck(); err != nil {
 					return errors.Trace(err)
 				}
 				continue
@@ -280,9 +284,15 @@ func (p *pullerImpl) GetResolvedTs() uint64 {
 	return atomic.LoadUint64(&p.resolvedTs)
 }
 
-func (p *pullerImpl) detectResolvedTsStuck(initialized bool) error {
-	if p.cfg.Debug.Puller.EnableResolvedTsStuckDetection && initialized {
+func (p *pullerImpl) detectResolvedTsStuck() error {
+	if p.cfg.Debug.Puller.EnableResolvedTsStuckDetection {
 		resolvedTs := p.tsTracker.Frontier()
+		// check if the resolvedTs is advancing,
+		// If the resolvedTs in Frontier is less than startResolvedTs, it means that the incremental scan has
+		// not complete yet. We need to make no decision in this scenario.
+		if resolvedTs <= p.startResolvedTs {
+			return nil
+		}
 		if resolvedTs == p.lastForwardResolvedTs {
 			log.Warn("ResolvedTs stuck detected in puller",
 				zap.String("namespace", p.changefeed.Namespace),