ticdc: Fix CPU surge problem in sorter

cdc/processor/sourcemanager/sorter/pebble/event_sorter.go

@@ -392,14 +392,16 @@
 		case batchEvent := <-batchCh:
 			// do batch commit
 			batch := batchEvent.batch
-			writeBytes.Observe(float64(len(batch.Repr())))
-			start := time.Now()
-			if err := batch.Commit(&pebbleWriteOptions); err != nil {
-				log.Panic("failed to commit pebble batch", zap.Error(err),
-					zap.String("namespace", s.changefeedID.Namespace),
-					zap.String("changefeed", s.changefeedID.ID))
+			if !batch.Empty() {
+				writeBytes.Observe(float64(len(batch.Repr())))
+				start := time.Now()
+				if err := batch.Commit(&pebbleWriteOptions); err != nil {
+					log.Panic("failed to commit pebble batch", zap.Error(err),
+						zap.String("namespace", s.changefeedID.Namespace),
+						zap.String("changefeed", s.changefeedID.ID))
+				}
+				writeDuration.Observe(time.Since(start).Seconds())
 			}
-			writeDuration.Observe(time.Since(start).Seconds())
 
 			// update resolved ts after commit successfully
 			batchResolved := batchEvent.batchResolved
@@ -440,11 +442,10 @@
 	s.wg.Add(1)
 	go s.batchCommitAndUpdateResolvedTs(batchCh, id)
 
-	batch := db.NewBatch()
-	newResolved := spanz.NewHashMap[model.Ts]()
-	startToCollectBatch := time.Now()
+	ticker := time.NewTicker(batchCommitInterval / 2)
+	defer ticker.Stop()
 
-	handleItem := func(item eventWithTableID) {
+	encodeItemAndBatch := func(batch *pebble.Batch, newResolved *spanz.HashMap[model.Ts], item eventWithTableID) {
 		if item.event.IsResolved() {
 			newResolved.ReplaceOrInsert(item.span, item.event.CRTs)
 			return
@@ -463,21 +464,36 @@
 		}
 	}
 
-	for {
-		for len(batch.Repr()) < batchCommitSize && time.Since(startToCollectBatch) < batchCommitInterval {
+	// Batch item and commit until batch size is larger than batchCommitSize,
+	// or the time since the last commit is larger than batchCommitInterval.
+	// Only return false when the sorter is closed.
+	doBatching := func() (*DBBatchEvent, bool) {
+		batch := db.NewBatch()
+		newResolved := spanz.NewHashMap[model.Ts]()
+		startToBatch := time.Now()
+		for {
 			select {
 			case item := <-inputCh:
-				handleItem(item)
+				encodeItemAndBatch(batch, newResolved, item)
+				if len(batch.Repr()) >= batchCommitSize {
+					return &DBBatchEvent{batch, newResolved}, true
+				}
 			case <-s.closed:
-				return
-			default:
+				return nil, false
+			case <-ticker.C:
+				if time.Since(startToBatch) >= batchCommitInterval {
+					return &DBBatchEvent{batch, newResolved}, true
+				}
 			}
 		}
-		batchCh <- &DBBatchEvent{batch, newResolved}
+	}
 
-		batch = db.NewBatch()
-		newResolved = spanz.NewHashMap[model.Ts]()
-		startToCollectBatch = time.Now()
+	for {
+		batchEvent, ok := doBatching()
+		if !ok {
+			return
+		}
+		batchCh <- batchEvent
 	}
 }