diff --git a/weed/mq/kafka/protocol/fetch_multibatch.go b/weed/mq/kafka/protocol/fetch_multibatch.go
index 0eff5da37..7f8fb8ff5 100644
--- a/weed/mq/kafka/protocol/fetch_multibatch.go
+++ b/weed/mq/kafka/protocol/fetch_multibatch.go
@@ -51,7 +51,7 @@ func (f *MultiBatchFetcher) FetchMultipleBatches(topicName string, partitionID i
 	currentOffset := startOffset
 	totalSize := int32(0)
 	batchCount := 0
-	
+
 	// Parameters for batch fetching - start smaller to respect maxBytes better
 	recordsPerBatch := int32(10) // Start with smaller batch size
 	maxBatchesPerFetch := 10     // Limit number of batches to avoid infinite loops
@@ -85,31 +85,19 @@ func (f *MultiBatchFetcher) FetchMultipleBatches(topicName string, partitionID i
 
 		// Estimate batch size before construction to better respect maxBytes
 		estimatedBatchSize := f.estimateBatchSize(smqRecords)
-		
-		// Check if this batch would exceed maxBytes BEFORE constructing it
-		if totalSize+estimatedBatchSize > maxBytes && batchCount > 0 {
-			fmt.Printf("DEBUG: MultiBatch - estimated batch would exceed limit (%d + %d > %d), stopping\n", 
-				totalSize, estimatedBatchSize, maxBytes)
-			break
-		}
-		
-		// Special case: If this is the first batch and it's already too big, 
-		// we still need to include it (Kafka behavior - always return at least some data)
-		if batchCount == 0 && estimatedBatchSize > maxBytes {
-			fmt.Printf("DEBUG: MultiBatch - first batch estimated size %d exceeds maxBytes %d, but including anyway\n", 
-				estimatedBatchSize, maxBytes)
-		}
+
+		// Note: we do not stop based on estimate; we will check actual size after constructing the batch
 
 		// Construct record batch
 		batch := f.constructSingleRecordBatch(currentOffset, smqRecords)
 		batchSize := int32(len(batch))
-		
-		fmt.Printf("DEBUG: MultiBatch - constructed batch %d: %d records, %d bytes (estimated %d), offset %d\n", 
+
+		fmt.Printf("DEBUG: MultiBatch - constructed batch %d: %d records, %d bytes (estimated %d), offset %d\n",
 			batchCount+1, len(smqRecords), batchSize, estimatedBatchSize, currentOffset)
 
 		// Double-check actual size doesn't exceed maxBytes
 		if totalSize+batchSize > maxBytes && batchCount > 0 {
-			fmt.Printf("DEBUG: MultiBatch - actual batch would exceed limit (%d + %d > %d), stopping\n", 
+			fmt.Printf("DEBUG: MultiBatch - actual batch would exceed limit (%d + %d > %d), stopping\n",
 				totalSize, batchSize, maxBytes)
 			break
 		}
@@ -134,7 +122,7 @@ func (f *MultiBatchFetcher) FetchMultipleBatches(topicName string, partitionID i
 		BatchCount:    batchCount,
 	}
 
-	fmt.Printf("DEBUG: MultiBatch - completed: %d batches, %d total bytes, next offset %d\n", 
+	fmt.Printf("DEBUG: MultiBatch - completed: %d batches, %d total bytes, next offset %d\n",
 		result.BatchCount, result.TotalSize, result.NextOffset)
 
 	return result, nil
@@ -450,42 +438,62 @@ func (f *MultiBatchFetcher) constructCompressedRecordBatch(baseOffset int64, com
 // estimateBatchSize estimates the size of a record batch before constructing it
 func (f *MultiBatchFetcher) estimateBatchSize(smqRecords []offset.SMQRecord) int32 {
 	if len(smqRecords) == 0 {
-		return 61 // empty batch size
+		return 61 // empty batch header size
 	}
 
-	// Record batch header: 61 bytes
+	// Record batch header: 61 bytes (base_offset + batch_length + leader_epoch + magic + crc + attributes +
+	// last_offset_delta + first_ts + max_ts + producer_id + producer_epoch + base_seq + record_count)
 	headerSize := int32(61)
-	
-	// Estimate records size
+
+	baseTs := smqRecords[0].GetTimestamp()
 	recordsSize := int32(0)
-	for _, record := range smqRecords {
-		// Each record has overhead: attributes(1) + timestamp_delta(varint) + offset_delta(varint) + headers(varint)
-		recordOverhead := int32(10) // rough estimate for varints and overhead
-		
-		keySize := int32(0)
-		if record.GetKey() != nil {
-			keySize = int32(len(record.GetKey())) + 5 // +5 for length varint
+	for i, rec := range smqRecords {
+		// attributes(1)
+		rb := int32(1)
+
+		// timestamp_delta(varint)
+		tsDelta := rec.GetTimestamp() - baseTs
+		rb += int32(len(encodeVarint(tsDelta)))
+
+		// offset_delta(varint)
+		rb += int32(len(encodeVarint(int64(i))))
+
+		// key length varint + data or -1
+		if k := rec.GetKey(); k != nil {
+			rb += int32(len(encodeVarint(int64(len(k))))) + int32(len(k))
 		} else {
-			keySize = 1 // -1 encoded as varint
+			rb += int32(len(encodeVarint(-1)))
 		}
-		
-		valueSize := int32(0)
-		if record.GetValue() != nil {
-			valueSize = int32(len(record.GetValue())) + 5 // +5 for length varint
+
+		// value length varint + data or -1
+		if v := rec.GetValue(); v != nil {
+			rb += int32(len(encodeVarint(int64(len(v))))) + int32(len(v))
 		} else {
-			valueSize = 1 // -1 encoded as varint
+			rb += int32(len(encodeVarint(-1)))
 		}
-		
-		// Record length itself is also encoded as varint
-		recordLength := recordOverhead + keySize + valueSize
-		recordLengthVarintSize := int32(5) // conservative estimate for varint
-		
-		recordsSize += recordLengthVarintSize + recordLength
+
+		// headers count (varint = 0)
+		rb += int32(len(encodeVarint(0)))
+
+		// prepend record length varint
+		recordsSize += int32(len(encodeVarint(int64(rb)))) + rb
 	}
-	
+
 	return headerSize + recordsSize
 }
 
+// sizeOfVarint returns the number of bytes encodeVarint would use for value
+func sizeOfVarint(value int64) int32 {
+	// ZigZag encode to match encodeVarint
+	u := uint64(uint64(value<<1) ^ uint64(value>>63))
+	size := int32(1)
+	for u >= 0x80 {
+		u >>= 7
+		size++
+	}
+	return size
+}
+
 // compressData compresses data using the specified codec (basic implementation)
 func (f *MultiBatchFetcher) compressData(data []byte, codec compression.CompressionCodec) ([]byte, error) {
 	// For Phase 5, implement basic compression support