diff --git a/internal/datanode/data_node.go b/internal/datanode/data_node.go
index 0601330164..3e9e343a87 100644
--- a/internal/datanode/data_node.go
+++ b/internal/datanode/data_node.go
@@ -77,6 +77,7 @@ const illegalRequestErrStr = "Illegal request"
 // makes sure DataNode implements types.DataNode
 var _ types.DataNode = (*DataNode)(nil)
 
+// Params from config.yaml
 var Params paramtable.GlobalParamTable
 
 // DataNode communicates with outside services and unioun all
@@ -89,9 +90,6 @@ var Params paramtable.GlobalParamTable
 //  `NodeID` is unique to each datanode.
 //  `State` is current statement of this data node, indicating whether it's healthy.
 //
-//  `vchan2SyncService` is a map of vchannlName to dataSyncService, so that datanode
-//  has ability to scale flowgraph.
-//  `vchan2FlushCh` holds flush-signal channels for every flowgraph.
 //  `clearSignal` is a signal channel for releasing the flowgraph resources.
 //  `segmentCache` stores all flushing and flushed segments.
 type DataNode struct {
@@ -101,10 +99,8 @@ type DataNode struct {
 	Role   string
 	State  atomic.Value // internalpb.StateCode_Initializing
 
-	// TODO struct
-	chanMut           sync.RWMutex
-	vchan2SyncService map[string]*dataSyncService // vchannel name
-	vchan2FlushChs    map[string]chan flushMsg    // vchannel name to flush channels
+	flowgraphManager *flowgraphManager
+	eventManagerMap  sync.Map // vchannel name -> channelEventManager
 
 	clearSignal        chan string // vchannel name
 	segmentCache       *Cache
@@ -138,15 +134,14 @@ func NewDataNode(ctx context.Context, factory msgstream.Factory) *DataNode {
 		segmentCache:       newCache(),
 		compactionExecutor: newCompactionExecutor(),
 
-		vchan2SyncService: make(map[string]*dataSyncService),
-		vchan2FlushChs:    make(map[string]chan flushMsg),
-		clearSignal:       make(chan string, 100),
+		flowgraphManager: newFlowgraphManager(),
+		clearSignal:      make(chan string, 100),
 	}
 	node.UpdateStateCode(internalpb.StateCode_Abnormal)
 	return node
 }
 
-// Set etcd client
+// SetEtcdClient sets etcd client for DataNode
 func (node *DataNode) SetEtcdClient(etcdCli *clientv3.Client) {
 	node.etcdCli = etcdCli
 }
@@ -287,125 +282,134 @@ func (node *DataNode) checkWatchedList() error {
 		return err
 	}
 	for i, val := range values {
-		node.handleWatchInfo(keys[i], []byte(val))
+		node.handleWatchInfo(&event{eventType: putEventType}, keys[i], []byte(val))
 	}
 	return nil
 }
 
 // handleChannelEvt handles event from kv watch event
 func (node *DataNode) handleChannelEvt(evt *clientv3.Event) {
+	var e *event
 	switch evt.Type {
 	case clientv3.EventTypePut: // datacoord shall put channels needs to be watched here
-		log.Debug("DataNode handleChannelEvt EventTypePut", zap.String("key", string(evt.Kv.Key)))
-		node.handleWatchInfo(string(evt.Kv.Key), evt.Kv.Value)
+		e = &event{
+			eventType: putEventType,
+		}
+
 	case clientv3.EventTypeDelete:
-		// guaranteed there is no "/" in channel name
-		parts := strings.Split(string(evt.Kv.Key), "/")
-		vchanName := parts[len(parts)-1]
-		log.Debug("DataNode handleChannelEvt EventTypeDelete",
-			zap.String("key", string(evt.Kv.Key)),
-			zap.String("vChanName", vchanName),
-			zap.Int64("node id", Params.DataNodeCfg.NodeID))
-		node.ReleaseDataSyncService(vchanName)
+		e = &event{
+			eventType: deleteEventType,
+		}
+	}
+	node.handleWatchInfo(e, string(evt.Kv.Key), evt.Kv.Value)
+}
+
+func (node *DataNode) handleWatchInfo(e *event, key string, data []byte) {
+	switch e.eventType {
+	case putEventType:
+		log.Info("DataNode is handling watchInfo put event", zap.String("key", key))
+
+		watchInfo, err := parsePutEventData(data)
+		if err != nil {
+			log.Warn("fail to handle watchInfo", zap.Int("event type", e.eventType), zap.String("key", key), zap.Error(err))
+			return
+		}
+
+		e.info = watchInfo
+		e.vChanName = watchInfo.GetVchan().GetChannelName()
+
+	case deleteEventType:
+		log.Info("DataNode is handling watchInfo delete event", zap.String("key", key))
+		e.vChanName = parseDeleteEventKey(key)
+	}
+
+	actualManager, loaded := node.eventManagerMap.LoadOrStore(e.vChanName, &channelEventManager{
+		eventChan:         make(chan event, 10),
+		closeChan:         make(chan struct{}),
+		handlePutEvent:    node.handlePutEvent,
+		handleDeleteEvent: node.handleDeleteEvent,
+	})
+
+	if !loaded {
+		actualManager.(*channelEventManager).Run()
+	}
+
+	actualManager.(*channelEventManager).handleEvent(*e)
+
+	// Whenever a delete event comes, this eventManger will be removed from map
+	if e.eventType == deleteEventType {
+		if m, loaded := node.eventManagerMap.LoadAndDelete(e.vChanName); loaded {
+			m.(*channelEventManager).Close()
+		}
 	}
 }
 
-func (node *DataNode) handleWatchInfo(key string, data []byte) {
+func parsePutEventData(data []byte) (*datapb.ChannelWatchInfo, error) {
 	watchInfo := datapb.ChannelWatchInfo{}
 	err := proto.Unmarshal(data, &watchInfo)
 	if err != nil {
-		log.Warn("fail to parse ChannelWatchInfo", zap.String("key", key), zap.Error(err))
-		return
+		return nil, fmt.Errorf("invalid event data: fail to parse ChannelWatchInfo, err: %v", err)
 	}
-	log.Debug("DataNode handleWatchInfo Unmarshal success", zap.String("key", key))
+
 	if watchInfo.State == datapb.ChannelWatchState_Complete {
-		log.Warn("DataNode handleWatchInfo State is already ChannelWatchState_Complete", zap.String("key", key))
-		return
+		return nil, fmt.Errorf("invalid event: event state is already ChannelWatchState_Compele")
 	}
+
 	if watchInfo.Vchan == nil {
-		log.Warn("found ChannelWatchInfo with nil VChannelInfo", zap.String("key", key))
-		return
+		return nil, fmt.Errorf("invalid event: ChannelWatchInfo with nil VChannelInfo")
 	}
-	log.Warn("DataNode handleWatchInfo try to NewDataSyncService", zap.String("key", key))
-	err = node.NewDataSyncService(watchInfo.Vchan)
-	if err != nil {
-		log.Warn("fail to create DataSyncService", zap.String("key", key), zap.Error(err))
-		return
-	}
-	log.Warn("DataNode handleWatchInfo NewDataSyncService success", zap.String("key", key))
 
-	watchInfo.State = datapb.ChannelWatchState_Complete
-	v, err := proto.Marshal(&watchInfo)
-	if err != nil {
-		log.Warn("DataNode handleWatchInfo fail to Marshal watchInfo", zap.String("key", key), zap.Error(err))
-		return
-	}
-	k := path.Join(Params.DataNodeCfg.ChannelWatchSubPath, fmt.Sprintf("%d", node.NodeID), watchInfo.GetVchan().GetChannelName())
-	log.Warn("DataNode handleWatchInfo try to Save", zap.String("key", key),
-		zap.String("k", k),
-		zap.String("v", string(v)))
-
-	err = node.watchKv.Save(k, string(v))
-	if err != nil {
-		log.Warn("DataNode handleWatchInfo fail to change WatchState to complete", zap.String("key", key), zap.Error(err))
-		node.ReleaseDataSyncService(key)
-	}
+	return &watchInfo, nil
 }
 
-// NewDataSyncService adds a new dataSyncService for new dmlVchannel and starts dataSyncService.
-func (node *DataNode) NewDataSyncService(vchan *datapb.VchannelInfo) error {
-	node.chanMut.RLock()
-	if _, ok := node.vchan2SyncService[vchan.GetChannelName()]; ok {
-		node.chanMut.RUnlock()
-		return nil
-	}
-	node.chanMut.RUnlock()
+func parseDeleteEventKey(key string) string {
+	parts := strings.Split(key, "/")
+	vChanName := parts[len(parts)-1]
+	return vChanName
+}
 
-	replica, err := newReplica(node.ctx, node.rootCoord, vchan.CollectionID)
+func (node *DataNode) handlePutEvent(watchInfo *datapb.ChannelWatchInfo) error {
+	vChanName := watchInfo.GetVchan().GetChannelName()
+
+	if err := node.flowgraphManager.addAndStart(node, watchInfo.GetVchan()); err != nil {
+		return fmt.Errorf("fail to add and start flowgraph for vChanName: %s, err: %v", vChanName, err)
+	}
+	log.Debug("handle put event: new data sync service success", zap.String("vChanName", vChanName))
+
+	watchInfo.State = datapb.ChannelWatchState_Complete
+	v, err := proto.Marshal(watchInfo)
 	if err != nil {
-		return err
+		return fmt.Errorf("fail to marshal watchInfo with complete state, vChanName: %s, err: %v", vChanName, err)
 	}
 
-	var alloc allocatorInterface = newAllocator(node.rootCoord)
+	k := path.Join(Params.DataNodeCfg.ChannelWatchSubPath, fmt.Sprintf("%d", node.NodeID), vChanName)
+	log.Debug("handle put event: try to save completed state", zap.String("key", k))
 
-	log.Debug("DataNode NewDataSyncService received Vchannel Info",
-		zap.Int64("collectionID", vchan.CollectionID),
-		zap.Int("Unflushed Segment Number", len(vchan.GetUnflushedSegments())),
-		zap.Int("Flushed Segment Number", len(vchan.GetFlushedSegments())),
-	)
-
-	flushCh := make(chan flushMsg, 100)
-
-	dataSyncService, err := newDataSyncService(node.ctx, flushCh, replica, alloc, node.msFactory, vchan, node.clearSignal, node.dataCoord, node.segmentCache, node.blobKv, node.compactionExecutor)
+	err = node.watchKv.Save(k, string(v))
+	// TODO DataNode unable to save into etcd, may need to panic
 	if err != nil {
-		log.Error("DataNode NewDataSyncService newDataSyncService failed",
-			zap.Error(err),
-		)
-		return err
+		node.releaseFlowgraph(vChanName)
+		return fmt.Errorf("fail to update completed state to etcd, vChanName: %s, err: %v", vChanName, err)
 	}
-
-	node.chanMut.Lock()
-	node.vchan2SyncService[vchan.GetChannelName()] = dataSyncService
-	node.vchan2FlushChs[vchan.GetChannelName()] = flushCh
-	node.chanMut.Unlock()
-
-	log.Info("DataNode NewDataSyncService success",
-		zap.Int64("Collection ID", vchan.GetCollectionID()),
-		zap.String("Vchannel name", vchan.GetChannelName()),
-	)
-	dataSyncService.start()
-
 	return nil
 }
 
+func (node *DataNode) handleDeleteEvent(vChanName string) {
+	node.releaseFlowgraph(vChanName)
+}
+
+func (node *DataNode) releaseFlowgraph(vChanName string) {
+	node.flowgraphManager.release(vChanName)
+}
+
 // BackGroundGC runs in background to release datanode resources
 func (node *DataNode) BackGroundGC(vChannelCh <-chan string) {
 	log.Info("DataNode Background GC Start")
 	for {
 		select {
-		case vChan := <-vChannelCh:
-			log.Info("GC flowgraph", zap.String("vChan", vChan))
-			node.ReleaseDataSyncService(vChan)
+		case vchanName := <-vChannelCh:
+			log.Info("GC flowgraph", zap.String("vChanName", vchanName))
+			node.releaseFlowgraph(vchanName)
 		case <-node.ctx.Done():
 			log.Info("DataNode ctx done")
 			return
@@ -413,23 +417,6 @@ func (node *DataNode) BackGroundGC(vChannelCh <-chan string) {
 	}
 }
 
-// ReleaseDataSyncService release flowgraph resources for a vchanName
-func (node *DataNode) ReleaseDataSyncService(vchanName string) {
-	log.Info("Release flowgraph resources begin", zap.String("Vchannel", vchanName))
-
-	node.chanMut.Lock()
-	dss, ok := node.vchan2SyncService[vchanName]
-	delete(node.vchan2SyncService, vchanName)
-	delete(node.vchan2FlushChs, vchanName)
-	node.chanMut.Unlock()
-
-	if ok {
-		// This is a time-consuming process, better to put outside of the lock
-		dss.close()
-	}
-	log.Debug("Release flowgraph resources end", zap.String("Vchannel", vchanName))
-}
-
 // FilterThreshold is the start time ouf DataNode
 var FilterThreshold Timestamp
 
@@ -541,51 +528,11 @@ func (node *DataNode) GetComponentStates(ctx context.Context) (*internalpb.Compo
 	return states, nil
 }
 
-func (node *DataNode) getChannelNamebySegmentID(segID UniqueID) string {
-	node.chanMut.RLock()
-	defer node.chanMut.RUnlock()
-	for name, dataSync := range node.vchan2SyncService {
-		if dataSync.replica.hasSegment(segID, true) {
-			return name
-		}
-	}
-	return ""
-}
-
-func (node *DataNode) getChannelNamesbyCollectionID(collID UniqueID) []string {
-	node.chanMut.RLock()
-	defer node.chanMut.RUnlock()
-
-	channels := make([]string, 0, len(node.vchan2SyncService))
-	for name, dataSync := range node.vchan2SyncService {
-		if dataSync.collectionID == collID {
-			channels = append(channels, name)
-		}
-	}
-	return channels
-}
-
 // ReadyToFlush tells wether DataNode is ready for flushing
 func (node *DataNode) ReadyToFlush() error {
 	if node.State.Load().(internalpb.StateCode) != internalpb.StateCode_Healthy {
 		return errors.New("DataNode not in HEALTHY state")
 	}
-
-	node.chanMut.RLock()
-	defer node.chanMut.RUnlock()
-	if len(node.vchan2SyncService) == 0 && len(node.vchan2FlushChs) == 0 {
-		// Healthy but Idle
-		msg := "DataNode HEALTHY but IDLE, please try WatchDmChannels to make it work"
-		log.Warn(msg)
-		return errors.New(msg)
-	}
-
-	if len(node.vchan2SyncService) != len(node.vchan2FlushChs) {
-		// TODO restart
-		msg := "DataNode HEALTHY but abnormal inside, restarting..."
-		log.Warn(msg)
-		return errors.New(msg)
-	}
 	return nil
 }
 
@@ -600,8 +547,8 @@ func (node *DataNode) FlushSegments(ctx context.Context, req *datapb.FlushSegmen
 		ErrorCode: commonpb.ErrorCode_UnexpectedError,
 	}
 
-	if err := node.ReadyToFlush(); err != nil {
-		status.Reason = err.Error()
+	if node.State.Load().(internalpb.StateCode) != internalpb.StateCode_Healthy {
+		status.Reason = "DataNode not in HEALTHY state"
 		return status, nil
 	}
 
@@ -613,16 +560,6 @@ func (node *DataNode) FlushSegments(ctx context.Context, req *datapb.FlushSegmen
 	processSegments := func(segmentIDs []UniqueID, flushed bool) bool {
 		noErr := true
 		for _, id := range segmentIDs {
-			chanName := node.getChannelNamebySegmentID(id)
-			if len(chanName) == 0 {
-				log.Warn("FlushSegments failed, cannot find segment in DataNode replica",
-					zap.Int64("collectionID", req.GetCollectionID()), zap.Int64("segmentID", id))
-
-				status.Reason = fmt.Sprintf("DataNode replica not find segment %d!", id)
-				noErr = false
-				continue
-			}
-
 			if node.segmentCache.checkIfCached(id) {
 				// Segment in flushing, ignore
 				log.Info("Segment flushing, ignore the flush request until flush is done.",
@@ -633,17 +570,15 @@ func (node *DataNode) FlushSegments(ctx context.Context, req *datapb.FlushSegmen
 
 			node.segmentCache.Cache(id)
 
-			node.chanMut.RLock()
-			flushChs, ok := node.vchan2FlushChs[chanName]
-			node.chanMut.RUnlock()
-			if !ok {
+			flushCh, err := node.flowgraphManager.getFlushCh(id)
+			if err != nil {
 				status.Reason = "DataNode abnormal, restarting"
-				log.Error("DataNode abnormal, no flushCh for a vchannel")
+				log.Error("DataNode abnormal, no flushCh for a vchannel", zap.Error(err))
 				noErr = false
 				continue
 			}
 
-			flushChs <- flushMsg{
+			flushCh <- flushMsg{
 				msgID:        req.Base.MsgID,
 				timestamp:    req.Base.Timestamp,
 				segmentID:    id,
@@ -677,15 +612,7 @@ func (node *DataNode) Stop() error {
 	node.UpdateStateCode(internalpb.StateCode_Abnormal)
 
 	node.cancel()
-
-	node.chanMut.RLock()
-	defer node.chanMut.RUnlock()
-	// close services
-	for _, syncService := range node.vchan2SyncService {
-		if syncService != nil {
-			(*syncService).close()
-		}
-	}
+	node.flowgraphManager.dropAll()
 
 	if node.closer != nil {
 		err := node.closer.Close()
@@ -796,7 +723,7 @@ func (node *DataNode) Compaction(ctx context.Context, req *datapb.CompactionPlan
 		ErrorCode: commonpb.ErrorCode_UnexpectedError,
 	}
 
-	ds, ok := node.vchan2SyncService[req.GetChannel()]
+	ds, ok := node.flowgraphManager.getFlowgraphService(req.GetChannel())
 	if !ok {
 		log.Warn("illegel compaction plan, channel not in this DataNode", zap.String("channel name", req.GetChannel()))
 		status.Reason = errIllegalCompactionPlan.Error()
diff --git a/internal/datanode/data_node_test.go b/internal/datanode/data_node_test.go
index debb45c3c3..7bd7bada42 100644
--- a/internal/datanode/data_node_test.go
+++ b/internal/datanode/data_node_test.go
@@ -29,11 +29,16 @@ import (
 	"time"
 
 	"github.com/milvus-io/milvus/internal/common"
-
+	etcdkv "github.com/milvus-io/milvus/internal/kv/etcd"
+	"github.com/milvus-io/milvus/internal/log"
 	"github.com/milvus-io/milvus/internal/msgstream"
-	"github.com/milvus-io/milvus/internal/proto/milvuspb"
 	"github.com/milvus-io/milvus/internal/types"
 
+	"github.com/milvus-io/milvus/internal/proto/commonpb"
+	"github.com/milvus-io/milvus/internal/proto/datapb"
+	"github.com/milvus-io/milvus/internal/proto/internalpb"
+	"github.com/milvus-io/milvus/internal/proto/milvuspb"
+
 	"github.com/milvus-io/milvus/internal/util/etcd"
 	"github.com/milvus-io/milvus/internal/util/metricsinfo"
 	"github.com/milvus-io/milvus/internal/util/sessionutil"
@@ -42,12 +47,6 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"go.uber.org/zap"
-
-	etcdkv "github.com/milvus-io/milvus/internal/kv/etcd"
-	"github.com/milvus-io/milvus/internal/log"
-	"github.com/milvus-io/milvus/internal/proto/commonpb"
-	"github.com/milvus-io/milvus/internal/proto/datapb"
-	"github.com/milvus-io/milvus/internal/proto/internalpb"
 )
 
 func TestMain(t *testing.M) {
@@ -62,6 +61,8 @@ func TestMain(t *testing.M) {
 
 func TestDataNode(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
 	node := newIDLEDataNodeMock(ctx)
 	etcdCli, err := etcd.GetEtcdClient(&Params.BaseParams)
 	assert.Nil(t, err)
@@ -132,39 +133,6 @@ func TestDataNode(t *testing.T) {
 		assert.Equal(t, commonpb.ErrorCode_Success, stat.Status.ErrorCode)
 	})
 
-	t.Run("Test NewDataSyncService", func(t *testing.T) {
-		t.Skip()
-		ctx, cancel := context.WithCancel(context.Background())
-		node2 := newIDLEDataNodeMock(ctx)
-		err = node2.Start()
-		assert.Nil(t, err)
-		dmChannelName := "fake-by-dev-rootcoord-dml-channel-test-NewDataSyncService"
-
-		vchan := &datapb.VchannelInfo{
-			CollectionID:      1,
-			ChannelName:       dmChannelName,
-			UnflushedSegments: []*datapb.SegmentInfo{},
-		}
-
-		require.Equal(t, 0, len(node2.vchan2FlushChs))
-		require.Equal(t, 0, len(node2.vchan2SyncService))
-
-		err := node2.NewDataSyncService(vchan)
-		assert.NoError(t, err)
-		assert.Equal(t, 1, len(node2.vchan2FlushChs))
-		assert.Equal(t, 1, len(node2.vchan2SyncService))
-
-		err = node2.NewDataSyncService(vchan)
-		assert.NoError(t, err)
-		assert.Equal(t, 1, len(node2.vchan2FlushChs))
-		assert.Equal(t, 1, len(node2.vchan2SyncService))
-
-		cancel()
-		<-node2.ctx.Done()
-		err = node2.Stop()
-		assert.Nil(t, err)
-	})
-
 	t.Run("Test FlushSegments", func(t *testing.T) {
 		dmChannelName := "fake-by-dev-rootcoord-dml-channel-test-FlushSegments"
 
@@ -185,12 +153,14 @@ func TestDataNode(t *testing.T) {
 			UnflushedSegments: []*datapb.SegmentInfo{},
 			FlushedSegments:   []*datapb.SegmentInfo{},
 		}
-		err := node1.NewDataSyncService(vchan)
-		assert.Nil(t, err)
 
-		service, ok := node1.vchan2SyncService[dmChannelName]
+		err := node1.flowgraphManager.addAndStart(node1, vchan)
+		require.Nil(t, err)
+
+		fgservice, ok := node1.flowgraphManager.getFlowgraphService(dmChannelName)
 		assert.True(t, ok)
-		err = service.replica.addNewSegment(0, 1, 1, dmChannelName, &internalpb.MsgPosition{}, &internalpb.MsgPosition{})
+
+		err = fgservice.replica.addNewSegment(0, 1, 1, dmChannelName, &internalpb.MsgPosition{}, &internalpb.MsgPosition{})
 		assert.Nil(t, err)
 
 		req := &datapb.FlushSegmentsRequest{
@@ -282,25 +252,6 @@ func TestDataNode(t *testing.T) {
 		status, err = node1.FlushSegments(node1.ctx, req)
 		assert.NoError(t, err)
 		assert.Equal(t, commonpb.ErrorCode_UnexpectedError, status.ErrorCode)
-
-		// manual inject meta error
-		node1.chanMut.Lock()
-		node1.vchan2FlushChs[dmChannelName+"1"] = node1.vchan2FlushChs[dmChannelName]
-		delete(node1.vchan2FlushChs, dmChannelName)
-		node1.chanMut.Unlock()
-		node1.segmentCache.Remove(0)
-
-		req = &datapb.FlushSegmentsRequest{
-			Base:         &commonpb.MsgBase{},
-			DbID:         0,
-			CollectionID: 1,
-			SegmentIDs:   []int64{0},
-		}
-
-		status, err = node1.FlushSegments(node1.ctx, req)
-		assert.NoError(t, err)
-		assert.Equal(t, commonpb.ErrorCode_UnexpectedError, status.ErrorCode)
-
 	})
 
 	t.Run("Test GetTimeTickChannel", func(t *testing.T) {
@@ -383,99 +334,13 @@ func TestDataNode(t *testing.T) {
 
 		for i, test := range testDataSyncs {
 			if i <= 2 {
-
-				err = node.NewDataSyncService(&datapb.VchannelInfo{CollectionID: 1, ChannelName: test.dmChannelName})
-
+				err = node.flowgraphManager.addAndStart(node, &datapb.VchannelInfo{CollectionID: 1, ChannelName: test.dmChannelName})
 				assert.Nil(t, err)
-
 				vchanNameCh <- test.dmChannelName
 			}
 		}
-
-		assert.Eventually(t, func() bool {
-			node.chanMut.Lock()
-			defer node.chanMut.Unlock()
-			return len(node.vchan2FlushChs) == 0
-		}, time.Second, time.Millisecond)
-
 		cancel()
 	})
-
-	t.Run("Test ReleaseDataSyncService", func(t *testing.T) {
-		dmChannelName := "fake-by-dev-rootcoord-dml-channel-test-NewDataSyncService"
-
-		vchan := &datapb.VchannelInfo{
-			CollectionID:      1,
-			ChannelName:       dmChannelName,
-			UnflushedSegments: []*datapb.SegmentInfo{},
-		}
-
-		err := node.NewDataSyncService(vchan)
-		require.NoError(t, err)
-		require.Equal(t, 1, len(node.vchan2FlushChs))
-		require.Equal(t, 1, len(node.vchan2SyncService))
-		time.Sleep(100 * time.Millisecond)
-
-		node.ReleaseDataSyncService(dmChannelName)
-		assert.Equal(t, 0, len(node.vchan2FlushChs))
-		assert.Equal(t, 0, len(node.vchan2SyncService))
-
-		s, ok := node.vchan2SyncService[dmChannelName]
-		assert.False(t, ok)
-		assert.Nil(t, s)
-	})
-
-	t.Run("Test GetChannelName", func(t *testing.T) {
-		ctx, cancel := context.WithCancel(context.Background())
-		node := newIDLEDataNodeMock(ctx)
-
-		testCollIDs := []UniqueID{0, 1, 2, 1}
-		testSegIDs := []UniqueID{10, 11, 12, 13}
-		testchanNames := []string{"a", "b", "c", "d"}
-
-		node.chanMut.Lock()
-		for i, name := range testchanNames {
-			replica := &SegmentReplica{
-				collectionID: testCollIDs[i],
-				newSegments:  make(map[UniqueID]*Segment),
-			}
-
-			err = replica.addNewSegment(testSegIDs[i], testCollIDs[i], 0, name, &internalpb.MsgPosition{}, nil)
-			assert.Nil(t, err)
-			node.vchan2SyncService[name] = &dataSyncService{collectionID: testCollIDs[i], replica: replica}
-		}
-		node.chanMut.Unlock()
-
-		type Test struct {
-			inCollID         UniqueID
-			expectedChannels []string
-
-			inSegID         UniqueID
-			expectedChannel string
-		}
-		tests := []Test{
-			{0, []string{"a"}, 10, "a"},
-			{1, []string{"b", "d"}, 11, "b"},
-			{2, []string{"c"}, 12, "c"},
-			{3, []string{}, 13, "d"},
-			{3, []string{}, 100, ""},
-		}
-
-		for _, test := range tests {
-			actualChannels := node.getChannelNamesbyCollectionID(test.inCollID)
-			assert.ElementsMatch(t, test.expectedChannels, actualChannels)
-
-			actualChannel := node.getChannelNamebySegmentID(test.inSegID)
-			assert.Equal(t, test.expectedChannel, actualChannel)
-		}
-
-		cancel()
-	})
-
-	cancel()
-	<-node.ctx.Done()
-	err = node.Stop()
-	require.Nil(t, err)
 }
 
 func TestWatchChannel(t *testing.T) {
@@ -495,6 +360,7 @@ func TestWatchChannel(t *testing.T) {
 	defer cancel()
 
 	t.Run("test watch channel", func(t *testing.T) {
+		// GOOSE TODO
 		kv := etcdkv.NewEtcdKV(etcdCli, Params.BaseParams.MetaRootPath)
 		oldInvalidCh := "datanode-etcd-test-by-dev-rootcoord-dml-channel-invalid"
 		path := fmt.Sprintf("%s/%d/%s", Params.DataNodeCfg.ChannelWatchSubPath, node.NodeID, oldInvalidCh)
@@ -540,29 +406,27 @@ func TestWatchChannel(t *testing.T) {
 
 		// wait for check goroutine received 2 events
 		<-c
-		node.chanMut.RLock()
-		_, has := node.vchan2SyncService[ch]
-		node.chanMut.RUnlock()
-		assert.True(t, has)
+		exist := node.flowgraphManager.exist(ch)
+		assert.True(t, exist)
 
 		err = kv.RemoveWithPrefix(fmt.Sprintf("%s/%d", Params.DataNodeCfg.ChannelWatchSubPath, node.NodeID))
 		assert.Nil(t, err)
 		//TODO there is not way to sync Release done, use sleep for now
 		time.Sleep(100 * time.Millisecond)
 
-		node.chanMut.RLock()
-		_, has = node.vchan2SyncService[ch]
-		node.chanMut.RUnlock()
-		assert.False(t, has)
+		exist = node.flowgraphManager.exist(ch)
+		assert.False(t, exist)
 	})
 
 	t.Run("handle watch info failed", func(t *testing.T) {
-		node.handleWatchInfo("test1", []byte{23})
+		e := &event{
+			eventType: putEventType,
+		}
 
-		node.chanMut.RLock()
-		_, has := node.vchan2SyncService["test1"]
-		assert.False(t, has)
-		node.chanMut.RUnlock()
+		node.handleWatchInfo(e, "test1", []byte{23})
+
+		exist := node.flowgraphManager.exist("test1")
+		assert.False(t, exist)
 
 		info := datapb.ChannelWatchInfo{
 			Vchan: nil,
@@ -570,12 +434,10 @@ func TestWatchChannel(t *testing.T) {
 		}
 		bs, err := proto.Marshal(&info)
 		assert.NoError(t, err)
-		node.handleWatchInfo("test2", bs)
+		node.handleWatchInfo(e, "test2", bs)
 
-		node.chanMut.RLock()
-		_, has = node.vchan2SyncService["test2"]
-		assert.False(t, has)
-		node.chanMut.RUnlock()
+		exist = node.flowgraphManager.exist("test2")
+		assert.False(t, exist)
 
 		info = datapb.ChannelWatchInfo{
 			Vchan: &datapb.VchannelInfo{},
@@ -587,12 +449,9 @@ func TestWatchChannel(t *testing.T) {
 		node.msFactory = &FailMessageStreamFactory{
 			node.msFactory,
 		}
-		node.handleWatchInfo("test3", bs)
-		node.chanMut.RLock()
-		_, has = node.vchan2SyncService["test3"]
-		assert.False(t, has)
-		node.chanMut.RUnlock()
-
+		node.handleWatchInfo(e, "test3", bs)
+		exist = node.flowgraphManager.exist("test3")
+		assert.False(t, exist)
 	})
 }
 
diff --git a/internal/datanode/event_manager.go b/internal/datanode/event_manager.go
new file mode 100644
index 0000000000..abdfc6aa5c
--- /dev/null
+++ b/internal/datanode/event_manager.go
@@ -0,0 +1,101 @@
+// Licensed to the LF AI & Data foundation under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package datanode
+
+import (
+	"time"
+
+	"github.com/milvus-io/milvus/internal/log"
+	"github.com/milvus-io/milvus/internal/proto/datapb"
+
+	"go.uber.org/zap"
+)
+
+const retryWatchInterval = 20 * time.Second
+
+type event struct {
+	eventType int
+	vChanName string
+	info      *datapb.ChannelWatchInfo
+}
+
+type channelEventManager struct {
+	eventChan         chan event
+	closeChan         chan struct{}
+	handlePutEvent    func(watchInfo *datapb.ChannelWatchInfo) error // node.handlePutEvent
+	handleDeleteEvent func(vChanName string)                         // node.handleDeleteEvent
+}
+
+const (
+	putEventType    = 1
+	deleteEventType = 2
+)
+
+func (e *channelEventManager) Run() {
+	go func() {
+		for {
+			select {
+			case event := <-e.eventChan:
+				switch event.eventType {
+				case putEventType:
+					// Trigger retry for-loop when fail to handle put event for the first time
+					if err := e.handlePutEvent(event.info); err != nil {
+						for {
+							log.Warn("handle put event fail, starting retry",
+								zap.String("vChanName", event.vChanName),
+								zap.String("retry interval", retryWatchInterval.String()),
+								zap.Error(err))
+
+							<-time.NewTimer(time.Second).C
+
+							select {
+							case e, ok := <-e.eventChan:
+								// When getting a delete event at next retry, exit retry loop
+								// When getting a put event, just continue the retry
+								if ok && e.eventType == deleteEventType {
+									log.Warn("delete event triggerred, terminating retry.",
+										zap.String("vChanName", event.vChanName))
+									return
+								}
+							default:
+							}
+
+							err = e.handlePutEvent(event.info)
+							if err == nil {
+								log.Debug("retry to handle put event successfully",
+									zap.String("vChanName", event.vChanName))
+								return
+							}
+						}
+					}
+				case deleteEventType:
+					e.handleDeleteEvent(event.vChanName)
+				}
+			case <-e.closeChan:
+				return
+			}
+		}
+	}()
+}
+
+func (e *channelEventManager) handleEvent(event event) {
+	e.eventChan <- event
+}
+
+func (e *channelEventManager) Close() {
+	close(e.closeChan)
+}
diff --git a/internal/datanode/flow_graph_manager.go b/internal/datanode/flow_graph_manager.go
new file mode 100644
index 0000000000..b27bdf978b
--- /dev/null
+++ b/internal/datanode/flow_graph_manager.go
@@ -0,0 +1,127 @@
+// Licensed to the LF AI & Data foundation under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package datanode
+
+import (
+	"fmt"
+	"sync"
+
+	"github.com/milvus-io/milvus/internal/log"
+	"github.com/milvus-io/milvus/internal/proto/datapb"
+
+	"go.uber.org/zap"
+)
+
+type flowgraphManager struct {
+	flowgraphs sync.Map // vChannelName -> dataSyncService
+}
+
+func newFlowgraphManager() *flowgraphManager {
+	return &flowgraphManager{}
+}
+
+func (fm *flowgraphManager) addAndStart(dn *DataNode, vchan *datapb.VchannelInfo) error {
+	log.Debug("received Vchannel Info",
+		zap.String("vChannelName", vchan.GetChannelName()),
+		zap.Int("Unflushed Segment Number", len(vchan.GetUnflushedSegments())),
+		zap.Int("Flushed Segment Number", len(vchan.GetFlushedSegments())),
+	)
+
+	if _, ok := fm.flowgraphs.Load(vchan.GetChannelName()); ok {
+		log.Warn("try to add an existed DataSyncService", zap.String("vChannelName", vchan.GetChannelName()))
+		return nil
+	}
+
+	replica, err := newReplica(dn.ctx, dn.rootCoord, vchan.GetCollectionID())
+	if err != nil {
+		log.Warn("new replica failed", zap.String("vChannelName", vchan.GetChannelName()), zap.Error(err))
+		return err
+	}
+
+	var alloc allocatorInterface = newAllocator(dn.rootCoord)
+
+	dataSyncService, err := newDataSyncService(dn.ctx, make(chan flushMsg, 100), replica, alloc, dn.msFactory, vchan, dn.clearSignal, dn.dataCoord, dn.segmentCache, dn.blobKv, dn.compactionExecutor)
+	if err != nil {
+		log.Warn("new data sync service fail", zap.String("vChannelName", vchan.GetChannelName()), zap.Error(err))
+		return err
+	}
+	log.Info("successfully created dataSyncService", zap.String("vChannelName", vchan.GetChannelName()))
+
+	dataSyncService.start()
+	log.Info("successfully started dataSyncService", zap.String("vChannelName", vchan.GetChannelName()))
+
+	fm.flowgraphs.Store(vchan.GetChannelName(), dataSyncService)
+	return nil
+}
+
+func (fm *flowgraphManager) release(vchanName string) {
+	log.Debug("release flowgraph resources begin", zap.String("vChannelName", vchanName))
+
+	if fg, loaded := fm.flowgraphs.LoadAndDelete(vchanName); loaded {
+		fg.(*dataSyncService).close()
+	}
+	log.Debug("release flowgraph resources end", zap.String("Vchannel", vchanName))
+}
+
+func (fm *flowgraphManager) getFlushCh(segID UniqueID) (chan<- flushMsg, error) {
+	var (
+		flushCh chan flushMsg
+		loaded  = false
+	)
+
+	fm.flowgraphs.Range(func(key, value interface{}) bool {
+		fg := value.(*dataSyncService)
+		if fg.replica.hasSegment(segID, true) {
+			loaded = true
+			flushCh = fg.flushCh
+			return false
+		}
+		return true
+	})
+
+	if loaded {
+		return flushCh, nil
+	}
+
+	return nil, fmt.Errorf("cannot find segment %d in all flowgraphs", segID)
+}
+
+func (fm *flowgraphManager) getFlowgraphService(vchan string) (*dataSyncService, bool) {
+	fg, ok := fm.flowgraphs.Load(vchan)
+	if ok {
+		return fg.(*dataSyncService), ok
+	}
+
+	return nil, ok
+}
+
+func (fm *flowgraphManager) exist(vchan string) bool {
+	_, exist := fm.getFlowgraphService(vchan)
+	return exist
+}
+
+func (fm *flowgraphManager) dropAll() {
+	log.Debug("start drop all flowgraph resources in DataNode")
+	fm.flowgraphs.Range(func(key, value interface{}) bool {
+		value.(*dataSyncService).close()
+		fm.flowgraphs.Delete(key.(string))
+
+		log.Debug("successfully dropped flowgraph", zap.String("vChannelName", key.(string)))
+		return true
+	})
+	log.Debug("end drop all flowgraph resources in DataNode")
+}
diff --git a/internal/datanode/flow_graph_manager_test.go b/internal/datanode/flow_graph_manager_test.go
new file mode 100644
index 0000000000..fd658ee737
--- /dev/null
+++ b/internal/datanode/flow_graph_manager_test.go
@@ -0,0 +1,128 @@
+// Licensed to the LF AI & Data foundation under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package datanode
+
+import (
+	"context"
+	"testing"
+
+	"github.com/milvus-io/milvus/internal/proto/datapb"
+	"github.com/milvus-io/milvus/internal/proto/internalpb"
+	"github.com/milvus-io/milvus/internal/util/etcd"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestFlowGraphManager(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.TODO())
+	defer cancel()
+
+	etcdCli, err := etcd.GetEtcdClient(&Params.BaseParams)
+	assert.Nil(t, err)
+	defer etcdCli.Close()
+
+	node := newIDLEDataNodeMock(ctx)
+	node.SetEtcdClient(etcdCli)
+	err = node.Init()
+	require.Nil(t, err)
+	err = node.Start()
+	require.Nil(t, err)
+
+	fm := newFlowgraphManager()
+	defer fm.dropAll()
+	t.Run("Test addAndStart", func(t *testing.T) {
+		vchanName := "by-dev-rootcoord-dml-test-flowgraphmanager-addAndStart"
+		vchan := &datapb.VchannelInfo{
+			CollectionID: 1,
+			ChannelName:  vchanName,
+		}
+		require.False(t, fm.exist(vchanName))
+
+		err := fm.addAndStart(node, vchan)
+		assert.NoError(t, err)
+		assert.True(t, fm.exist(vchanName))
+
+		fm.dropAll()
+	})
+
+	t.Run("Test Release", func(t *testing.T) {
+		vchanName := "by-dev-rootcoord-dml-test-flowgraphmanager-Release"
+		vchan := &datapb.VchannelInfo{
+			CollectionID: 1,
+			ChannelName:  vchanName,
+		}
+		require.False(t, fm.exist(vchanName))
+
+		err := fm.addAndStart(node, vchan)
+		assert.NoError(t, err)
+		assert.True(t, fm.exist(vchanName))
+
+		fm.release(vchanName)
+
+		assert.False(t, fm.exist(vchanName))
+		fm.dropAll()
+	})
+
+	t.Run("Test getFlushCh", func(t *testing.T) {
+		vchanName := "by-dev-rootcoord-dml-test-flowgraphmanager-getFlushCh"
+		vchan := &datapb.VchannelInfo{
+			CollectionID: 1,
+			ChannelName:  vchanName,
+		}
+		require.False(t, fm.exist(vchanName))
+
+		err := fm.addAndStart(node, vchan)
+		assert.NoError(t, err)
+		assert.True(t, fm.exist(vchanName))
+
+		fg, ok := fm.getFlowgraphService(vchanName)
+		require.True(t, ok)
+		err = fg.replica.addNewSegment(100, 1, 10, vchanName, &internalpb.MsgPosition{}, &internalpb.MsgPosition{})
+		require.NoError(t, err)
+
+		tests := []struct {
+			isvalid bool
+			inSegID UniqueID
+
+			description string
+		}{
+			{true, 100, "valid input for existed segmentID 100"},
+			{false, 101, "invalid input for not existed segmentID 101"},
+		}
+
+		for _, test := range tests {
+			t.Run(test.description, func(t *testing.T) {
+				ch, err := fm.getFlushCh(test.inSegID)
+
+				if test.isvalid {
+					assert.NoError(t, err)
+					assert.NotNil(t, ch)
+				} else {
+					assert.Error(t, err)
+					assert.Nil(t, ch)
+				}
+			})
+		}
+	})
+
+	t.Run("Test getFlowgraphService", func(t *testing.T) {
+		fg, ok := fm.getFlowgraphService("channel-not-exist")
+		assert.False(t, ok)
+		assert.Nil(t, fg)
+	})
+}