diff --git a/eth/tracers/live/filter.go b/eth/tracers/live/filter.go
index 956115df15..d09a5b9df8 100644
--- a/eth/tracers/live/filter.go
+++ b/eth/tracers/live/filter.go
@@ -41,11 +41,14 @@ type filter struct {
 	backend    tracers.Backend
 	kvdb       ethdb.Database
 	frdb       *rawdb.Freezer
+	blockCh    chan uint64
+	stopCh     chan struct{}
 	tables     map[string]bool
 	traces     map[string][]*traceResult
 	tracer     *native.MuxTracer
 	latest     atomic.Uint64
 	offset     atomic.Uint64
+	finalized  atomic.Uint64
 	hash       common.Hash
 	once       sync.Once
 	offsetFile string
@@ -132,6 +135,8 @@ func newFilter(cfg json.RawMessage, backend tracers.Backend) (*tracing.Hooks, []
 		backend:    backend,
 		kvdb:       kvdb,
 		frdb:       frdb,
+		blockCh:    make(chan uint64, 100),
+		stopCh:     make(chan struct{}),
 		tables:     tables,
 		traces:     traces,
 		tracer:     t,
@@ -176,20 +181,25 @@ func newFilter(cfg json.RawMessage, backend tracers.Backend) (*tracing.Hooks, []
 			Service:   &filterAPI{backend: backend, filter: f},
 		},
 	}
+
+	// Initialize head if it doesn't exist
+	head, _ := f.getFreezerHeadTail()
+	if head == 0 {
+		f.updateFreezerHead(f.latest.Load())
+	}
+	go f.freeze()
+
 	return hooks, apis, nil
 }
 
 func (f *filter) OnBlockStart(ev tracing.BlockEvent) {
 	// track the latest block number
 	blknum := ev.Block.NumberU64()
-	// latest := f.latest.Load()
-	// if blknum < latest {
-	// 	// TODO: handle the case of setHead
-	// 	log.Error("OnBlockStart received an old block", "latest", latest, "number", blknum)
-	// 	return
-	// }
 	f.latest.Store(blknum)
 	f.hash = ev.Block.Hash()
+	if ev.Finalized != nil {
+		f.finalized.Store(ev.Finalized.Number.Uint64())
+	}
 
 	// reset local cache
 	txs := ev.Block.Transactions().Len()
@@ -204,18 +214,6 @@ func (f *filter) OnBlockStart(ev tracing.BlockEvent) {
 			os.WriteFile(f.offsetFile, []byte(fmt.Sprintf("%d", blknum)), 0666)
 		}
 	})
-
-	// // truncate the freezer db if the block number is less than the latest
-	// if blknum <= latest {
-	// 	frozen, _ := f.frdb.Ancients()
-	// 	offset := f.offset.Load()
-	// 	log.Info("Reorg detected", "number", blknum, "latest", latest, "offset", offset, "frozen", frozen)
-	// 	if _, err := f.frdb.TruncateHead(blknum - offset); err != nil {
-	// 		log.Error("Failed to truncate filter tracer db", "error", err)
-	// 		// TODO: how to handle this error?
-	// 		return
-	// 	}
-	// }
 }
 
 func (f *filter) OnTxStart(env *tracing.VMContext, tx *types.Transaction, from common.Address) {
@@ -240,7 +238,7 @@ func (f *filter) OnTxEnd(receipt *types.Receipt, err error) {
 
 func (f *filter) OnBlockEnd(err error) {
 	if err != nil {
-		log.Warn("OnBlockEnd", "latest", f.latest.Load(), "err", err)
+		log.Warn("OnBlockEnd", "latest", f.latest.Load(), "error", err)
 	}
 	batch := f.kvdb.NewBatch()
 
@@ -250,32 +248,60 @@ func (f *filter) OnBlockEnd(err error) {
 		data, err := json.Marshal(traces)
 		if err != nil {
 			log.Error("Failed to marshal traces", "error", err)
+			break
 		}
 		batch.Put(toKVKey(name, number, hash), data)
 	}
 	if err := batch.Write(); err != nil {
-		log.Error("Failed to write", "err", err)
+		log.Error("Failed to write", "error", err)
+		return
 	}
 
-	// f.frdb.ModifyAncients(func(w ethdb.AncientWriteOp) error {
-	// 	latest := f.latest.Load()
-	// 	offset := f.offset.Load()
-	// 	number := latest - offset
-	// 	for name, traces := range f.traces {
-	// 		data, err := json.Marshal(traces)
-	// 		if err != nil {
-	// 			log.Error("Failed to marshal traces", "error", err)
-	// 		}
-	// 		table := toTraceTable(name)
-	// 		if err := w.AppendRaw(table, number, data); err != nil {
-	// 			log.Error("Failed to write block traces", "number", latest, "table", table, "error", err)
-	// 		}
-	// 	}
-	// 	return nil
-	// })
+	select {
+	case f.blockCh <- f.finalized.Load():
+	default:
+		// Channel is full, log a warning
+		log.Warn("Block channel is full, skipping finalized block notification")
+	}
 }
 
 func (f *filter) readBlockTraces(ctx context.Context, name string, blknum uint64) ([]*traceResult, error) {
+	if blknum > f.latest.Load() {
+		return nil, errors.New("notfound")
+	}
+	if blknum < f.offset.Load() {
+		return nil, errors.New("historical data not available")
+	}
+
+	_, tail := f.getFreezerHeadTail()
+
+	// If tail is 0 (not found in kvdb), use the offset
+	if tail == 0 {
+		tail = f.offset.Load()
+	}
+
+	// Determine whether to read from kvdb or frdb
+	var (
+		data []byte
+		err  error
+	)
+	if blknum >= tail {
+		// Data is in kvdb
+		data, err = f.readFromKVDB(ctx, name, blknum)
+	} else {
+		// Data is in frdb
+		data, err = f.readFromFRDB(name, blknum)
+	}
+	if err != nil {
+		return nil, err
+	}
+
+	var traces []*traceResult
+	err = json.Unmarshal(data, &traces)
+	return traces, err
+}
+
+func (f *filter) readFromKVDB(ctx context.Context, name string, blknum uint64) ([]byte, error) {
 	header, err := f.backend.HeaderByNumber(ctx, rpc.BlockNumber(blknum))
 	if err != nil {
 		return nil, err
@@ -284,36 +310,28 @@ func (f *filter) readBlockTraces(ctx context.Context, name string, blknum uint64
 	kvKey := toKVKey(name, blknum, header.Hash())
 	data, err := f.kvdb.Get(kvKey)
 	if err != nil {
-		return nil, err
+		return nil, fmt.Errorf("traces not found in kvdb for block %d: %w", blknum, err)
 	}
-	var traces []*traceResult
-	err = json.Unmarshal(data, &traces)
-	return traces, err
+	return data, err
+}
 
-	// table := toTraceTable(name)
-	// if _, ok := f.tables[table]; !ok {
-	// 	return nil, errors.New("tracer not found")
-	// }
-	//
-	// if blknum < f.offset.Load() || blknum > f.latest.Load() {
-	// 	return nil, nil
-	// }
-	//
-	// var data []byte
-	// err := f.frdb.ReadAncients(func(reader ethdb.AncientReaderOp) error {
-	// 	var err error
-	// 	data, err = reader.Ancient(table, blknum-f.offset.Load())
-	// 	return err
-	// })
-	// if err != nil {
-	// 	return nil, err
-	// }
-	// var traces []*traceResult
-	// err = json.Unmarshal(data, &traces)
-	// return traces, err
+func (f *filter) readFromFRDB(name string, blknum uint64) ([]byte, error) {
+	table := toTraceTable(name)
+	var data []byte
+	err := f.frdb.ReadAncients(func(reader ethdb.AncientReaderOp) error {
+		var err error
+		data, err = reader.Ancient(table, blknum-f.offset.Load())
+		return err
+	})
+	if err != nil {
+		return nil, fmt.Errorf("traces not found in frdb for block %d: %w", blknum, err)
+	}
+	return data, nil
 }
 
 func (f *filter) Close() {
+	close(f.stopCh)
+
 	if err := f.kvdb.Close(); err != nil {
 		log.Error("Close kvdb failed", "err", err)
 	}
diff --git a/eth/tracers/live/filter_freeze.go b/eth/tracers/live/filter_freeze.go
new file mode 100644
index 0000000000..78503d8462
--- /dev/null
+++ b/eth/tracers/live/filter_freeze.go
@@ -0,0 +1,154 @@
+package live
+
+import (
+	"context"
+	"encoding/binary"
+	"fmt"
+
+	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/log"
+	"github.com/ethereum/go-ethereum/rpc"
+)
+
+const (
+	freezeThreshold = 64
+	kvdbHeadKey     = "FilterFreezerHead"
+	kvdbTailKey     = "FilterFreezerTail"
+)
+
+func (f *filter) freeze() {
+	var lastFinalized uint64
+	for {
+		select {
+		case <-f.stopCh:
+			return
+		case finalizedBlock := <-f.blockCh:
+			if finalizedBlock <= lastFinalized {
+				continue
+			}
+			lastFinalized = finalizedBlock
+
+			head, tail := f.getFreezerHeadTail()
+
+			// If tail is 0 (not found in kvdb), use the offset
+			if tail == 0 {
+				tail = f.offset.Load()
+			}
+
+			// Freeze at most freezeThreshold blocks
+			freezeUpTo := finalizedBlock
+			freezeUpTo = min(freezeUpTo, tail+freezeThreshold)
+
+			log.Info("Move traces from kvdb to frdb", "from", tail, "to", freezeUpTo)
+			for blknum := tail; blknum < freezeUpTo; blknum++ {
+				if err := f.moveBlockToFreezer(blknum); err != nil {
+					log.Error("Failed to move block to freezer", "block", blknum, "error", err)
+					break
+				}
+			}
+
+			// Update head and tail
+			if freezeUpTo > tail {
+				if err := f.updateFreezerTail(freezeUpTo); err != nil {
+					log.Error("Failed to update freezer tail", "error", err)
+				}
+			}
+			if freezeUpTo > head {
+				if err := f.updateFreezerHead(freezeUpTo); err != nil {
+					log.Error("Failed to update freezer head", "error", err)
+				}
+			}
+		}
+	}
+}
+
+func (f *filter) getFreezerHeadTail() (head, tail uint64) {
+	headBytes, _ := f.kvdb.Get([]byte(kvdbHeadKey))
+	tailBytes, _ := f.kvdb.Get([]byte(kvdbTailKey))
+
+	if len(headBytes) > 0 {
+		head = binary.BigEndian.Uint64(headBytes)
+	}
+	if len(tailBytes) > 0 {
+		tail = binary.BigEndian.Uint64(tailBytes)
+	}
+	return
+}
+
+func (f *filter) updateFreezerHead(head uint64) error {
+	headBytes := make([]byte, 8)
+	binary.BigEndian.PutUint64(headBytes, head)
+	return f.kvdb.Put([]byte(kvdbHeadKey), headBytes)
+}
+
+func (f *filter) updateFreezerTail(tail uint64) error {
+	tailBytes := make([]byte, 8)
+	binary.BigEndian.PutUint64(tailBytes, tail)
+	return f.kvdb.Put([]byte(kvdbTailKey), tailBytes)
+}
+
+func (f *filter) moveBlockToFreezer(blknum uint64) error {
+	header, err := f.backend.HeaderByNumber(context.Background(), rpc.BlockNumber(blknum))
+	if err != nil {
+		return err
+	}
+
+	offset := f.offset.Load()
+	for name := range f.tracer.Tracers() {
+		kvKey := toKVKey(name, blknum, header.Hash())
+		data, err := f.kvdb.Get(kvKey)
+		if err != nil {
+			return err
+		}
+
+		table := toTraceTable(name)
+		n, err := f.frdb.ModifyAncients(func(op ethdb.AncientWriteOp) error {
+			return op.AppendRaw(table, blknum-offset, data)
+		})
+		if err != nil {
+			return err
+		}
+		log.Info("Move from kvdb to frdb", "blknum", blknum, "size", n)
+
+		// Delete all entries for this prefix from kvdb, ignore error
+		prefix := append([]byte(name), encodeBlockNumber(blknum)...)
+		if err := f.deleteKVDBEntriesWithPrefix(prefix); err != nil {
+			log.Error("Failed to delete entries from kvdb", "error", err)
+		}
+	}
+
+	return nil
+}
+
+func (f *filter) deleteKVDBEntriesWithPrefix(prefix []byte) error {
+	batch := f.kvdb.NewBatch()
+	it := f.kvdb.NewIterator(prefix, nil)
+	defer it.Release()
+
+	for it.Next() {
+		if err := batch.Delete(it.Key()); err != nil {
+			return fmt.Errorf("failed to add delete operation to batch: %w", err)
+		}
+
+		// Write batch if it's getting too large
+		if batch.ValueSize() > ethdb.IdealBatchSize {
+			if err := batch.Write(); err != nil {
+				return fmt.Errorf("failed to write batch: %w", err)
+			}
+			batch.Reset()
+		}
+	}
+
+	if err := it.Error(); err != nil {
+		return fmt.Errorf("iterator error: %w", err)
+	}
+
+	// Write any remaining batch operations
+	if batch.ValueSize() > 0 {
+		if err := batch.Write(); err != nil {
+			return fmt.Errorf("failed to write final batch: %w", err)
+		}
+	}
+
+	return nil
+}