placement: do not delete orphan peers if some peers selected by RuleFit is down or pending (#4067)

server/schedule/checker/rule_checker.go

@@ -269,12 +269,27 @@ func (c *RuleChecker) fixOrphanPeers(region *core.RegionInfo, fit *placement.Reg
 	if len(fit.OrphanPeers) == 0 {
 		return nil, nil
 	}
-	// remove orphan peers only when all rules are satisfied (count+role)
+	// remove orphan peers only when all rules are satisfied (count+role) and all peers selected
+	// by RuleFits is not pending or down.
 	for _, rf := range fit.RuleFits {
 		if !rf.IsSatisfied() {
 			checkerCounter.WithLabelValues("rule_checker", "skip-remove-orphan-peer").Inc()
 			return nil, nil
 		}
+		for _, p := range rf.Peers {
+			for _, pendingPeer := range region.GetPendingPeers() {
+				if pendingPeer.Id == p.Id {
+					checkerCounter.WithLabelValues("rule_checker", "skip-remove-orphan-peer").Inc()
+					return nil, nil
+				}
+			}
+			for _, downPeer := range region.GetDownPeers() {
+				if downPeer.Peer.Id == p.Id {
+					checkerCounter.WithLabelValues("rule_checker", "skip-remove-orphan-peer").Inc()
+					return nil, nil
+				}
+			}
+		}
 	}
 	checkerCounter.WithLabelValues("rule_checker", "remove-orphan-peer").Inc()
 	peer := fit.OrphanPeers[0]

server/schedule/checker/rule_checker_test.go

@@ -600,3 +600,45 @@ func (s *testRuleCheckerSuite) TestFixOfflinePeer(c *C) {
 	s.ruleManager.SetRule(rule)
 	c.Assert(s.rc.Check(region), IsNil)
 }
+
+// Ref https://github.com/tikv/pd/issues/4045
+func (s *testRuleCheckerSuite) TestSkipFixOrphanPeerIfSelectedPeerisPendingOrDown(c *C) {
+	s.cluster.AddLabelsStore(1, 1, map[string]string{"host": "host1"})
+	s.cluster.AddLabelsStore(2, 1, map[string]string{"host": "host1"})
+	s.cluster.AddLabelsStore(3, 1, map[string]string{"host": "host2"})
+	s.cluster.AddLabelsStore(4, 1, map[string]string{"host": "host4"})
+	s.cluster.AddLeaderRegionWithRange(1, "", "", 1, 2, 3, 4)
+
+	// set peer3 and peer4 to pending
+	r1 := s.cluster.GetRegion(1)
+	r1 = r1.Clone(core.WithPendingPeers([]*metapb.Peer{r1.GetStorePeer(3), r1.GetStorePeer(4)}))
+	s.cluster.PutRegion(r1)
+
+	// should not remove extra peer
+	op := s.rc.Check(s.cluster.GetRegion(1))
+	c.Assert(op, IsNil)
+
+	// set peer3 to down-peer
+	r1 = r1.Clone(core.WithPendingPeers([]*metapb.Peer{r1.GetStorePeer(4)}))
+	r1 = r1.Clone(core.WithDownPeers([]*pdpb.PeerStats{
+		{
+			Peer:        r1.GetStorePeer(3),
+			DownSeconds: 42,
+		},
+	}))
+	s.cluster.PutRegion(r1)
+
+	// should not remove extra peer
+	op = s.rc.Check(s.cluster.GetRegion(1))
+	c.Assert(op, IsNil)
+
+	// set peer3 to normal
+	r1 = r1.Clone(core.WithDownPeers(nil))
+	s.cluster.PutRegion(r1)
+
+	// should remove extra peer now
+	var remove operator.RemovePeer
+	op = s.rc.Check(s.cluster.GetRegion(1))
+	c.Assert(op.Step(0), FitsTypeOf, remove)
+	c.Assert(op.Desc(), Equals, "remove-orphan-peer")
+}