Move the remove dispatcher logic to op controller (pingcap#369)

* update some filed name

* move the remove logic to operator_controller.go

maintainer/maintainer.go

@@ -48,11 +48,17 @@ import (
 	"go.uber.org/zap"
 )
 
-// Maintainer is response for handle changefeed replication tasks, Maintainer should:
-// 1. schedules tables to ticdc watcher
+// Maintainer is response for handle changefeed replication tasksMaintainer should:
+// 1. schedules tables to dispatcher manager
 // 2. calculate changefeed checkpoint ts
 // 3. send changefeed status to coordinator
 // 4. handle heartbeat reported by dispatcher
+// there are four threads in maintainer:
+// 1. controller thread , handled in dynstream, it handles the main logic of the maintainer, like barrier, heartbeat
+// 2. scheduler thread, handled in threadpool, it schedules the tables to dispatcher manager
+// 3. operator controller thread, handled in threadpool, it runs the operators
+// 4. checker controller, handled in threadpool, it runs the checkers to dynamically adjust the schedule
+// all threads are read/write information from/to the ReplicationDB
 type Maintainer struct {
 	id         model.ChangeFeedID
 	config     *config.ChangeFeedInfo

maintainer/maintainer_controller.go

@@ -39,6 +39,7 @@ import (
 )
 
 // Controller schedules and balance tables
+// there are 3 main components in the controller, scheduler, ReplicationDB and operator controller
 type Controller struct {
 	//  initialTables hold all tables that before controller bootstrapped
 	initialTables []commonEvent.Table
@@ -93,6 +94,37 @@ func NewController(changefeedID string,
 	return s
 }
 
+// HandleStatus handle the status report from the node
+func (c *Controller) HandleStatus(from node.ID, statusList []*heartbeatpb.TableSpanStatus) {
+	for _, status := range statusList {
+		dispatcherID := common.NewDispatcherIDFromPB(status.ID)
+		stm := c.GetTask(dispatcherID)
+		if stm == nil {
+			log.Warn("no span found, ignore",
+				zap.String("changefeed", c.changefeedID),
+				zap.String("from", from.String()),
+				zap.Any("status", status),
+				zap.String("span", dispatcherID.String()))
+			if status.ComponentStatus == heartbeatpb.ComponentState_Working {
+				// if the span is not found, and the status is working, we need to remove it from dispatcher
+				_ = c.messageCenter.SendCommand(replica.NewRemoveInferiorMessage(from, c.changefeedID, status.ID))
+			}
+			continue
+		}
+		c.operatorController.UpdateOperatorStatus(dispatcherID, from, status)
+		nodeID := stm.GetNodeID()
+		if nodeID != from {
+			// todo: handle the case that the node id is mismatch
+			log.Warn("node id not match",
+				zap.String("changefeed", c.changefeedID),
+				zap.Any("from", from),
+				zap.Stringer("node", nodeID))
+			continue
+		}
+		stm.UpdateStatus(status)
+	}
+}
+
 func (c *Controller) GetTasksBySchemaID(schemaID int64) []*replica.SpanReplication {
 	return c.replicationDB.GetTasksBySchemaID(schemaID)
 }
@@ -207,25 +239,22 @@ func (c *Controller) GetTask(dispatcherID common.DispatcherID) *replica.SpanRepl
 	return c.replicationDB.GetTaskByID(dispatcherID)
 }
 
-// RemoveAllTasks remove all tasks, todo: move these 3 methods to the operator controller
+// RemoveAllTasks remove all tasks
 func (c *Controller) RemoveAllTasks() {
-	for _, replicaSet := range c.replicationDB.TryRemoveAll() {
-		c.operatorController.ReplicaSetRemoved(operator.NewRemoveDispatcherOperator(c.replicationDB, replicaSet))
-	}
+	c.operatorController.RemoveAllTasks()
 }
 
+// RemoveTasksBySchemaID remove all tasks by schema id
 func (c *Controller) RemoveTasksBySchemaID(schemaID int64) {
-	for _, replicaSet := range c.replicationDB.TryRemoveBySchemaID(schemaID) {
-		c.operatorController.ReplicaSetRemoved(operator.NewRemoveDispatcherOperator(c.replicationDB, replicaSet))
-	}
+	c.operatorController.RemoveTasksBySchemaID(schemaID)
 }
 
+// RemoveTasksByTableIDs remove all tasks by table id
 func (c *Controller) RemoveTasksByTableIDs(tables ...int64) {
-	for _, replicaSet := range c.replicationDB.TryRemoveByTableIDs(tables...) {
-		c.operatorController.ReplicaSetRemoved(operator.NewRemoveDispatcherOperator(c.replicationDB, replicaSet))
-	}
+	c.operatorController.RemoveTasksByTableIDs(tables...)
 }
 
+// GetTasksByTableIDs get all tasks by table id
 func (c *Controller) GetTasksByTableIDs(tableIDs ...int64) []*replica.SpanReplication {
 	return c.replicationDB.GetTasksByTableIDs(tableIDs...)
 }
@@ -236,6 +265,7 @@ func (c *Controller) UpdateSchemaID(tableID, newSchemaID int64) {
 	c.replicationDB.UpdateSchemaID(tableID, newSchemaID)
 }
 
+// RemoveNode is called when a node is removed
 func (c *Controller) RemoveNode(id node.ID) {
 	c.operatorController.OnNodeRemoved(id)
 }
@@ -245,36 +275,6 @@ func (c *Controller) ScheduleFinished() bool {
 	return c.replicationDB.GetAbsentSize() == 0 && c.operatorController.OperatorSize() == 0
 }
 
-func (c *Controller) HandleStatus(from node.ID, statusList []*heartbeatpb.TableSpanStatus) {
-	for _, status := range statusList {
-		dispatcherID := common.NewDispatcherIDFromPB(status.ID)
-		stm := c.GetTask(dispatcherID)
-		if stm == nil {
-			log.Warn("no span found, ignore",
-				zap.String("changefeed", c.changefeedID),
-				zap.String("from", from.String()),
-				zap.Any("status", status),
-				zap.String("span", dispatcherID.String()))
-			if status.ComponentStatus == heartbeatpb.ComponentState_Working {
-				// if the span is not found, and the status is working, we need to remove it from dispatcher
-				_ = c.messageCenter.SendCommand(replica.NewRemoveInferiorMessage(from, c.changefeedID, status.ID))
-			}
-			continue
-		}
-		c.operatorController.UpdateOperatorStatus(dispatcherID, from, status)
-		nodeID := stm.GetNodeID()
-		if nodeID != from {
-			// todo: handle the case that the node id is mismatch
-			log.Warn("node id not match",
-				zap.String("changefeed", c.changefeedID),
-				zap.Any("from", from),
-				zap.Stringer("node", nodeID))
-			continue
-		}
-		stm.UpdateStatus(status)
-	}
-}
-
 func (c *Controller) TaskSize() int {
 	return c.replicationDB.TaskSize()
 }

maintainer/operator/operator_add.go

@@ -94,6 +94,6 @@ func (m *AddDispatcherOperator) PostFinish() {
 }
 
 func (m *AddDispatcherOperator) String() string {
-	return fmt.Sprintf("add dispatcher operator: %s, dest:%s ",
+	return fmt.Sprintf("add dispatcher operator: %s, dest:%s",
 		m.replicaSet.ID, m.dest)
 }

maintainer/operator/operator_controller.go

@@ -27,6 +27,8 @@ import (
 	"go.uber.org/zap"
 )
 
+// Controller is the operator controller, it manages all operators.
+// And the Controller is responsible for the execution of the operator.
 type Controller struct {
 	changefeedID  string
 	replicationDB *replica.ReplicationDB
@@ -80,21 +82,35 @@ func (oc *Controller) Execute() time.Time {
 	}
 }
 
-// ReplicaSetRemoved if the replica set is removed,
-// the controller will remove the operator. add a new operator to the controller.
-func (oc *Controller) ReplicaSetRemoved(op *RemoveDispatcherOperator) {
+// RemoveAllTasks remove all tasks, and notify all operators to stop.
+// it is only called by the barrier when the changefeed is stopped.
+func (oc *Controller) RemoveAllTasks() {
 	oc.lock.Lock()
 	defer oc.lock.Unlock()
 
-	if old, ok := oc.operators[op.ID()]; ok {
-		log.Info("replica set is removed , replace the old one",
-			zap.String("changefeed", oc.changefeedID),
-			zap.String("replicaset", op.ID().String()),
-			zap.String("operator", op.String()))
-		old.OnTaskRemoved()
-		delete(oc.operators, op.ID())
+	for _, replicaSet := range oc.replicationDB.TryRemoveAll() {
+		oc.removeReplicaSet(NewRemoveDispatcherOperator(oc.replicationDB, replicaSet))
+	}
+}
+
+// RemoveTasksBySchemaID remove all tasks by schema id.
+// it is only by the barrier when the schema is dropped by ddl
+func (oc *Controller) RemoveTasksBySchemaID(schemaID int64) {
+	oc.lock.Lock()
+	defer oc.lock.Unlock()
+	for _, replicaSet := range oc.replicationDB.TryRemoveBySchemaID(schemaID) {
+		oc.removeReplicaSet(NewRemoveDispatcherOperator(oc.replicationDB, replicaSet))
+	}
+}
+
+// RemoveTasksByTableIDs remove all tasks by table ids.
+// it is only called by the barrier when the table is dropped by ddl
+func (oc *Controller) RemoveTasksByTableIDs(tables ...int64) {
+	oc.lock.Lock()
+	defer oc.lock.Unlock()
+	for _, replicaSet := range oc.replicationDB.TryRemoveByTableIDs(tables...) {
+		oc.removeReplicaSet(NewRemoveDispatcherOperator(oc.replicationDB, replicaSet))
 	}
-	oc.operators[op.ID()] = op
 }
 
 // AddOperator adds an operator to the controller, if the operator already exists, return false.
@@ -110,7 +126,7 @@ func (oc *Controller) AddOperator(op Operator) bool {
 	}
 	span := oc.replicationDB.GetTaskByID(op.ID())
 	if span == nil {
-		log.Warn("span not found",
+		log.Warn("add operator failed, span not found",
 			zap.String("changefeed", oc.changefeedID),
 			zap.String("operator", op.String()))
 		return false
@@ -198,3 +214,17 @@ func (oc *Controller) pollQueueingOperator() (Operator, bool) {
 	heap.Push(&oc.runningQueue, item)
 	return op, true
 }
+
+// ReplicaSetRemoved if the replica set is removed,
+// the controller will remove the operator. add a new operator to the controller.
+func (oc *Controller) removeReplicaSet(op *RemoveDispatcherOperator) {
+	if old, ok := oc.operators[op.ID()]; ok {
+		log.Info("replica set is removed , replace the old one",
+			zap.String("changefeed", oc.changefeedID),
+			zap.String("replicaset", op.ID().String()),
+			zap.String("operator", op.String()))
+		old.OnTaskRemoved()
+		delete(oc.operators, op.ID())
+	}
+	oc.operators[op.ID()] = op
+}

maintainer/operator/operator_move.go

@@ -154,6 +154,6 @@ func (m *MoveDispatcherOperator) String() string {
 	m.lck.Lock()
 	defer m.lck.Unlock()
 
-	return fmt.Sprintf("move dispatcher operator: %s, origin:%s, dest:%s ",
+	return fmt.Sprintf("move dispatcher operator: %s, origin:%s, dest:%s",
 		m.replicaSet.ID, m.origin, m.dest)
 }

maintainer/operator/operator_split.go

@@ -83,6 +83,6 @@ func (m *SplitDispatcherOperator) Start() {
 
 func (m *SplitDispatcherOperator) String() string {
 	// todo add split region span
-	return fmt.Sprintf("move dispatcher operator: %s, dest:%v ",
+	return fmt.Sprintf("move dispatcher operator: %s, splitSpans:%v",
 		m.replicaSet.ID, m.splitSpans)
 }

maintainer/replica/span_replication.go

@@ -92,9 +92,8 @@ func NewWorkingReplicaSet(
 	return r
 }
 
-func (r *SpanReplication) UpdateStatus(status any) {
-	if status != nil {
-		newStatus := status.(*heartbeatpb.TableSpanStatus)
+func (r *SpanReplication) UpdateStatus(newStatus *heartbeatpb.TableSpanStatus) {
+	if newStatus != nil {
 		if newStatus.CheckpointTs >= r.status.CheckpointTs {
 			r.status = newStatus
 		}
@@ -124,12 +123,8 @@ func (r *SpanReplication) NewAddInferiorMessage(server node.ID) *messaging.Targe
 			ChangefeedID: r.ChangefeedID.ID,
 			Config: &heartbeatpb.DispatcherConfig{
 				DispatcherID: r.ID.ToPB(),
-				Span: &heartbeatpb.TableSpan{
-					TableID:  r.Span.TableID,
-					StartKey: r.Span.StartKey,
-					EndKey:   r.Span.EndKey,
-				},
-				StartTs: r.status.CheckpointTs,
+				Span:         r.Span,
+				StartTs:      r.status.CheckpointTs,
 			},
 			ScheduleAction: heartbeatpb.ScheduleAction_Create,
 		})