This is an automated cherry-pick of #8264

close #8263

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

pkg/schedule/operator/operator.go

@@ -311,10 +311,11 @@ func (o *Operator) Check(region *core.RegionInfo) OpStep {
 	defer func() { _ = o.CheckTimeout() }()
 	for step := atomic.LoadInt32(&o.currentStep); int(step) < len(o.steps); step++ {
 		if o.steps[int(step)].IsFinish(region) {
-			if atomic.CompareAndSwapInt64(&(o.stepsTime[step]), 0, time.Now().UnixNano()) {
+			current := time.Now()
+			if atomic.CompareAndSwapInt64(&(o.stepsTime[step]), 0, current.UnixNano()) {
 				startTime, _ := o.getCurrentTimeAndStep()
 				operatorStepDuration.WithLabelValues(reflect.TypeOf(o.steps[int(step)]).Name()).
-					Observe(time.Unix(0, o.stepsTime[step]).Sub(startTime).Seconds())
+					Observe(current.Sub(startTime).Seconds())
 			}
 			atomic.StoreInt32(&o.currentStep, step+1)
 		} else {

pkg/schedule/operator/operator_test.go

@@ -17,7 +17,11 @@ package operator
 import (
 	"context"
 	"encoding/json"
+<<<<<<< HEAD
 	"fmt"
+=======
+	"sync"
+>>>>>>> e767c012f (schedule: fix datarace in `operator.check` (#8264))
 	"sync/atomic"
 	"testing"
 	"time"
@@ -529,3 +533,27 @@ func (suite *operatorTestSuite) TestRecord() {
 	suite.Equal(now, ob.FinishTime)
 	suite.Greater(ob.duration.Seconds(), time.Second.Seconds())
 }
+
+func TestOperatorCheckConcurrently(t *testing.T) {
+	re := require.New(t)
+	region := newTestRegion(1, 1, [2]uint64{1, 1}, [2]uint64{2, 2})
+	// addPeer1, transferLeader1, removePeer3
+	steps := []OpStep{
+		AddPeer{ToStore: 1, PeerID: 1},
+		TransferLeader{FromStore: 3, ToStore: 1},
+		RemovePeer{FromStore: 3},
+	}
+	op := NewTestOperator(1, &metapb.RegionEpoch{}, OpAdmin|OpLeader|OpRegion, steps...)
+	re.Equal(constant.Urgent, op.GetPriorityLevel())
+	checkSteps(re, op, steps)
+	op.Start()
+	var wg sync.WaitGroup
+	for i := 0; i < 10; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			re.Nil(op.Check(region))
+		}()
+	}
+	wg.Wait()
+}

pkg/schedule/operator_controller.go

@@ -417,10 +417,17 @@ func (oc *OperatorController) checkAddOperator(isPromoting bool, ops ...*operato
 			operatorWaitCounter.WithLabelValues(op.Desc(), "unexpected-status").Inc()
 			return false
 		}
+<<<<<<< HEAD:pkg/schedule/operator_controller.go
 		if !isPromoting && oc.wopStatus.ops[op.Desc()] >= oc.cluster.GetOpts().GetSchedulerMaxWaitingOperator() {
 			log.Debug("exceed max return false", zap.Uint64("waiting", oc.wopStatus.ops[op.Desc()]), zap.String("desc", op.Desc()), zap.Uint64("max", oc.cluster.GetOpts().GetSchedulerMaxWaitingOperator()))
 			operatorWaitCounter.WithLabelValues(op.Desc(), "exceed-max").Inc()
 			return false
+=======
+		if !isPromoting && oc.wopStatus.getCount(op.Desc()) >= oc.config.GetSchedulerMaxWaitingOperator() {
+			log.Debug("exceed max return false", zap.Uint64("waiting", oc.wopStatus.getCount(op.Desc())), zap.String("desc", op.Desc()), zap.Uint64("max", oc.config.GetSchedulerMaxWaitingOperator()))
+			operatorCounter.WithLabelValues(op.Desc(), "exceed-max-waiting").Inc()
+			return false, ExceedWaitLimit
+>>>>>>> e767c012f (schedule: fix datarace in `operator.check` (#8264)):pkg/schedule/operator/operator_controller.go
 		}
 
 		if op.SchedulerKind() == operator.OpAdmin || op.IsLeaveJointStateOperator() {