ms, tso: pick some enhancements (tikv#113)

* Add keyspace and keyspace group info to the time fallback log. (tikv#6581) ref tikv#5895 Add keyspace and keyspace group info to the time fallback log to help debugging time fallback issue in multi-timeline scenario. Signed-off-by: Bin Shi <[email protected]> * config, server: fault injection in TSO Proxy (tikv#6588) ref tikv#5895 Add failure test cases. Signed-off-by: Bin Shi <[email protected]> * mcs, tso: fix stream Send() and CloseSend() data race issue in TSO proxy (tikv#6591) close tikv#6590 No need to call SendClose(), because TSO proxy will cancel the stream context which will cause the corresponding grpc stream on the server side to exit. Signed-off-by: Bin Shi <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com> * tests: make TestSplitKeyspaceGroup stable (tikv#6584) close tikv#6571 Signed-off-by: lhy1024 <[email protected]> --------- Signed-off-by: Bin Shi <[email protected]> Signed-off-by: lhy1024 <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com> Co-authored-by: lhy1024 <[email protected]>
JmPotato · Jun 13, 2023 · 188d0d8 · 188d0d8
1 parent 1007f82
commit 188d0d8
Show file tree

Hide file tree

Showing 11 changed files with 203 additions and 54 deletions.
diff --git a/client/tso_dispatcher.go b/client/tso_dispatcher.go
@@ -739,8 +739,10 @@ func (c *tsoClient) compareAndSwapTS(dcLocation string, physical, firstLogical i
 	// to compare with the new TSO's first logical. For example, if we have a TSO resp with logical 10, count 5, then
 	// all TSOs we get will be [6, 7, 8, 9, 10].
 	if tsoutil.TSLessEqual(physical, firstLogical, lastPhysical, lastLogical) {
-		panic(errors.Errorf("%s timestamp fallback, newly acquired ts (%d, %d) is less or equal to last one (%d, %d)",
-			dcLocation, physical, firstLogical, lastPhysical, lastLogical))
+		panic(errors.Errorf(
+			"%s timestamp fallback, new ts (%d, %d) <= the last one (%d, %d). keyspace: %d, keyspace group: %d",
+			dcLocation, physical, firstLogical, lastPhysical, lastLogical,
+			c.svcDiscovery.GetKeyspaceID(), c.svcDiscovery.GetKeyspaceGroupID()))
 	}
 	lastTSOPointer.physical = physical
 	// Same as above, we save the largest logical part here.

diff --git a/pkg/keyspace/keyspace_test.go b/pkg/keyspace/keyspace_test.go
@@ -80,7 +80,7 @@ func (suite *keyspaceTestSuite) SetupTest() {
 	allocator := mockid.NewIDAllocator()
 	kgm := NewKeyspaceGroupManager(suite.ctx, store, nil, 0)
 	suite.manager = NewKeyspaceManager(suite.ctx, store, nil, allocator, &mockConfig{}, kgm)
-	suite.NoError(kgm.Bootstrap())
+	suite.NoError(kgm.Bootstrap(suite.ctx))
 	suite.NoError(suite.manager.Bootstrap())
 }
 

diff --git a/pkg/keyspace/tso_keyspace_group.go b/pkg/keyspace/tso_keyspace_group.go
@@ -18,6 +18,7 @@ import (
 	"context"
 	"encoding/json"
 	"strconv"
+	"strings"
 	"sync"
 	"time"
 
@@ -51,9 +52,11 @@ const (
 
 // GroupManager is the manager of keyspace group related data.
 type GroupManager struct {
-	ctx    context.Context
-	cancel context.CancelFunc
-	wg     sync.WaitGroup
+	ctx       context.Context
+	cancel    context.CancelFunc
+	wg        sync.WaitGroup
+	client    *clientv3.Client
+	clusterID uint64
 
 	sync.RWMutex
 	// groups is the cache of keyspace group related information.
@@ -90,24 +93,24 @@ func NewKeyspaceGroupManager(
 		cancel:             cancel,
 		store:              store,
 		groups:             groups,
+		client:             client,
+		clusterID:          clusterID,
 		nodesBalancer:      balancer.GenByPolicy[string](defaultBalancerPolicy),
 		serviceRegistryMap: make(map[string]string),
 	}
 
 	// If the etcd client is not nil, start the watch loop for the registered tso servers.
 	// The PD(TSO) Client relies on this info to discover tso servers.
-	if client != nil {
-		m.initTSONodesWatcher(client, clusterID)
-		m.wg.Add(2)
+	if m.client != nil {
+		m.initTSONodesWatcher(m.client, m.clusterID)
+		m.wg.Add(1)
 		go m.tsoNodesWatcher.StartWatchLoop()
-		go m.allocNodesToAllKeyspaceGroups()
 	}
-
 	return m
 }
 
 // Bootstrap saves default keyspace group info and init group mapping in the memory.
-func (m *GroupManager) Bootstrap() error {
+func (m *GroupManager) Bootstrap(ctx context.Context) error {
 	// Force the membership restriction that the default keyspace must belong to default keyspace group.
 	// Have no information to specify the distribution of the default keyspace group replicas, so just
 	// leave the replica/member list empty. The TSO service will assign the default keyspace group replica
@@ -137,6 +140,11 @@ func (m *GroupManager) Bootstrap() error {
 		m.groups[userKind].Put(group)
 	}
 
+	// It will only alloc node when the group manager is on API leader.
+	if m.client != nil {
+		m.wg.Add(1)
+		go m.allocNodesToAllKeyspaceGroups(ctx)
+	}
 	return nil
 }
 
@@ -146,7 +154,7 @@ func (m *GroupManager) Close() {
 	m.wg.Wait()
 }
 
-func (m *GroupManager) allocNodesToAllKeyspaceGroups() {
+func (m *GroupManager) allocNodesToAllKeyspaceGroups(ctx context.Context) {
 	defer logutil.LogPanic()
 	defer m.wg.Done()
 	ticker := time.NewTicker(allocNodesToKeyspaceGroupsInterval)
@@ -158,7 +166,7 @@ func (m *GroupManager) allocNodesToAllKeyspaceGroups() {
 	log.Info("start to alloc nodes to all keyspace groups")
 	for {
 		select {
-		case <-m.ctx.Done():
+		case <-ctx.Done():
 			log.Info("stop to alloc nodes to all keyspace groups")
 			return
 		case <-ticker.C:
@@ -338,11 +346,6 @@ func (m *GroupManager) saveKeyspaceGroups(keyspaceGroups []*endpoint.KeyspaceGro
 				Members:   keyspaceGroup.Members,
 				Keyspaces: keyspaceGroup.Keyspaces,
 			}
-			if oldKG.IsSplitting() {
-				newKG.SplitState = &endpoint.SplitState{
-					SplitSource: oldKG.SplitState.SplitSource,
-				}
-			}
 			err = m.store.SaveKeyspaceGroup(txn, newKG)
 			if err != nil {
 				return err
@@ -380,6 +383,8 @@ func (m *GroupManager) getKeyspaceConfigByKindLocked(userKind endpoint.UserKind)
 	return config, nil
 }
 
+var failpointOnce sync.Once
+
 // UpdateKeyspaceForGroup updates the keyspace field for the keyspace group.
 func (m *GroupManager) UpdateKeyspaceForGroup(userKind endpoint.UserKind, groupID string, keyspaceID uint32, mutation int) error {
 	// when server is not in API mode, we don't need to update the keyspace for keyspace group
@@ -391,6 +396,12 @@ func (m *GroupManager) UpdateKeyspaceForGroup(userKind endpoint.UserKind, groupI
 		return err
 	}
 
+	failpoint.Inject("externalAllocNode", func(val failpoint.Value) {
+		failpointOnce.Do(func() {
+			addrs := val.(string)
+			m.SetNodesForKeyspaceGroup(utils.DefaultKeyspaceGroupID, strings.Split(addrs, ","))
+		})
+	})
 	m.Lock()
 	defer m.Unlock()
 	return m.updateKeyspaceForGroupLocked(userKind, id, keyspaceID, mutation)
@@ -425,7 +436,6 @@ func (m *GroupManager) updateKeyspaceForGroupLocked(userKind endpoint.UserKind,
 		if err := m.saveKeyspaceGroups([]*endpoint.KeyspaceGroup{kg}, true); err != nil {
 			return err
 		}
-
 		m.groups[userKind].Put(kg)
 	}
 	return nil
@@ -696,8 +706,10 @@ func (m *GroupManager) AllocNodesForKeyspaceGroup(id uint32, desiredReplicaCount
 func (m *GroupManager) SetNodesForKeyspaceGroup(id uint32, nodes []string) error {
 	m.Lock()
 	defer m.Unlock()
-	return m.store.RunInTxn(m.ctx, func(txn kv.Txn) error {
-		kg, err := m.store.LoadKeyspaceGroup(txn, id)
+	var kg *endpoint.KeyspaceGroup
+	err := m.store.RunInTxn(m.ctx, func(txn kv.Txn) error {
+		var err error
+		kg, err = m.store.LoadKeyspaceGroup(txn, id)
 		if err != nil {
 			return err
 		}
@@ -714,6 +726,11 @@ func (m *GroupManager) SetNodesForKeyspaceGroup(id uint32, nodes []string) error
 		kg.Members = members
 		return m.store.SaveKeyspaceGroup(txn, kg)
 	})
+	if err != nil {
+		return err
+	}
+	m.groups[endpoint.StringUserKind(kg.UserKind)].Put(kg)
+	return nil
 }
 
 // IsExistNode checks if the node exists.

diff --git a/pkg/keyspace/tso_keyspace_group_test.go b/pkg/keyspace/tso_keyspace_group_test.go
@@ -48,7 +48,7 @@ func (suite *keyspaceGroupTestSuite) SetupTest() {
 	idAllocator := mockid.NewIDAllocator()
 	cluster := mockcluster.NewCluster(suite.ctx, config.NewTestOptions())
 	suite.kg = NewKeyspaceManager(suite.ctx, store, cluster, idAllocator, &mockConfig{}, suite.kgm)
-	suite.NoError(suite.kgm.Bootstrap())
+	suite.NoError(suite.kgm.Bootstrap(suite.ctx))
 }
 
 func (suite *keyspaceGroupTestSuite) TearDownTest() {

diff --git a/server/cluster/cluster.go b/server/cluster/cluster.go
@@ -270,7 +270,7 @@ func (c *RaftCluster) Start(s Server) error {
 	}
 
 	if s.IsAPIServiceMode() {
-		err = c.keyspaceGroupManager.Bootstrap()
+		err = c.keyspaceGroupManager.Bootstrap(c.ctx)
 		if err != nil {
 			return err
 		}

diff --git a/server/config/config.go b/server/config/config.go
@@ -85,9 +85,9 @@ type Config struct {
 	// Set this to 0 will disable TSO Proxy.
 	// Set this to the negative value to disable the limit.
 	MaxConcurrentTSOProxyStreamings int `toml:"max-concurrent-tso-proxy-streamings" json:"max-concurrent-tso-proxy-streamings"`
-	// TSOProxyClientRecvTimeout is the timeout for the TSO proxy to receive a tso request from a client via grpc TSO stream.
+	// TSOProxyRecvFromClientTimeout is the timeout for the TSO proxy to receive a tso request from a client via grpc TSO stream.
 	// After the timeout, the TSO proxy will close the grpc TSO stream.
-	TSOProxyClientRecvTimeout typeutil.Duration `toml:"tso-proxy-client-recv-timeout" json:"tso-proxy-client-recv-timeout"`
+	TSOProxyRecvFromClientTimeout typeutil.Duration `toml:"tso-proxy-recv-from-client-timeout" json:"tso-proxy-recv-from-client-timeout"`
 
 	// TSOSaveInterval is the interval to save timestamp.
 	TSOSaveInterval typeutil.Duration `toml:"tso-save-interval" json:"tso-save-interval"`
@@ -228,7 +228,7 @@ const (
 	defaultDRWaitStoreTimeout = time.Minute
 
 	defaultMaxConcurrentTSOProxyStreamings = 5000
-	defaultTSOProxyClientRecvTimeout       = 1 * time.Hour
+	defaultTSOProxyRecvFromClientTimeout   = 1 * time.Hour
 
 	defaultTSOSaveInterval = time.Duration(defaultLeaderLease) * time.Second
 	// defaultTSOUpdatePhysicalInterval is the default value of the config `TSOUpdatePhysicalInterval`.
@@ -458,7 +458,7 @@ func (c *Config) Adjust(meta *toml.MetaData, reloading bool) error {
 	}
 
 	configutil.AdjustInt(&c.MaxConcurrentTSOProxyStreamings, defaultMaxConcurrentTSOProxyStreamings)
-	configutil.AdjustDuration(&c.TSOProxyClientRecvTimeout, defaultTSOProxyClientRecvTimeout)
+	configutil.AdjustDuration(&c.TSOProxyRecvFromClientTimeout, defaultTSOProxyRecvFromClientTimeout)
 
 	configutil.AdjustInt64(&c.LeaderLease, defaultLeaderLease)
 	configutil.AdjustDuration(&c.TSOSaveInterval, defaultTSOSaveInterval)
@@ -1266,9 +1266,9 @@ func (c *Config) GetMaxConcurrentTSOProxyStreamings() int {
 	return c.MaxConcurrentTSOProxyStreamings
 }
 
-// GetTSOProxyClientRecvTimeout returns the TSO proxy client receive timeout.
-func (c *Config) GetTSOProxyClientRecvTimeout() time.Duration {
-	return c.TSOProxyClientRecvTimeout.Duration
+// GetTSOProxyRecvFromClientTimeout returns timeout value for TSO proxy receiving from the client.
+func (c *Config) GetTSOProxyRecvFromClientTimeout() time.Duration {
+	return c.TSOProxyRecvFromClientTimeout.Duration
 }
 
 // GetTSOUpdatePhysicalInterval returns TSO update physical interval.

diff --git a/server/grpc_service.go b/server/grpc_service.go
@@ -69,7 +69,7 @@ var (
 	ErrNotFoundTSOAddr                  = status.Errorf(codes.NotFound, "not found tso address")
 	ErrForwardTSOTimeout                = status.Errorf(codes.DeadlineExceeded, "forward tso request timeout")
 	ErrMaxCountTSOProxyRoutinesExceeded = status.Errorf(codes.ResourceExhausted, "max count of concurrent tso proxy routines exceeded")
-	ErrTSOProxyClientRecvTimeout        = status.Errorf(codes.DeadlineExceeded, "tso proxy client recv timeout. stream closed by server")
+	ErrTSOProxyRecvFromClientTimeout    = status.Errorf(codes.DeadlineExceeded, "tso proxy timeout when receiving from client; stream closed by server")
 )
 
 // GrpcServer wraps Server to provide grpc service.
@@ -409,9 +409,6 @@ func (s *GrpcServer) forwardTSO(stream pdpb.PD_TsoServer) error {
 	)
 	defer func() {
 		s.concurrentTSOProxyStreamings.Add(-1)
-		if forwardStream != nil {
-			forwardStream.CloseSend()
-		}
 		// cancel the forward stream
 		if cancel != nil {
 			cancel()
@@ -433,7 +430,7 @@ func (s *GrpcServer) forwardTSO(stream pdpb.PD_TsoServer) error {
 		default:
 		}
 
-		request, err := server.Recv(s.GetTSOProxyClientRecvTimeout())
+		request, err := server.Recv(s.GetTSOProxyRecvFromClientTimeout())
 		if err == io.EOF {
 			return nil
 		}
@@ -450,9 +447,6 @@ func (s *GrpcServer) forwardTSO(stream pdpb.PD_TsoServer) error {
 			return errors.WithStack(ErrNotFoundTSOAddr)
 		}
 		if forwardStream == nil || lastForwardedHost != forwardedHost {
-			if forwardStream != nil {
-				forwardStream.CloseSend()
-			}
 			if cancel != nil {
 				cancel()
 			}
@@ -548,6 +542,11 @@ func (s *GrpcServer) forwardTSORequestAsync(
 		DcLocation: request.GetDcLocation(),
 	}
 
+	failpoint.Inject("tsoProxySendToTSOTimeout", func() {
+		<-ctxTimeout.Done()
+		failpoint.Return()
+	})
+
 	if err := forwardStream.Send(tsopbReq); err != nil {
 		select {
 		case <-ctxTimeout.Done():
@@ -563,23 +562,21 @@ func (s *GrpcServer) forwardTSORequestAsync(
 	default:
 	}
 
+	failpoint.Inject("tsoProxyRecvFromTSOTimeout", func() {
+		<-ctxTimeout.Done()
+		failpoint.Return()
+	})
+
 	response, err := forwardStream.Recv()
 	if err != nil {
 		if strings.Contains(err.Error(), errs.NotLeaderErr) {
 			s.tsoPrimaryWatcher.ForceLoad()
 		}
-		select {
-		case <-ctxTimeout.Done():
-			return
-		case tsoRespCh <- &tsopbTSOResponse{err: err}:
-		}
-		return
 	}
-
 	select {
 	case <-ctxTimeout.Done():
 		return
-	case tsoRespCh <- &tsopbTSOResponse{response: response}:
+	case tsoRespCh <- &tsopbTSOResponse{response: response, err: err}:
 	}
 }
 
@@ -607,6 +604,10 @@ func (s *tsoServer) Send(m *pdpb.TsoResponse) error {
 	done := make(chan error, 1)
 	go func() {
 		defer logutil.LogPanic()
+		failpoint.Inject("tsoProxyFailToSendToClient", func() {
+			done <- errors.New("injected error")
+			failpoint.Return()
+		})
 		done <- s.stream.Send(m)
 	}()
 	select {
@@ -625,6 +626,11 @@ func (s *tsoServer) Recv(timeout time.Duration) (*pdpb.TsoRequest, error) {
 	if atomic.LoadInt32(&s.closed) == 1 {
 		return nil, io.EOF
 	}
+	failpoint.Inject("tsoProxyRecvFromClientTimeout", func(val failpoint.Value) {
+		if customTimeoutInSeconds, ok := val.(int); ok {
+			timeout = time.Duration(customTimeoutInSeconds) * time.Second
+		}
+	})
 	requestCh := make(chan *pdpbTSORequest, 1)
 	go func() {
 		defer logutil.LogPanic()
@@ -640,7 +646,7 @@ func (s *tsoServer) Recv(timeout time.Duration) (*pdpb.TsoRequest, error) {
 		return req.request, nil
 	case <-time.After(timeout):
 		atomic.StoreInt32(&s.closed, 1)
-		return nil, ErrTSOProxyClientRecvTimeout
+		return nil, ErrTSOProxyRecvFromClientTimeout
 	}
 }
 

diff --git a/server/server.go b/server/server.go
@@ -1844,9 +1844,9 @@ func (s *Server) GetMaxConcurrentTSOProxyStreamings() int {
 	return s.cfg.GetMaxConcurrentTSOProxyStreamings()
 }
 
-// GetTSOProxyClientRecvTimeout returns the TSO proxy client receive timeout.
-func (s *Server) GetTSOProxyClientRecvTimeout() time.Duration {
-	return s.cfg.GetTSOProxyClientRecvTimeout()
+// GetTSOProxyRecvFromClientTimeout returns timeout value for TSO proxy receiving from the client.
+func (s *Server) GetTSOProxyRecvFromClientTimeout() time.Duration {
+	return s.cfg.GetTSOProxyRecvFromClientTimeout()
 }
 
 // GetLeaderLease returns the leader lease.