client: update the leader even if the connection creation fails

client/client.go

@@ -744,16 +744,18 @@ func (c *client) checkLeaderHealth(ctx context.Context) {
 	if client := c.pdSvcDiscovery.GetServingEndpointClientConn(); client != nil {
 		healthCli := healthpb.NewHealthClient(client)
 		resp, err := healthCli.Check(ctx, &healthpb.HealthCheckRequest{Service: ""})
-		rpcErr, ok := status.FromError(err)
 		failpoint.Inject("unreachableNetwork1", func() {
 			resp = nil
 			err = status.New(codes.Unavailable, "unavailable").Err()
 		})
+		rpcErr, ok := status.FromError(err)
 		if (ok && isNetworkError(rpcErr.Code())) || resp.GetStatus() != healthpb.HealthCheckResponse_SERVING {
 			atomic.StoreInt32(&(c.leaderNetworkFailure), int32(1))
 		} else {
 			atomic.StoreInt32(&(c.leaderNetworkFailure), int32(0))
 		}
+	} else {
+		atomic.StoreInt32(&(c.leaderNetworkFailure), int32(1))
 	}
 }
 

client/grpcutil/grpcutil.go

@@ -21,6 +21,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/pingcap/errors"
+	"github.com/pingcap/failpoint"
 	"github.com/pingcap/log"
 	"github.com/tikv/pd/client/errs"
 	"github.com/tikv/pd/client/tlsutil"
@@ -88,6 +90,12 @@ func GetOrCreateGRPCConn(ctx context.Context, clientConns *sync.Map, addr string
 	dCtx, cancel := context.WithTimeout(ctx, dialTimeout)
 	defer cancel()
 	cc, err := GetClientConn(dCtx, addr, tlsConfig, opt...)
+	failpoint.Inject("unreachableNetwork2", func(val failpoint.Value) {
+		if val, ok := val.(string); ok && val == addr {
+			cc = nil
+			err = errors.Errorf("unreachable network")
+		}
+	})
 	if err != nil {
 		return nil, err
 	}

client/pd_service_discovery.go

@@ -614,7 +614,6 @@ func (c *pdServiceDiscovery) switchLeader(addrs []string) error {
 
 	if _, err := c.GetOrCreateGRPCConn(addr); err != nil {
 		log.Warn("[pd] failed to connect leader", zap.String("leader", addr), errs.ZapError(err))
-		return err
 	}
 	// Set PD leader and Global TSO Allocator (which is also the PD leader)
 	c.leader.Store(addr)

client/tso_client.go

@@ -171,9 +171,10 @@ func (c *tsoClient) GetTSOAllocatorClientConnByDCLocation(dcLocation string) (*g
 	if !ok {
 		panic(fmt.Sprintf("the allocator leader in %s should exist", dcLocation))
 	}
+	// todo: if we support local tso forward, we should get or create client conns.
 	cc, ok := c.svcDiscovery.GetClientConns().Load(url)
 	if !ok {
-		panic(fmt.Sprintf("the client connection of %s in %s should exist", url, dcLocation))
+		return nil, url.(string)
 	}
 	return cc.(*grpc.ClientConn), url.(string)
 }

client/tso_dispatcher.go

@@ -254,7 +254,7 @@ func (c *tsoClient) checkAllocator(
 		requestForwarded.WithLabelValues(forwardedHostTrim, addrTrim).Set(0)
 	}()
 	cc, u := c.GetTSOAllocatorClientConnByDCLocation(dc)
-	healthCli := healthpb.NewHealthClient(cc)
+	var healthCli healthpb.HealthClient
 	ticker := time.NewTicker(time.Second)
 	defer ticker.Stop()
 	for {
@@ -263,20 +263,25 @@ func (c *tsoClient) checkAllocator(
 			log.Info("[tso] the leader of the allocator leader is changed", zap.String("dc", dc), zap.String("origin", url), zap.String("new", u))
 			return
 		}
-		healthCtx, healthCancel := context.WithTimeout(dispatcherCtx, c.option.timeout)
-		resp, err := healthCli.Check(healthCtx, &healthpb.HealthCheckRequest{Service: ""})
-		failpoint.Inject("unreachableNetwork", func() {
-			resp.Status = healthpb.HealthCheckResponse_UNKNOWN
-		})
-		healthCancel()
-		if err == nil && resp.GetStatus() == healthpb.HealthCheckResponse_SERVING {
-			// create a stream of the original allocator
-			cctx, cancel := context.WithCancel(dispatcherCtx)
-			stream, err := c.tsoStreamBuilderFactory.makeBuilder(cc).build(cctx, cancel, c.option.timeout)
-			if err == nil && stream != nil {
-				log.Info("[tso] recover the original tso stream since the network has become normal", zap.String("dc", dc), zap.String("url", url))
-				updateAndClear(url, &tsoConnectionContext{url, stream, cctx, cancel})
-				return
+		if healthCli == nil && cc != nil {
+			healthCli = healthpb.NewHealthClient(cc)
+		}
+		if healthCli != nil {
+			healthCtx, healthCancel := context.WithTimeout(dispatcherCtx, c.option.timeout)
+			resp, err := healthCli.Check(healthCtx, &healthpb.HealthCheckRequest{Service: ""})
+			failpoint.Inject("unreachableNetwork", func() {
+				resp.Status = healthpb.HealthCheckResponse_UNKNOWN
+			})
+			healthCancel()
+			if err == nil && resp.GetStatus() == healthpb.HealthCheckResponse_SERVING {
+				// create a stream of the original allocator
+				cctx, cancel := context.WithCancel(dispatcherCtx)
+				stream, err := c.tsoStreamBuilderFactory.makeBuilder(cc).build(cctx, cancel, c.option.timeout)
+				if err == nil && stream != nil {
+					log.Info("[tso] recover the original tso stream since the network has become normal", zap.String("dc", dc), zap.String("url", url))
+					updateAndClear(url, &tsoConnectionContext{url, stream, cctx, cancel})
+					return
+				}
 			}
 		}
 		select {
@@ -285,7 +290,7 @@ func (c *tsoClient) checkAllocator(
 		case <-ticker.C:
 			// To ensure we can get the latest allocator leader
 			// and once the leader is changed, we can exit this function.
-			_, u = c.GetTSOAllocatorClientConnByDCLocation(dc)
+			cc, u = c.GetTSOAllocatorClientConnByDCLocation(dc)
 		}
 	}
 }
@@ -597,29 +602,32 @@ func (c *tsoClient) tryConnectToTSO(
 	for i := 0; i < maxRetryTimes; i++ {
 		c.svcDiscovery.ScheduleCheckMemberChanged()
 		cc, url = c.GetTSOAllocatorClientConnByDCLocation(dc)
-		cctx, cancel := context.WithCancel(dispatcherCtx)
-		stream, err = c.tsoStreamBuilderFactory.makeBuilder(cc).build(cctx, cancel, c.option.timeout)
-		failpoint.Inject("unreachableNetwork", func() {
-			stream = nil
-			err = status.New(codes.Unavailable, "unavailable").Err()
-		})
-		if stream != nil && err == nil {
-			updateAndClear(url, &tsoConnectionContext{url, stream, cctx, cancel})
-			return nil
-		}
-
-		if err != nil && c.option.enableForwarding {
-			// The reason we need to judge if the error code is equal to "Canceled" here is that
-			// when we create a stream we use a goroutine to manually control the timeout of the connection.
-			// There is no need to wait for the transport layer timeout which can reduce the time of unavailability.
-			// But it conflicts with the retry mechanism since we use the error code to decide if it is caused by network error.
-			// And actually the `Canceled` error can be regarded as a kind of network error in some way.
-			if rpcErr, ok := status.FromError(err); ok && (isNetworkError(rpcErr.Code()) || rpcErr.Code() == codes.Canceled) {
-				networkErrNum++
+		if cc != nil {
+			cctx, cancel := context.WithCancel(dispatcherCtx)
+			stream, err = c.tsoStreamBuilderFactory.makeBuilder(cc).build(cctx, cancel, c.option.timeout)
+			failpoint.Inject("unreachableNetwork", func() {
+				stream = nil
+				err = status.New(codes.Unavailable, "unavailable").Err()
+			})
+			if stream != nil && err == nil {
+				updateAndClear(url, &tsoConnectionContext{url, stream, cctx, cancel})
+				return nil
 			}
-		}
 
-		cancel()
+			if err != nil && c.option.enableForwarding {
+				// The reason we need to judge if the error code is equal to "Canceled" here is that
+				// when we create a stream we use a goroutine to manually control the timeout of the connection.
+				// There is no need to wait for the transport layer timeout which can reduce the time of unavailability.
+				// But it conflicts with the retry mechanism since we use the error code to decide if it is caused by network error.
+				// And actually the `Canceled` error can be regarded as a kind of network error in some way.
+				if rpcErr, ok := status.FromError(err); ok && (isNetworkError(rpcErr.Code()) || rpcErr.Code() == codes.Canceled) {
+					networkErrNum++
+				}
+			}
+			cancel()
+		} else {
+			networkErrNum++
+		}
 		select {
 		case <-dispatcherCtx.Done():
 			return err

tests/integrations/client/client_test.go

@@ -518,7 +518,7 @@ func TestCustomTimeout(t *testing.T) {
 	re.Less(time.Since(start), 2*time.Second)
 }
 
-func TestGetRegionFromFollowerClient(t *testing.T) {
+func TestGetRegionByFollowerForwarding(t *testing.T) {
 	re := require.New(t)
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -544,7 +544,7 @@ func TestGetRegionFromFollowerClient(t *testing.T) {
 }
 
 // case 1: unreachable -> normal
-func TestGetTsoFromFollowerClient1(t *testing.T) {
+func TestGetTsoByFollowerForwarding1(t *testing.T) {
 	re := require.New(t)
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -575,7 +575,7 @@ func TestGetTsoFromFollowerClient1(t *testing.T) {
 }
 
 // case 2: unreachable -> leader transfer -> normal
-func TestGetTsoFromFollowerClient2(t *testing.T) {
+func TestGetTsoByFollowerForwarding2(t *testing.T) {
 	re := require.New(t)
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -609,6 +609,101 @@ func TestGetTsoFromFollowerClient2(t *testing.T) {
 	checkTS(re, cli, lastTS)
 }
 
+// case 3: network partition between client and follower A -> transfer leader to follower A -> normal
+func TestGetTsoAndRegionByFollowerForwarding(t *testing.T) {
+	re := require.New(t)
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	pd.LeaderHealthCheckInterval = 100 * time.Millisecond
+	cluster, err := tests.NewTestCluster(ctx, 3)
+	re.NoError(err)
+	defer cluster.Destroy()
+
+	endpoints := runServer(re, cluster)
+	re.NotEmpty(cluster.WaitLeader())
+	leader := cluster.GetLeaderServer()
+	grpcPDClient := testutil.MustNewGrpcClient(re, leader.GetAddr())
+	testutil.Eventually(re, func() bool {
+		regionHeartbeat, err := grpcPDClient.RegionHeartbeat(ctx)
+		re.NoError(err)
+		regionID := regionIDAllocator.alloc()
+		region := &metapb.Region{
+			Id: regionID,
+			RegionEpoch: &metapb.RegionEpoch{
+				ConfVer: 1,
+				Version: 1,
+			},
+			Peers: peers,
+		}
+		req := &pdpb.RegionHeartbeatRequest{
+			Header: newHeader(leader.GetServer()),
+			Region: region,
+			Leader: peers[0],
+		}
+		err = regionHeartbeat.Send(req)
+		re.NoError(err)
+		_, err = regionHeartbeat.Recv()
+		return err == nil
+	})
+	follower := cluster.GetServer(cluster.GetFollower())
+	re.NoError(failpoint.Enable("github.com/tikv/pd/client/grpcutil/unreachableNetwork2", fmt.Sprintf("return(\"%s\")", follower.GetAddr())))
+
+	cli := setupCli(re, ctx, endpoints, pd.WithForwardingOption(true))
+	var lastTS uint64
+	testutil.Eventually(re, func() bool {
+		physical, logical, err := cli.GetTS(context.TODO())
+		if err == nil {
+			lastTS = tsoutil.ComposeTS(physical, logical)
+			return true
+		}
+		t.Log(err)
+		return false
+	})
+	lastTS = checkTS(re, cli, lastTS)
+	r, err := cli.GetRegion(context.Background(), []byte("a"))
+	re.NoError(err)
+	re.NotNil(r)
+	leader.GetServer().GetMember().ResignEtcdLeader(leader.GetServer().Context(),
+		leader.GetServer().Name(), follower.GetServer().Name())
+	re.NotEmpty(cluster.WaitLeader())
+	testutil.Eventually(re, func() bool {
+		physical, logical, err := cli.GetTS(context.TODO())
+		if err == nil {
+			lastTS = tsoutil.ComposeTS(physical, logical)
+			return true
+		}
+		t.Log(err)
+		return false
+	})
+	lastTS = checkTS(re, cli, lastTS)
+	testutil.Eventually(re, func() bool {
+		r, err = cli.GetRegion(context.Background(), []byte("a"))
+		if err == nil && r != nil {
+			return true
+		}
+		return false
+	})
+
+	re.NoError(failpoint.Disable("github.com/tikv/pd/client/grpcutil/unreachableNetwork2"))
+	testutil.Eventually(re, func() bool {
+		physical, logical, err := cli.GetTS(context.TODO())
+		if err == nil {
+			lastTS = tsoutil.ComposeTS(physical, logical)
+			return true
+		}
+		t.Log(err)
+		return false
+	})
+	lastTS = checkTS(re, cli, lastTS)
+	testutil.Eventually(re, func() bool {
+		r, err = cli.GetRegion(context.Background(), []byte("a"))
+		if err == nil && r != nil {
+			return true
+		}
+		return false
+	})
+}
+
 func checkTS(re *require.Assertions, cli pd.Client, lastTS uint64) uint64 {
 	for i := 0; i < tsoRequestRound; i++ {
 		physical, logical, err := cli.GetTS(context.TODO())