schedule: fix location label in placement rule not work well with rul…

…e checker (#6633)

close #6637

Signed-off-by: Ryan Leung <rleungx@gmail.com>

pkg/schedule/checker/replica_strategy.go

@@ -93,6 +93,9 @@ func (s *ReplicaStrategy) SelectStoreToAdd(coLocationStores []*core.StoreInfo, e
 // SelectStoreToFix returns a store to replace down/offline old peer. The location
 // placement after scheduling is allowed to be worse than original.
 func (s *ReplicaStrategy) SelectStoreToFix(coLocationStores []*core.StoreInfo, old uint64) (uint64, bool) {
+	if len(coLocationStores) == 0 {
+		return 0, false
+	}
 	// trick to avoid creating a slice with `old` removed.
 	s.swapStoreToFirst(coLocationStores, old)
 	return s.SelectStoreToAdd(coLocationStores[1:])
@@ -101,6 +104,9 @@ func (s *ReplicaStrategy) SelectStoreToFix(coLocationStores []*core.StoreInfo, o
 // SelectStoreToImprove returns a store to replace oldStore. The location
 // placement after scheduling should be better than original.
 func (s *ReplicaStrategy) SelectStoreToImprove(coLocationStores []*core.StoreInfo, old uint64) (uint64, bool) {
+	if len(coLocationStores) == 0 {
+		return 0, false
+	}
 	// trick to avoid creating a slice with `old` removed.
 	s.swapStoreToFirst(coLocationStores, old)
 	oldStore := s.cluster.GetStore(old)

pkg/schedule/checker/rule_checker.go

@@ -391,19 +391,26 @@ func (c *RuleChecker) allowLeader(fit *placement.RegionFit, peer *metapb.Peer) b
 }
 
 func (c *RuleChecker) fixBetterLocation(region *core.RegionInfo, rf *placement.RuleFit) (*operator.Operator, error) {
-	if len(rf.Rule.LocationLabels) == 0 || rf.Rule.Count <= 1 {
+	if len(rf.Rule.LocationLabels) == 0 {
 		return nil, nil
 	}
 
 	isWitness := rf.Rule.IsWitness && c.isWitnessEnabled()
 	// If the peer to be moved is a witness, since no snapshot is needed, we also reuse the fast failover logic.
 	strategy := c.strategy(region, rf.Rule, isWitness)
-	ruleStores := c.getRuleFitStores(rf)
-	oldStore := strategy.SelectStoreToRemove(ruleStores)
+	regionStores := c.cluster.GetRegionStores(region)
+	oldStore := strategy.SelectStoreToRemove(regionStores)
 	if oldStore == 0 {
 		return nil, nil
 	}
-	newStore, filterByTempState := strategy.SelectStoreToImprove(ruleStores, oldStore)
+	var coLocationStores []*core.StoreInfo
+	for _, s := range regionStores {
+		if placement.MatchLabelConstraints(s, rf.Rule.LabelConstraints) {
+			coLocationStores = append(coLocationStores, s)
+		}
+	}
+
+	newStore, filterByTempState := strategy.SelectStoreToImprove(coLocationStores, oldStore)
 	if newStore == 0 {
 		log.Debug("no replacement store", zap.Uint64("region-id", region.GetID()))
 		c.handleFilterState(region, filterByTempState)

pkg/schedule/checker/rule_checker_test.go

@@ -1332,3 +1332,62 @@ func (suite *ruleCheckerTestSuite) TestPendingList() {
 	_, exist = suite.rc.pendingList.Get(1)
 	suite.False(exist)
 }
+
+func (suite *ruleCheckerTestSuite) TestLocationLabels() {
+	suite.cluster.AddLabelsStore(1, 1, map[string]string{"zone": "z1", "rack": "r1", "host": "h1"})
+	suite.cluster.AddLabelsStore(2, 1, map[string]string{"zone": "z1", "rack": "r1", "host": "h1"})
+	suite.cluster.AddLabelsStore(3, 1, map[string]string{"zone": "z1", "rack": "r2", "host": "h1"})
+	suite.cluster.AddLabelsStore(4, 1, map[string]string{"zone": "z1", "rack": "r2", "host": "h1"})
+	suite.cluster.AddLabelsStore(5, 1, map[string]string{"zone": "z2", "rack": "r3", "host": "h2"})
+	suite.cluster.AddLabelsStore(6, 1, map[string]string{"zone": "z2", "rack": "r3", "host": "h2"})
+	suite.cluster.AddLeaderRegionWithRange(1, "", "", 1, 2, 5)
+	rule1 := &placement.Rule{
+		GroupID: "pd",
+		ID:      "test1",
+		Role:    placement.Leader,
+		Count:   1,
+		LabelConstraints: []placement.LabelConstraint{
+			{
+				Key:    "zone",
+				Op:     placement.In,
+				Values: []string{"z1"},
+			},
+		},
+		LocationLabels: []string{"rack"},
+	}
+	rule2 := &placement.Rule{
+		GroupID: "pd",
+		ID:      "test2",
+		Role:    placement.Voter,
+		Count:   1,
+		LabelConstraints: []placement.LabelConstraint{
+			{
+				Key:    "zone",
+				Op:     placement.In,
+				Values: []string{"z1"},
+			},
+		},
+		LocationLabels: []string{"rack"},
+	}
+	rule3 := &placement.Rule{
+		GroupID: "pd",
+		ID:      "test3",
+		Role:    placement.Voter,
+		Count:   1,
+		LabelConstraints: []placement.LabelConstraint{
+			{
+				Key:    "zone",
+				Op:     placement.In,
+				Values: []string{"z2"},
+			},
+		},
+		LocationLabels: []string{"rack"},
+	}
+	suite.ruleManager.SetRule(rule1)
+	suite.ruleManager.SetRule(rule2)
+	suite.ruleManager.SetRule(rule3)
+	suite.ruleManager.DeleteRule("pd", "default")
+	op := suite.rc.Check(suite.cluster.GetRegion(1))
+	suite.NotNil(op)
+	suite.Equal("move-to-better-location", op.Desc())
+}