schedule: fix datarace in operator.check (#8264) (#8578)

close #8263 Signed-off-by: husharp <[email protected]> Co-authored-by: husharp <[email protected]> Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
tikv · Sep 10, 2024 · 9fb2837 · 9fb2837
1 parent 002b454
commit 9fb2837
Show file tree

Hide file tree

Showing 2 changed files with 27 additions and 2 deletions.
diff --git a/pkg/schedule/operator/operator.go b/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 {

diff --git a/pkg/schedule/operator/operator_test.go b/pkg/schedule/operator/operator_test.go
@@ -18,6 +18,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"sync"
 	"sync/atomic"
 	"testing"
 	"time"
@@ -529,3 +530,26 @@ func (suite *operatorTestSuite) TestRecord() {
 	suite.Equal(now, ob.FinishTime)
 	suite.Greater(ob.duration.Seconds(), time.Second.Seconds())
 }
+
+func (suite *operatorTestSuite) TestOperatorCheckConcurrently() {
+	region := suite.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...)
+	suite.Equal(constant.Urgent, op.GetPriorityLevel())
+	suite.checkSteps(op, steps)
+	op.Start()
+	var wg sync.WaitGroup
+	for i := 0; i < 10; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			suite.Nil(op.Check(region))
+		}()
+	}
+	wg.Wait()
+}