scheduler(ticdc): revert 3b8d55 and do not return error when resolved…

…Ts less than checkpoint (pingcap#9953)

ref pingcap#9830, ref pingcap#9926

cdc/scheduler/internal/v3/agent/table.go

@@ -75,7 +75,7 @@ func (t *tableSpan) getTableSpanStatus(collectStat bool) tablepb.TableStatus {
 
 func newAddTableResponseMessage(status tablepb.TableStatus) *schedulepb.Message {
 	if status.Checkpoint.ResolvedTs < status.Checkpoint.CheckpointTs {
-		log.Panic("schedulerv3: resolved ts should not less than checkpoint ts",
+		log.Warn("schedulerv3: resolved ts should not less than checkpoint ts",
 			zap.Any("tableStatus", status),
 			zap.Any("checkpoint", status.Checkpoint.CheckpointTs),
 			zap.Any("resolved", status.Checkpoint.ResolvedTs))
@@ -100,7 +100,7 @@ func newRemoveTableResponseMessage(status tablepb.TableStatus) *schedulepb.Messa
 			// Advance resolved ts to checkpoint ts if table is removed.
 			status.Checkpoint.ResolvedTs = status.Checkpoint.CheckpointTs
 		} else {
-			log.Panic("schedulerv3: resolved ts should not less than checkpoint ts",
+			log.Warn("schedulerv3: resolved ts should not less than checkpoint ts",
 				zap.Any("tableStatus", status),
 				zap.Any("checkpoint", status.Checkpoint.CheckpointTs),
 				zap.Any("resolved", status.Checkpoint.ResolvedTs))

cdc/scheduler/internal/v3/replication/replication_set.go

@@ -162,9 +162,7 @@ func NewReplicationSet(
 			return nil, r.inconsistentError(table, captureID,
 				"schedulerv3: table id inconsistent")
 		}
-		if err := r.updateCheckpointAndStats(table.Checkpoint, table.Stats); err != nil {
-			return nil, errors.Trace(err)
-		}
+		r.updateCheckpointAndStats(table.Checkpoint, table.Stats)
 
 		switch table.State {
 		case tablepb.TableStateReplicating:
@@ -500,8 +498,8 @@ func (r *ReplicationSet) pollOnPrepare(
 		}
 	case tablepb.TableStateReplicating:
 		if r.Primary == captureID {
-			err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
-			return nil, false, err
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
+			return nil, false, nil
 		}
 	case tablepb.TableStateStopping, tablepb.TableStateStopped:
 		if r.Primary == captureID {
@@ -614,9 +612,7 @@ func (r *ReplicationSet) pollOnCommit(
 
 	case tablepb.TableStateStopped, tablepb.TableStateAbsent:
 		if r.Primary == captureID {
-			if err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats); err != nil {
-				return nil, false, errors.Trace(err)
-			}
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
 			original := r.Primary
 			r.clearPrimary()
 			if !r.hasRole(RoleSecondary) {
@@ -688,9 +684,7 @@ func (r *ReplicationSet) pollOnCommit(
 
 	case tablepb.TableStateReplicating:
 		if r.Primary == captureID {
-			if err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats); err != nil {
-				return nil, false, errors.Trace(err)
-			}
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
 			if r.hasRole(RoleSecondary) {
 				// Original primary is not stopped, ask for stopping.
 				return &schedulepb.Message{
@@ -725,8 +719,8 @@ func (r *ReplicationSet) pollOnCommit(
 
 	case tablepb.TableStateStopping:
 		if r.Primary == captureID && r.hasRole(RoleSecondary) {
-			err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
-			return nil, false, err
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
+			return nil, false, nil
 		} else if r.isInRole(captureID, RoleUndetermined) {
 			log.Info("schedulerv3: capture is stopping during Commit",
 				zap.String("namespace", r.Changefeed.Namespace),
@@ -755,8 +749,8 @@ func (r *ReplicationSet) pollOnReplicating(
 	switch input.State {
 	case tablepb.TableStateReplicating:
 		if r.Primary == captureID {
-			err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
-			return nil, false, err
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
+			return nil, false, nil
 		}
 		return nil, false, r.multiplePrimaryError(
 			input, captureID, "schedulerv3: multiple primary")
@@ -767,10 +761,7 @@ func (r *ReplicationSet) pollOnReplicating(
 	case tablepb.TableStateStopping:
 	case tablepb.TableStateStopped:
 		if r.Primary == captureID {
-			if err := r.updateCheckpointAndStats(input.Checkpoint, input.Stats); err != nil {
-				return nil, false, errors.Trace(err)
-			}
-
+			r.updateCheckpointAndStats(input.Checkpoint, input.Stats)
 			// Primary is stopped, but we still has secondary.
 			// Clear primary and promote secondary when it's prepared.
 			log.Info("schedulerv3: primary is stopped during Replicating",
@@ -1000,7 +991,7 @@ func (r *ReplicationSet) handleCaptureShutdown(
 
 func (r *ReplicationSet) updateCheckpointAndStats(
 	checkpoint tablepb.Checkpoint, stats tablepb.Stats,
-) error {
+) {
 	if checkpoint.ResolvedTs < checkpoint.CheckpointTs {
 		log.Warn("schedulerv3: resolved ts should not less than checkpoint ts",
 			zap.String("namespace", r.Changefeed.Namespace),
@@ -1028,11 +1019,8 @@ func (r *ReplicationSet) updateCheckpointAndStats(
 			zap.Any("replicationSet", r),
 			zap.Any("checkpointTs", r.Checkpoint.CheckpointTs),
 			zap.Any("resolvedTs", r.Checkpoint.ResolvedTs))
-		return errors.ErrInvalidCheckpointTs.GenWithStackByArgs(r.Checkpoint.CheckpointTs,
-			r.Checkpoint.ResolvedTs)
 	}
 	r.Stats = stats
-	return nil
 }
 
 // SetHeap is a max-heap, it implements heap.Interface.

cdc/scheduler/internal/v3/replication/replication_set_test.go

@@ -1455,3 +1455,29 @@ func TestReplicationSetHeap_MinK(t *testing.T) {
 	require.Equal(t, expectedTables, tables)
 	require.Equal(t, 0, h.Len())
 }
+
+func TestUpdateCheckpointAndStats(t *testing.T) {
+	cases := []struct {
+		checkpoint tablepb.Checkpoint
+		stats      tablepb.Stats
+	}{
+		{
+			checkpoint: tablepb.Checkpoint{
+				CheckpointTs: 1,
+				ResolvedTs:   2,
+			},
+			stats: tablepb.Stats{},
+		},
+		{
+			checkpoint: tablepb.Checkpoint{
+				CheckpointTs: 2,
+				ResolvedTs:   1,
+			},
+			stats: tablepb.Stats{},
+		},
+	}
+	r := &ReplicationSet{}
+	for _, c := range cases {
+		r.updateCheckpointAndStats(c.checkpoint, c.stats)
+	}
+}