Merge pull request #13467 from chaochn47/fix_exclude_alarms

server/etcdserver/api/etcdhttp: exclude the same alarm type activated by multiple peers

CHANGELOG-3.4.md

@@ -5,12 +5,21 @@ Previous change logs can be found at [CHANGELOG-3.3](https://github.com/etcd-io/
 
 The minimum recommended etcd versions to run in **production** are 3.2.28+, 3.3.18+, 3.4.2+, and 3.5.1+.
 
-See [code changes](https://github.com/etcd-io/etcd/compare/v3.4.17...v3.4.18) and [v3.4 upgrade guide](https://etcd.io/docs/latest/upgrades/upgrade_3_4/) for any breaking changes.
+<hr>
+
+## v3.4.19 (TODO)
+
+See [code changes](https://github.com/etcd-io/etcd/compare/v3.4.18...v3.4.19) and [v3.4 upgrade guide](https://etcd.io/docs/latest/upgrades/upgrade_3_4/) for any breaking changes.
+
+### etcd server
+- Fix [exclude the same alarm type activated by multiple peers](https://github.com/etcd-io/etcd/pull/13475).
 
 <hr>
 
 ## v3.4.18 (2021-10-15)
 
+See [code changes](https://github.com/etcd-io/etcd/compare/v3.4.17...v3.4.18) and [v3.4 upgrade guide](https://etcd.io/docs/latest/upgrades/upgrade_3_4/) for any breaking changes.
+
 ### Metrics, Monitoring
 
 See [List of metrics](https://etcd.io/docs/latest/metrics/) for all metrics per release.

CHANGELOG-3.5.md

@@ -6,6 +6,15 @@ Previous change logs can be found at [CHANGELOG-3.4](https://github.com/etcd-io/
 The minimum recommended etcd versions to run in **production** are 3.2.28+, 3.3.18+, 3.4.2+, and 3.5.1+.
 
 
+<hr>
+
+## [v3.5.2](https://github.com/etcd-io/etcd/releases/tag/v3.5.2) (TODO)
+
+See [code changes](https://github.com/etcd-io/etcd/compare/v3.5.1...v3.5.2) and [v3.5 upgrade guide](https://etcd.io/docs/latest/upgrades/upgrade_3_5/) for any breaking changes.
+
+### etcd server
+- Fix [exclude the same alarm type activated by multiple peers](https://github.com/etcd-io/etcd/pull/13476).
+
 <hr>
 
 ## [v3.5.1](https://github.com/etcd-io/etcd/releases/tag/v3.5.1) (2021-10-15)

CHANGELOG-3.6.md

@@ -33,6 +33,7 @@ See [code changes](https://github.com/etcd-io/etcd/compare/v3.5.0...v3.6.0).
 
 - Add [`etcd --log-format`](https://github.com/etcd-io/etcd/pull/13339) flag to support log format.
 - Fix [non mutating requests pass through quotaKVServer when NOSPACE](https://github.com/etcd-io/etcd/pull/13435)
+- Fix [exclude the same alarm type activated by multiple peers](https://github.com/etcd-io/etcd/pull/13467).
 
 ### tools/benchmark
 

server/etcdserver/api/etcdhttp/metrics.go

@@ -138,8 +138,7 @@ func checkHealth(lg *zap.Logger, srv etcdserver.ServerV2, excludedAlarms AlarmSe
 		for _, v := range as {
 			alarmName := v.Alarm.String()
 			if _, found := excludedAlarms[alarmName]; found {
-				lg.Debug("/health excluded alarm", zap.String("alarm", alarmName))
+				lg.Debug("/health excluded alarm", zap.String("alarm", v.String()))
-				delete(excludedAlarms, alarmName)
 				continue
 			}
 
@@ -157,10 +156,6 @@ func checkHealth(lg *zap.Logger, srv etcdserver.ServerV2, excludedAlarms AlarmSe
 		}
 	}
 
-	if len(excludedAlarms) > 0 {
-		lg.Warn("fail exclude alarms from health check", zap.String("exclude alarms", fmt.Sprintf("%+v", excludedAlarms)))
-	}
-
 	if uint64(srv.Leader()) == raft.None {
 		h.Health = "false"
 		h.Reason = "RAFT NO LEADER"

server/etcdserver/api/etcdhttp/metrics_test.go

@@ -77,6 +77,12 @@ func TestHealthHandler(t *testing.T) {
 			http.StatusOK,
 			"true",
 		},
+		{
+			[]*pb.AlarmMember{{MemberID: uint64(1), Alarm: pb.AlarmType_NOSPACE}, {MemberID: uint64(2), Alarm: pb.AlarmType_NOSPACE}, {MemberID: uint64(3), Alarm: pb.AlarmType_NOSPACE}},
+			"/health?exclude=NOSPACE",
+			http.StatusOK,
+			"true",
+		},
 		{
 			[]*pb.AlarmMember{{MemberID: uint64(0), Alarm: pb.AlarmType_NOSPACE}, {MemberID: uint64(1), Alarm: pb.AlarmType_CORRUPT}},
 			"/health?exclude=NOSPACE",