support linearizable renew lease for 3.4

Cherry pick etcd-io#13932 to 3.4.

When etcdserver receives a LeaseRenew request, it may be still in
progress of processing the LeaseGrantRequest on exact the same
leaseID. Accordingly it may return a TTL=0 to client due to the
leaseID not found error. So the leader should wait for the appliedID
to be available before processing client requests.

Signed-off-by: Benjamin Wang <wachao@vmware.com>

etcdserver/api/v3rpc/rpctypes/error.go

@@ -74,6 +74,7 @@ var (
 	ErrGRPCTimeout                    = status.New(codes.Unavailable, "etcdserver: request timed out").Err()
 	ErrGRPCTimeoutDueToLeaderFail     = status.New(codes.Unavailable, "etcdserver: request timed out, possibly due to previous leader failure").Err()
 	ErrGRPCTimeoutDueToConnectionLost = status.New(codes.Unavailable, "etcdserver: request timed out, possibly due to connection lost").Err()
+	ErrGRPCTimeoutWaitAppliedIndex    = status.New(codes.Unavailable, "etcdserver: request timed out, waiting for the applied index took too long").Err()
 	ErrGRPCUnhealthy                  = status.New(codes.Unavailable, "etcdserver: unhealthy cluster").Err()
 	ErrGRPCCorrupt                    = status.New(codes.DataLoss, "etcdserver: corrupt cluster").Err()
 	ErrGPRCNotSupportedForLearner     = status.New(codes.Unavailable, "etcdserver: rpc not supported for learner").Err()
@@ -132,6 +133,7 @@ var (
 		ErrorDesc(ErrGRPCTimeout):                    ErrGRPCTimeout,
 		ErrorDesc(ErrGRPCTimeoutDueToLeaderFail):     ErrGRPCTimeoutDueToLeaderFail,
 		ErrorDesc(ErrGRPCTimeoutDueToConnectionLost): ErrGRPCTimeoutDueToConnectionLost,
+		ErrorDesc(ErrGRPCTimeoutWaitAppliedIndex):    ErrGRPCTimeoutWaitAppliedIndex,
 		ErrorDesc(ErrGRPCUnhealthy):                  ErrGRPCUnhealthy,
 		ErrorDesc(ErrGRPCCorrupt):                    ErrGRPCCorrupt,
 		ErrorDesc(ErrGPRCNotSupportedForLearner):     ErrGPRCNotSupportedForLearner,
@@ -192,6 +194,7 @@ var (
 	ErrTimeout                    = Error(ErrGRPCTimeout)
 	ErrTimeoutDueToLeaderFail     = Error(ErrGRPCTimeoutDueToLeaderFail)
 	ErrTimeoutDueToConnectionLost = Error(ErrGRPCTimeoutDueToConnectionLost)
+	ErrTimeoutWaitAppliedIndex    = Error(ErrGRPCTimeoutWaitAppliedIndex)
 	ErrUnhealthy                  = Error(ErrGRPCUnhealthy)
 	ErrCorrupt                    = Error(ErrGRPCCorrupt)
 	ErrBadLeaderTransferee        = Error(ErrGRPCBadLeaderTransferee)

etcdserver/api/v3rpc/util.go

@@ -53,6 +53,7 @@ var toGRPCErrorMap = map[error]error{
 	etcdserver.ErrTimeout:                    rpctypes.ErrGRPCTimeout,
 	etcdserver.ErrTimeoutDueToLeaderFail:     rpctypes.ErrGRPCTimeoutDueToLeaderFail,
 	etcdserver.ErrTimeoutDueToConnectionLost: rpctypes.ErrGRPCTimeoutDueToConnectionLost,
+	etcdserver.ErrTimeoutWaitAppliedIndex:    rpctypes.ErrGRPCTimeoutWaitAppliedIndex,
 	etcdserver.ErrUnhealthy:                  rpctypes.ErrGRPCUnhealthy,
 	etcdserver.ErrKeyNotFound:                rpctypes.ErrGRPCKeyNotFound,
 	etcdserver.ErrCorrupt:                    rpctypes.ErrGRPCCorrupt,

etcdserver/errors.go

@@ -26,6 +26,7 @@ var (
 	ErrTimeout                    = errors.New("etcdserver: request timed out")
 	ErrTimeoutDueToLeaderFail     = errors.New("etcdserver: request timed out, possibly due to previous leader failure")
 	ErrTimeoutDueToConnectionLost = errors.New("etcdserver: request timed out, possibly due to connection lost")
+	ErrTimeoutWaitAppliedIndex    = errors.New("etcdserver: request timed out, waiting for the applied index took too long")
 	ErrTimeoutLeaderTransfer      = errors.New("etcdserver: request timed out, leader transfer took too long")
 	ErrLeaderChanged              = errors.New("etcdserver: leader changed")
 	ErrNotEnoughStartedMembers    = errors.New("etcdserver: re-configuration failed due to not enough started members")

etcdserver/server_test.go

@@ -1850,3 +1850,59 @@ func (s *sendMsgAppRespTransporter) Send(m []raftpb.Message) {
 	}
 	s.sendC <- send
 }
+
+func TestWaitAppliedIndex(t *testing.T) {
+	cases := []struct {
+		name           string
+		appliedIndex   uint64
+		committedIndex uint64
+		action         func(s *EtcdServer)
+		ExpectedError  error
+	}{
+		{
+			name:           "The applied Id is already equal to the commitId",
+			appliedIndex:   10,
+			committedIndex: 10,
+			action: func(s *EtcdServer) {
+				s.applyWait.Trigger(10)
+			},
+			ExpectedError: nil,
+		},
+		{
+			name:           "The etcd server has already stopped",
+			appliedIndex:   10,
+			committedIndex: 12,
+			action: func(s *EtcdServer) {
+				s.stopping <- struct{}{}
+			},
+			ExpectedError: ErrStopped,
+		},
+		{
+			name:           "Timed out waiting for the applied index",
+			appliedIndex:   10,
+			committedIndex: 12,
+			action:         nil,
+			ExpectedError:  ErrTimeoutWaitAppliedIndex,
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			s := &EtcdServer{
+				appliedIndex:   tc.appliedIndex,
+				committedIndex: tc.committedIndex,
+				stopping:       make(chan struct{}, 1),
+				applyWait:      wait.NewTimeList(),
+			}
+
+			if tc.action != nil {
+				go tc.action(s)
+			}
+
+			err := s.waitAppliedIndex()
+
+			if err != tc.ExpectedError {
+				t.Errorf("Unexpected error, want (%v), got (%v)", tc.ExpectedError, err)
+			}
+		})
+	}
+}

etcdserver/v3_server.go

@@ -40,6 +40,10 @@ const (
 	// We should stop accepting new proposals if the gap growing to a certain point.
 	maxGapBetweenApplyAndCommitIndex = 5000
 	traceThreshold                   = 100 * time.Millisecond
+
+	// The timeout for the node to catch up its applied index, and is used in
+	// lease related operations, such as LeaseRenew and LeaseTimeToLive.
+	applyTimeout = time.Second
 )
 
 type RaftKV interface {
@@ -257,6 +261,18 @@ func (s *EtcdServer) LeaseGrant(ctx context.Context, r *pb.LeaseGrantRequest) (*
 	return resp.(*pb.LeaseGrantResponse), nil
 }
 
+func (s *EtcdServer) waitAppliedIndex() error {
+	select {
+	case <-s.ApplyWait():
+	case <-s.stopping:
+		return ErrStopped
+	case <-time.After(applyTimeout):
+		return ErrTimeoutWaitAppliedIndex
+	}
+
+	return nil
+}
+
 func (s *EtcdServer) LeaseRevoke(ctx context.Context, r *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error) {
 	resp, err := s.raftRequestOnce(ctx, pb.InternalRaftRequest{LeaseRevoke: r})
 	if err != nil {
@@ -266,26 +282,32 @@ func (s *EtcdServer) LeaseRevoke(ctx context.Context, r *pb.LeaseRevokeRequest)
 }
 
 func (s *EtcdServer) LeaseRenew(ctx context.Context, id lease.LeaseID) (int64, error) {
-	ttl, err := s.lessor.Renew(id)
-	if err == nil { // already requested to primary lessor(leader)
-		return ttl, nil
-	}
-	if err != lease.ErrNotPrimary {
-		return -1, err
+	if s.isLeader() {
+		if err := s.waitAppliedIndex(); err != nil {
+			return 0, err
+		}
+
+		ttl, err := s.lessor.Renew(id)
+		if err == nil { // already requested to primary lessor(leader)
+			return ttl, nil
+		}
+		if err != lease.ErrNotPrimary {
+			return -1, err
+		}
 	}
 
 	cctx, cancel := context.WithTimeout(ctx, s.Cfg.ReqTimeout())
 	defer cancel()
 
 	// renewals don't go through raft; forward to leader manually
-	for cctx.Err() == nil && err != nil {
+	for cctx.Err() == nil {
 		leader, lerr := s.waitLeader(cctx)
 		if lerr != nil {
 			return -1, lerr
 		}
 		for _, url := range leader.PeerURLs {
 			lurl := url + leasehttp.LeasePrefix
-			ttl, err = leasehttp.RenewHTTP(cctx, id, lurl, s.peerRt)
+			ttl, err := leasehttp.RenewHTTP(cctx, id, lurl, s.peerRt)
 			if err == nil || err == lease.ErrLeaseNotFound {
 				return ttl, err
 			}
@@ -299,7 +321,10 @@ func (s *EtcdServer) LeaseRenew(ctx context.Context, id lease.LeaseID) (int64, e
 }
 
 func (s *EtcdServer) LeaseTimeToLive(ctx context.Context, r *pb.LeaseTimeToLiveRequest) (*pb.LeaseTimeToLiveResponse, error) {
-	if s.Leader() == s.ID() {
+	if s.isLeader() {
+		if err := s.waitAppliedIndex(); err != nil {
+			return nil, err
+		}
 		// primary; timetolive directly from leader
 		le := s.lessor.Lookup(lease.LeaseID(r.ID))
 		if le == nil {

integration/v3_lease_test.go

@@ -362,13 +362,13 @@ func testLeaseStress(t *testing.T, stresser func(context.Context, pb.LeaseClient
 	defer cancel()
 	errc := make(chan error)
 
-	for i := 0; i < 30; i++ {
+	for i := 0; i < 100; i++ {
 		for j := 0; j < 3; j++ {
 			go func(i int) { errc <- stresser(ctx, toGRPC(clus.Client(i)).Lease) }(j)
 		}
 	}
 
-	for i := 0; i < 90; i++ {
+	for i := 0; i < 300; i++ {
 		if err := <-errc; err != nil {
 			t.Fatal(err)
 		}