From 002b454d186925821cd20330ac654d6381e76c30 Mon Sep 17 00:00:00 2001
From: Ti Chi Robot <ti-community-prow-bot@tidb.io>
Date: Tue, 10 Sep 2024 15:55:14 +0800
Subject: [PATCH] schedule: fix the down peer cannot be repaired (#7996)
 (#8196)

close tikv/pd#7808

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

Co-authored-by: Ryan Leung <rleungx@gmail.com>
Co-authored-by: ShuNing <nolouch@gmail.com>
Co-authored-by: ti-chi-bot[bot] <108142056+ti-chi-bot[bot]@users.noreply.github.com>
---
 pkg/schedule/checker/replica_strategy.go  |  7 ++-
 pkg/schedule/checker/rule_checker_test.go | 60 +++++++++++++++++++++++
 2 files changed, 66 insertions(+), 1 deletion(-)

diff --git a/pkg/schedule/checker/replica_strategy.go b/pkg/schedule/checker/replica_strategy.go
index 8b9b055f0a7..b8037612531 100644
--- a/pkg/schedule/checker/replica_strategy.go
+++ b/pkg/schedule/checker/replica_strategy.go
@@ -95,7 +95,12 @@ func (s *ReplicaStrategy) SelectStoreToAdd(coLocationStores []*core.StoreInfo, e
 func (s *ReplicaStrategy) SelectStoreToFix(coLocationStores []*core.StoreInfo, old uint64) (uint64, bool) {
 	// trick to avoid creating a slice with `old` removed.
 	s.swapStoreToFirst(coLocationStores, old)
-	return s.SelectStoreToAdd(coLocationStores[1:])
+	// If the coLocationStores only has one store, no need to remove.
+	// Otherwise, the other stores will be filtered.
+	if len(coLocationStores) > 1 {
+		coLocationStores = coLocationStores[1:]
+	}
+	return s.SelectStoreToAdd(coLocationStores)
 }
 
 // SelectStoreToImprove returns a store to replace oldStore. The location
diff --git a/pkg/schedule/checker/rule_checker_test.go b/pkg/schedule/checker/rule_checker_test.go
index d4d37de2c3c..f3ef8ed8d23 100644
--- a/pkg/schedule/checker/rule_checker_test.go
+++ b/pkg/schedule/checker/rule_checker_test.go
@@ -2061,3 +2061,63 @@ func (suite *ruleCheckerTestSuite) TestRemoveOrphanPeer() {
 	suite.NotNil(op)
 	suite.Equal("remove-orphan-peer", op.Desc())
 }
+
+func (suite *ruleCheckerTestSuite) TestIssue7808() {
+	re := suite.Require()
+	suite.cluster.AddLabelsStore(1, 1, map[string]string{"host": "host1", "disk_type": "mix"})
+	suite.cluster.AddLabelsStore(2, 1, map[string]string{"host": "host2", "disk_type": "mix"})
+	suite.cluster.AddLabelsStore(3, 1, map[string]string{"host": "host3", "disk_type": "ssd"})
+	suite.cluster.AddLabelsStore(4, 1, map[string]string{"host": "host4", "disk_type": "ssd"})
+	suite.cluster.AddLabelsStore(5, 1, map[string]string{"host": "host5", "disk_type": "ssd"})
+	suite.cluster.AddLeaderRegionWithRange(1, "", "", 3, 4, 1)
+	err := suite.ruleManager.SetRules([]*placement.Rule{
+		{
+			GroupID: "pd",
+			ID:      "1",
+			Role:    placement.Voter,
+			Count:   2,
+			LabelConstraints: []placement.LabelConstraint{
+				{
+					Key: "disk_type",
+					Values: []string{
+						"ssd",
+					},
+					Op: placement.In,
+				},
+			},
+			LocationLabels: []string{"host"},
+			IsolationLevel: "host",
+		},
+		{
+			GroupID: "pd",
+			ID:      "2",
+			Role:    placement.Follower,
+			Count:   1,
+			LabelConstraints: []placement.LabelConstraint{
+				{
+					Key: "disk_type",
+					Values: []string{
+						"mix",
+					},
+					Op: placement.In,
+				},
+			},
+			LocationLabels: []string{"host"},
+			IsolationLevel: "host",
+		},
+	})
+	re.NoError(err)
+	err = suite.ruleManager.DeleteRule("pd", "default")
+	re.NoError(err)
+	suite.cluster.SetStoreDown(1)
+	region := suite.cluster.GetRegion(1)
+	downPeer := []*pdpb.PeerStats{
+		{Peer: region.GetStorePeer(1), DownSeconds: 6000},
+	}
+	region = region.Clone(core.WithDownPeers(downPeer))
+	suite.cluster.PutRegion(region)
+	op := suite.rc.Check(suite.cluster.GetRegion(1))
+	re.NotNil(op)
+	re.Equal("fast-replace-rule-down-peer", op.Desc())
+	re.Contains(op.Brief(), "mv peer: store [1] to [2]")
+}