*: fix scheduling can not immediately start after transfer leader (#4875

) (#4973)

close #4769, ref #4875

Signed-off-by: ti-chi-bot <[email protected]>
Signed-off-by: Ryan Leung <[email protected]>

Co-authored-by: Ryan Leung <[email protected]>

server/cluster/cluster.go

@@ -538,6 +538,11 @@ func (c *RaftCluster) HandleStoreHeartbeat(stats *pdpb.StoreStats) error {
 	return nil
 }
 
+// IsPrepared return true if the prepare checker is ready.
+func (c *RaftCluster) IsPrepared() bool {
+	return c.prepareChecker.isPrepared()
+}
+
 var regionGuide = core.GenerateRegionGuideFunc(true)
 
 // processRegionHeartbeat updates the region information.
@@ -608,7 +613,7 @@ func (c *RaftCluster) processRegionHeartbeat(region *core.RegionInfo) error {
 		regionEventCounter.WithLabelValues("update_cache").Inc()
 	}
 
-	if isNew {
+	if !c.IsPrepared() && isNew {
 		c.prepareChecker.collect(region)
 	}
 
@@ -656,7 +661,6 @@ func (c *RaftCluster) updateStoreStatusLocked(id uint64) {
 	c.core.UpdateStoreStatus(id, leaderCount, regionCount, pendingPeerCount, leaderRegionSize, regionSize)
 }
 
-//nolint:unused
 func (c *RaftCluster) getClusterID() uint64 {
 	c.RLock()
 	defer c.RUnlock()
@@ -1381,13 +1385,6 @@ func (c *RaftCluster) GetComponentManager() *component.Manager {
 	return c.componentManager
 }
 
-// isPrepared if the cluster information is collected
-func (c *RaftCluster) isPrepared() bool {
-	c.RLock()
-	defer c.RUnlock()
-	return c.prepareChecker.check(c)
-}
-
 // GetStoresLoads returns load stats of all stores.
 func (c *RaftCluster) GetStoresLoads() map[uint64][]float64 {
 	c.RLock()
@@ -1447,10 +1444,11 @@ func (c *RaftCluster) FitRegion(region *core.RegionInfo) *placement.RegionFit {
 }
 
 type prepareChecker struct {
+	sync.RWMutex
 	reactiveRegions map[uint64]int
 	start           time.Time
 	sum             int
-	isPrepared      bool
+	prepared        bool
 }
 
 func newPrepareChecker() *prepareChecker {
@@ -1461,12 +1459,18 @@ func newPrepareChecker() *prepareChecker {
 }
 
 // Before starting up the scheduler, we need to take the proportion of the regions on each store into consideration.
-func (checker *prepareChecker) check(c *RaftCluster) bool {
-	if checker.isPrepared || time.Since(checker.start) > collectTimeout {
+func (checker *prepareChecker) check(c *core.BasicCluster) bool {
+	checker.Lock()
+	defer checker.Unlock()
+	if checker.prepared {
+		return true
+	}
+	if time.Since(checker.start) > collectTimeout {
+		checker.prepared = true
 		return true
 	}
 	// The number of active regions should be more than total region of all stores * collectFactor
-	if float64(c.core.GetRegionCount())*collectFactor > float64(checker.sum) {
+	if float64(c.GetRegionCount())*collectFactor > float64(checker.sum) {
 		return false
 	}
 	for _, store := range c.GetStores() {
@@ -1475,21 +1479,29 @@ func (checker *prepareChecker) check(c *RaftCluster) bool {
 		}
 		storeID := store.GetID()
 		// For each store, the number of active regions should be more than total region of the store * collectFactor
-		if float64(c.core.GetStoreRegionCount(storeID))*collectFactor > float64(checker.reactiveRegions[storeID]) {
+		if float64(c.GetStoreRegionCount(storeID))*collectFactor > float64(checker.reactiveRegions[storeID]) {
 			return false
 		}
 	}
-	checker.isPrepared = true
+	checker.prepared = true
 	return true
 }
 
 func (checker *prepareChecker) collect(region *core.RegionInfo) {
+	checker.Lock()
+	defer checker.Unlock()
 	for _, p := range region.GetPeers() {
 		checker.reactiveRegions[p.GetStoreId()]++
 	}
 	checker.sum++
 }
 
+func (checker *prepareChecker) isPrepared() bool {
+	checker.RLock()
+	defer checker.RUnlock()
+	return checker.prepared
+}
+
 // GetHotWriteRegions gets hot write regions' info.
 func (c *RaftCluster) GetHotWriteRegions() *statistics.StoreHotPeersInfos {
 	c.RLock()

server/cluster/coordinator.go

@@ -252,6 +252,9 @@ func (c *coordinator) drivePushOperator() {
 
 func (c *coordinator) run() {
 	ticker := time.NewTicker(runSchedulerCheckInterval)
+	failpoint.Inject("changeCoordinatorTicker", func() {
+		ticker = time.NewTicker(100 * time.Millisecond)
+	})
 	defer ticker.Stop()
 	log.Info("coordinator starts to collect cluster information")
 	for {
@@ -569,7 +572,7 @@ func (c *coordinator) resetHotSpotMetrics() {
 }
 
 func (c *coordinator) shouldRun() bool {
-	return c.cluster.isPrepared()
+	return c.cluster.prepareChecker.check(c.cluster.core)
 }
 
 func (c *coordinator) addScheduler(scheduler schedule.Scheduler, args ...string) error {

server/cluster/coordinator_test.go

@@ -189,9 +189,8 @@ func (s *testCoordinatorSuite) TestBasic(c *C) {
 }
 
 func (s *testCoordinatorSuite) TestDispatch(c *C) {
-	tc, co, cleanup := prepare(nil, func(tc *testCluster) { tc.prepareChecker.isPrepared = true }, nil, c)
+	tc, co, cleanup := prepare(nil, func(tc *testCluster) { tc.prepareChecker.prepared = true }, nil, c)
 	defer cleanup()
-
 	// Transfer peer from store 4 to store 1.
 	c.Assert(tc.addRegionStore(4, 40), IsNil)
 	c.Assert(tc.addRegionStore(3, 30), IsNil)

server/core/region.go

@@ -55,6 +55,7 @@ type RegionInfo struct {
 	approximateKeys   int64
 	interval          *pdpb.TimeInterval
 	replicationStatus *replication_modepb.RegionReplicationStatus
+	fromHeartbeat     bool
 }
 
 // NewRegionInfo creates RegionInfo with region's meta and leader peer.
@@ -98,7 +99,7 @@ const (
 )
 
 // RegionFromHeartbeat constructs a Region from region heartbeat.
-func RegionFromHeartbeat(heartbeat *pdpb.RegionHeartbeatRequest) *RegionInfo {
+func RegionFromHeartbeat(heartbeat *pdpb.RegionHeartbeatRequest, opts ...RegionCreateOption) *RegionInfo {
 	// Convert unit to MB.
 	// If region is empty or less than 1MB, use 1MB instead.
 	regionSize := heartbeat.GetApproximateSize() / (1 << 20)
@@ -122,6 +123,10 @@ func RegionFromHeartbeat(heartbeat *pdpb.RegionHeartbeatRequest) *RegionInfo {
 		replicationStatus: heartbeat.GetReplicationStatus(),
 	}
 
+	for _, opt := range opts {
+		opt(region)
+	}
+
 	if region.writtenKeys >= ImpossibleFlowSize || region.writtenBytes >= ImpossibleFlowSize {
 		region.writtenKeys = 0
 		region.writtenBytes = 0
@@ -439,6 +444,11 @@ func (r *RegionInfo) GetReplicationStatus() *replication_modepb.RegionReplicatio
 	return r.replicationStatus
 }
 
+// IsFromHeartbeat returns whether the region info is from the region heartbeat.
+func (r *RegionInfo) IsFromHeartbeat() bool {
+	return r.fromHeartbeat
+}
+
 // RegionGuideFunc is a function that determines which follow-up operations need to be performed based on the origin
 // and new region information.
 type RegionGuideFunc func(region, origin *RegionInfo, traceRegionFlow bool) (isNew, saveKV, saveCache, needSync bool)
@@ -523,6 +533,9 @@ func GenerateRegionGuideFunc(enableLog bool) RegionGuideFunc {
 					region.GetReplicationStatus().GetStateId() != origin.GetReplicationStatus().GetStateId()) {
 				saveCache = true
 			}
+			if !origin.IsFromHeartbeat() {
+				isNew = true
+			}
 		}
 		return
 	}

server/core/region_option.go

@@ -276,3 +276,10 @@ func WithInterval(interval *pdpb.TimeInterval) RegionCreateOption {
 		region.interval = interval
 	}
 }
+
+// SetFromHeartbeat sets if the region info comes from the region heartbeat.
+func SetFromHeartbeat(fromHeartbeat bool) RegionCreateOption {
+	return func(region *RegionInfo) {
+		region.fromHeartbeat = fromHeartbeat
+	}
+}

server/grpc_service.go

@@ -586,7 +586,7 @@ func (s *Server) RegionHeartbeat(stream pdpb.PD_RegionHeartbeatServer) error {
 			lastBind = time.Now()
 		}
 
-		region := core.RegionFromHeartbeat(request)
+		region := core.RegionFromHeartbeat(request, core.SetFromHeartbeat(true))
 		if region.GetLeader() == nil {
 			log.Error("invalid request, the leader is nil", zap.Reflect("request", request), errs.ZapError(errs.ErrLeaderNil))
 			regionHeartbeatCounter.WithLabelValues(storeAddress, storeLabel, "report", "invalid-leader").Inc()

server/region_syncer/client.go

@@ -201,9 +201,10 @@ func (s *RegionSyncer) StartSyncWithLeader(addr string) {
 							core.SetWrittenKeys(stats[i].KeysWritten),
 							core.SetReadBytes(stats[i].BytesRead),
 							core.SetReadKeys(stats[i].KeysRead),
+							core.SetFromHeartbeat(false),
 						)
 					} else {
-						region = core.NewRegionInfo(r, regionLeader)
+						region = core.NewRegionInfo(r, regionLeader, core.SetFromHeartbeat(false))
 					}
 
 					origin, err := bc.PreCheckPutRegion(region)

tests/server/region_syncer/region_syncer_test.go

@@ -19,6 +19,7 @@ import (
 	"time"
 
 	. "github.com/pingcap/check"
+	"github.com/pingcap/failpoint"
 	"github.com/pingcap/kvproto/pkg/metapb"
 	"github.com/tikv/pd/pkg/mock/mockid"
 	"github.com/tikv/pd/pkg/testutil"
@@ -223,5 +224,72 @@ func (s *serverTestSuite) TestFullSyncWithAddMember(c *C) {
 	c.Assert(err, IsNil)
 	c.Assert(cluster.WaitLeader(), Equals, "pd2")
 	loadRegions := pd2.GetServer().GetRaftCluster().GetRegions()
-	c.Assert(len(loadRegions), Equals, regionLen)
+	c.Assert(loadRegions, HasLen, regionLen)
+}
+
+func (s *serverTestSuite) TestPrepareChecker(c *C) {
+	c.Assert(failpoint.Enable("github.com/tikv/pd/server/cluster/changeCoordinatorTicker", `return(true)`), IsNil)
+	defer failpoint.Disable("github.com/tikv/pd/server/cluster/changeCoordinatorTicker")
+	cluster, err := tests.NewTestCluster(s.ctx, 1, func(conf *config.Config, serverName string) { conf.PDServerCfg.UseRegionStorage = true })
+	defer cluster.Destroy()
+	c.Assert(err, IsNil)
+
+	err = cluster.RunInitialServers()
+	c.Assert(err, IsNil)
+	cluster.WaitLeader()
+	leaderServer := cluster.GetServer(cluster.GetLeader())
+	c.Assert(leaderServer.BootstrapCluster(), IsNil)
+	rc := leaderServer.GetServer().GetRaftCluster()
+	c.Assert(rc, NotNil)
+	regionLen := 110
+	regions := initRegions(regionLen)
+	for _, region := range regions {
+		err = rc.HandleRegionHeartbeat(region)
+		c.Assert(err, IsNil)
+	}
+
+	// ensure flush to region storage
+	time.Sleep(3 * time.Second)
+	c.Assert(leaderServer.GetRaftCluster().IsPrepared(), IsTrue)
+
+	// join new PD
+	pd2, err := cluster.Join(s.ctx)
+	c.Assert(err, IsNil)
+	err = pd2.Run()
+	c.Assert(err, IsNil)
+	// waiting for synchronization to complete
+	time.Sleep(3 * time.Second)
+	err = cluster.ResignLeader()
+	c.Assert(err, IsNil)
+	c.Assert(cluster.WaitLeader(), Equals, "pd2")
+	leaderServer = cluster.GetServer(cluster.GetLeader())
+	rc = leaderServer.GetServer().GetRaftCluster()
+	for _, region := range regions {
+		err = rc.HandleRegionHeartbeat(region)
+		c.Assert(err, IsNil)
+	}
+	time.Sleep(time.Second)
+	c.Assert(rc.IsPrepared(), IsTrue)
+}
+
+func initRegions(regionLen int) []*core.RegionInfo {
+	allocator := &idAllocator{allocator: mockid.NewIDAllocator()}
+	regions := make([]*core.RegionInfo, 0, regionLen)
+	for i := 0; i < regionLen; i++ {
+		r := &metapb.Region{
+			Id: allocator.alloc(),
+			RegionEpoch: &metapb.RegionEpoch{
+				ConfVer: 1,
+				Version: 1,
+			},
+			StartKey: []byte{byte(i)},
+			EndKey:   []byte{byte(i + 1)},
+			Peers: []*metapb.Peer{
+				{Id: allocator.alloc(), StoreId: uint64(0)},
+				{Id: allocator.alloc(), StoreId: uint64(0)},
+			},
+		}
+		regions = append(regions, core.NewRegionInfo(r, r.Peers[0]))
+	}
+	return regions
 }