fallback to follower when leader is busy (tikv#916)

* fallback to follower when leader is busy Signed-off-by: you06 <[email protected]> Co-authored-by: cfzjywxk <[email protected]>
you06 · Aug 7, 2023 · cf84acf · cf84acf
1 parent a47e1c2
commit cf84acf
Show file tree

Hide file tree

Showing 3 changed files with 196 additions and 24 deletions.
diff --git a/internal/locate/region_cache.go b/internal/locate/region_cache.go
@@ -1289,9 +1289,11 @@ func (c *RegionCache) reloadRegion(regionID uint64) {
 		// ignore error and use old region info.
 		logutil.Logger(bo.GetCtx()).Error("load region failure",
 			zap.Uint64("regionID", regionID), zap.Error(err))
+		c.mu.RLock()
 		if oldRegion := c.getRegionByIDFromCache(regionID); oldRegion != nil {
 			oldRegion.asyncReload.Store(false)
 		}
+		c.mu.RUnlock()
 		return
 	}
 	c.mu.Lock()

diff --git a/internal/locate/region_request.go b/internal/locate/region_request.go
@@ -389,38 +389,77 @@ func (state *accessKnownLeader) onNoLeader(selector *replicaSelector) {
 // the leader will be updated to replicas[0] and give it another chance.
 type tryFollower struct {
 	stateBase
-	leaderIdx AccessIndex
-	lastIdx   AccessIndex
+	// if the leader is unavailable, but it still holds the leadership, fallbackFromLeader is true and replica read is enabled.
+	fallbackFromLeader bool
+	leaderIdx          AccessIndex
+	lastIdx            AccessIndex
+	labels             []*metapb.StoreLabel
 }
 
 func (state *tryFollower) next(bo *retry.Backoffer, selector *replicaSelector) (*RPCContext, error) {
-	var targetReplica *replica
-	// Search replica that is not attempted from the last accessed replica
-	for i := 1; i < len(selector.replicas); i++ {
-		idx := AccessIndex((int(state.lastIdx) + i) % len(selector.replicas))
-		if idx == state.leaderIdx {
-			continue
+	filterReplicas := func(fn func(*replica) bool) (AccessIndex, *replica) {
+		for i := 0; i < len(selector.replicas); i++ {
+			idx := AccessIndex((int(state.lastIdx) + i) % len(selector.replicas))
+			if idx == state.leaderIdx {
+				continue
+			}
+			selectReplica := selector.replicas[idx]
+			if fn(selectReplica) && selectReplica.store.getLivenessState() != unreachable {
+				return idx, selectReplica
+			}
 		}
-		targetReplica = selector.replicas[idx]
-		// Each follower is only tried once
-		if !targetReplica.isExhausted(1) && targetReplica.store.getLivenessState() != unreachable {
+		return -1, nil
+	}
+
+	if len(state.labels) > 0 {
+		idx, selectReplica := filterReplicas(func(selectReplica *replica) bool {
+			return selectReplica.store.IsLabelsMatch(state.labels)
+		})
+		if selectReplica != nil && idx >= 0 {
+			state.lastIdx = idx
+			selector.targetIdx = idx
+		}
+		// labels only take effect for first try.
+		state.labels = nil
+	}
+	if selector.targetIdx < 0 {
+		// Search replica that is not attempted from the last accessed replica
+		idx, selectReplica := filterReplicas(func(selectReplica *replica) bool {
+			return !selectReplica.isExhausted(1)
+		})
+		if selectReplica != nil && idx >= 0 {
 			state.lastIdx = idx
 			selector.targetIdx = idx
-			break
 		}
 	}
+
 	// If all followers are tried and fail, backoff and retry.
 	if selector.targetIdx < 0 {
 		metrics.TiKVReplicaSelectorFailureCounter.WithLabelValues("exhausted").Inc()
 		selector.invalidateRegion()
 		return nil, nil
 	}
-	return selector.buildRPCContext(bo)
+	rpcCtx, err := selector.buildRPCContext(bo)
+	if err != nil || rpcCtx == nil {
+		return rpcCtx, err
+	}
+	if state.fallbackFromLeader {
+		staleRead := false
+		rpcCtx.contextPatcher.staleRead = &staleRead
+		replicaRead := true
+		rpcCtx.contextPatcher.replicaRead = &replicaRead
+	}
+	return rpcCtx, nil
 }
 
 func (state *tryFollower) onSendSuccess(selector *replicaSelector) {
-	if !selector.region.switchWorkLeaderToPeer(selector.targetReplica().peer) {
-		panic("the store must exist")
+	if !state.fallbackFromLeader {
+		peer := selector.targetReplica().peer
+		if !selector.region.switchWorkLeaderToPeer(peer) {
+			logutil.BgLogger().Warn("the store must exist",
+				zap.Uint64("store", peer.StoreId),
+				zap.Uint64("peer", peer.Id))
+		}
 	}
 }
 
@@ -542,6 +581,10 @@ type accessFollower struct {
 	learnerOnly bool
 }
 
+// Follower read will try followers first, if no follower is available, it will fallback to leader.
+// Specially, for stale read, it tries local peer(can be either leader or follower), then use snapshot read in the leader,
+// if the leader read receive server-is-busy and connection errors, the region cache is still valid,
+// and the state will be changed to tryFollower, which will read by replica read.
 func (state *accessFollower) next(bo *retry.Backoffer, selector *replicaSelector) (*RPCContext, error) {
 	replicaSize := len(selector.replicas)
 	resetStaleRead := false
@@ -610,14 +653,30 @@ func (state *accessFollower) next(bo *retry.Backoffer, selector *replicaSelector
 	// If there is no candidate, fallback to the leader.
 	if selector.targetIdx < 0 {
 		leader := selector.replicas[state.leaderIdx]
-		leaderInvalid := leader.isEpochStale() || state.IsLeaderExhausted(leader)
+		leaderEpochStale := leader.isEpochStale()
+		leaderInvalid := leaderEpochStale || state.IsLeaderExhausted(leader)
 		if len(state.option.labels) > 0 {
 			logutil.Logger(bo.GetCtx()).Warn("unable to find stores with given labels",
 				zap.Uint64("region", selector.region.GetID()),
 				zap.Bool("leader-invalid", leaderInvalid),
 				zap.Any("labels", state.option.labels))
 		}
 		if leaderInvalid {
+			// In stale-read, the request will fallback to leader after the local follower failure.
+			// If the leader is also unavailable, we can fallback to the follower and use replica-read flag again,
+			// The remote follower not tried yet, and the local follower can retry without stale-read flag.
+			if state.isStaleRead {
+				selector.state = &tryFollower{
+					fallbackFromLeader: true,
+					leaderIdx:          state.leaderIdx,
+					lastIdx:            state.leaderIdx,
+					labels:             state.option.labels,
+				}
+				if leaderEpochStale {
+					selector.regionCache.scheduleReloadRegion(selector.region)
+				}
+				return nil, stateChanged{}
+			}
 			metrics.TiKVReplicaSelectorFailureCounter.WithLabelValues("exhausted").Inc()
 			selector.invalidateRegion()
 			return nil, nil
@@ -668,22 +727,23 @@ func (state *accessFollower) isCandidate(idx AccessIndex, replica *replica) bool
 	if replica.isEpochStale() || replica.isExhausted(1) || replica.store.getLivenessState() == unreachable {
 		return false
 	}
-	// The request can only be sent to the leader.
-	if state.option.leaderOnly && idx == state.leaderIdx {
-		return true
+	if state.option.leaderOnly {
+		// The request can only be sent to the leader.
+		return idx == state.leaderIdx
+	} else if !state.tryLeader && idx == state.leaderIdx {
+		// The request cannot be sent to leader.
+		return false
 	}
-	// Choose a replica with matched labels.
-	followerCandidate := !state.option.leaderOnly && (state.tryLeader || idx != state.leaderIdx) &&
-		replica.store.IsLabelsMatch(state.option.labels) && (!state.learnerOnly || replica.peer.Role == metapb.PeerRole_Learner)
-	if !followerCandidate {
+	if state.learnerOnly && replica.peer.Role == metapb.PeerRole_Learner {
+		// The request can only be sent to the learner.
 		return false
 	}
 	// And If the leader store is abnormal to be accessed under `ReplicaReadPreferLeader` mode, we should choose other valid followers
 	// as candidates to serve the Read request.
 	if state.option.preferLeader && replica.store.isSlow() {
 		return false
 	}
-	// If the stores are limited, check if the store is in the list.
+	// Choose a replica with matched labels.
 	return replica.store.IsStoreMatch(state.option.stores)
 }
 
@@ -1064,6 +1124,9 @@ func (s *replicaSelector) onServerIsBusy(
 		// Mark the server is busy (the next incoming READs could be redirect
 		// to expected followers. )
 		ctx.Store.markAlreadySlow()
+		if s.canFallback2Follower() {
+			return true, nil
+		}
 	}
 	err = bo.Backoff(retry.BoTiKVServerBusy, errors.Errorf("server is busy, ctx: %v", ctx))
 	if err != nil {
@@ -1072,6 +1135,23 @@ func (s *replicaSelector) onServerIsBusy(
 	return true, nil
 }
 
+// For some reasons, the leader is unreachable by now, try followers instead.
+// the state is changed in accessFollower.next when leader is unavailable.
+func (s *replicaSelector) canFallback2Follower() bool {
+	if s == nil || s.state == nil {
+		return false
+	}
+	state, ok := s.state.(*accessFollower)
+	if !ok {
+		return false
+	}
+	if !state.isStaleRead {
+		return false
+	}
+	// can fallback to follower only when the leader is exhausted.
+	return state.lastIdx == state.leaderIdx && state.IsLeaderExhausted(s.replicas[state.leaderIdx])
+}
+
 func (s *replicaSelector) invalidateRegion() {
 	if s.region != nil {
 		s.region.invalidate(Other)

diff --git a/internal/locate/region_request3_test.go b/internal/locate/region_request3_test.go
@@ -1199,3 +1199,93 @@ func (s *testRegionRequestToThreeStoresSuite) TestStaleReadFallback() {
 	s.True(ok)
 	s.Equal(getResp.Value, value)
 }
+
+func (s *testRegionRequestToThreeStoresSuite) TestStaleReadFallback2Follower() {
+	leaderStore, _ := s.loadAndGetLeaderStore()
+	leaderLabel := []*metapb.StoreLabel{
+		{
+			Key:   "id",
+			Value: strconv.FormatUint(leaderStore.StoreID(), 10),
+		},
+	}
+	var followerID *uint64
+	for _, storeID := range s.storeIDs {
+		if storeID != leaderStore.storeID {
+			followerID = &storeID
+			break
+		}
+	}
+	s.NotNil(followerID)
+	followerLabel := []*metapb.StoreLabel{
+		{
+			Key:   "id",
+			Value: strconv.FormatUint(*followerID, 10),
+		},
+	}
+
+	regionLoc, err := s.cache.LocateRegionByID(s.bo, s.regionID)
+	s.Nil(err)
+	s.NotNil(regionLoc)
+
+	dataIsNotReady := false
+	s.regionRequestSender.client = &fnClient{fn: func(ctx context.Context, addr string, req *tikvrpc.Request, timeout time.Duration) (response *tikvrpc.Response, err error) {
+		select {
+		case <-ctx.Done():
+			return nil, errors.New("timeout")
+		default:
+		}
+		if dataIsNotReady && req.StaleRead {
+			dataIsNotReady = false
+			return &tikvrpc.Response{Resp: &kvrpcpb.GetResponse{RegionError: &errorpb.Error{
+				DataIsNotReady: &errorpb.DataIsNotReady{},
+			}}}, nil
+		}
+		if addr == leaderStore.addr {
+			return &tikvrpc.Response{Resp: &kvrpcpb.GetResponse{RegionError: &errorpb.Error{
+				ServerIsBusy: &errorpb.ServerIsBusy{},
+			}}}, nil
+		}
+		if !req.ReplicaRead {
+			return &tikvrpc.Response{Resp: &kvrpcpb.GetResponse{RegionError: &errorpb.Error{
+				NotLeader: &errorpb.NotLeader{},
+			}}}, nil
+		}
+		return &tikvrpc.Response{Resp: &kvrpcpb.GetResponse{Value: []byte(addr)}}, nil
+	}}
+
+	for _, localLeader := range []bool{true, false} {
+		dataIsNotReady = true
+		// data is not ready, then server is busy in the first round,
+		// directly server is busy in the second round.
+		for i := 0; i < 2; i++ {
+			req := tikvrpc.NewReplicaReadRequest(tikvrpc.CmdGet, &kvrpcpb.GetRequest{Key: []byte("key")}, kv.ReplicaReadLeader, nil)
+			req.ReadReplicaScope = oracle.GlobalTxnScope
+			req.TxnScope = oracle.GlobalTxnScope
+			req.EnableStaleRead()
+			req.ReplicaReadType = kv.ReplicaReadMixed
+			var ops []StoreSelectorOption
+			if localLeader {
+				ops = append(ops, WithMatchLabels(leaderLabel))
+			} else {
+				ops = append(ops, WithMatchLabels(followerLabel))
+			}
+
+			ctx, _ := context.WithTimeout(context.Background(), 10000*time.Second)
+			bo := retry.NewBackoffer(ctx, -1)
+			s.Nil(err)
+			resp, _, _, err := s.regionRequestSender.SendReqCtx(bo, req, regionLoc.Region, time.Second, tikvrpc.TiKV, ops...)
+			s.Nil(err)
+
+			regionErr, err := resp.GetRegionError()
+			s.Nil(err)
+			s.Nil(regionErr)
+			getResp, ok := resp.Resp.(*kvrpcpb.GetResponse)
+			s.True(ok)
+			if localLeader {
+				s.NotEqual(getResp.Value, []byte("store"+leaderLabel[0].Value))
+			} else {
+				s.Equal(getResp.Value, []byte("store"+followerLabel[0].Value))
+			}
+		}
+	}
+}