GitHub user RongtongJin added a comment to the discussion: controller通知broker变更失败，导致两个broker一直是master
> 嗯是的，但是broker-a和broker-b都把自己变更成了master ... ,在看板中也是两个broker都是master... 0.0 > > 这是broker-a的变更日志： > > > 2023-06-08 09:26:06 INFO ReplicasManager_ExecutorService_2 - Change broker > > [id:1][address:192.168.105.104:10911] to SYNC_MASTER, newMasterBrokerId:1, > > newMasterAddress:192.168.105.104:10911, newMasterEpoch:3, > > syncStateSetEpoch:2 > > > 2023-06-08 09:26:06 INFO main - Set PopReviveService Status to true > > 这是broker-b的变更日志： > > > 2023-06-08 09:26:03 INFO ReplicasManager_ExecutorService_2 - Change broker > > [id:1][address:192.168.105.105:10911] to SYNC_MASTER, newMasterBrokerId:1, > > newMasterAddress:192.168.105.105:10911, newMasterEpoch:3, > > syncStateSetEpoch:2 > > > 2023-06-08 09:26:03 INFO main - Set PopReviveService Status to true > > 以下分别是broker-a和broker-b的配置文件: > > > brokerClusterName=MyCluster > > brokerName=broker-a > > deleteWhen=04 > > fileReservedTime=48 > > flushDiskType=ASYNC_FLUSH > > brokerRole=SLAVE > > brokerId=-1 > > enableControllerMode=true > > controllerAddr=192.168.105.111:9877;192.168.105.112:9877;192.168.105.113:9877 > > allAckInSyncStateSet=true > > listenPort=10911 > > storePathRootDir=/tmp/rmqstore/node00 > > storePathCommitLog=/tmp/rmqstore/node00/commitlog > > > brokerClusterName=MyCluster > > brokerName=broker-b > > deleteWhen=04 > > fileReservedTime=48 > > flushDiskType=ASYNC_FLUSH > > brokerRole=SLAVE > > brokerId=-1 > > enableControllerMode=true > > controllerAddr=192.168.105.111:9877;192.168.105.112:9877;192.168.105.113:9877 > > allAckInSyncStateSet=true > > listenPort=10911 > > storePathRootDir=/tmp/rmqstore/node00 > > storePathCommitLog=/tmp/rmqstore/node00/commitlog > > 还有broker对namesrc和controller的连接也是好的，日志如下： > > > 2023-06-08 09:26:16 INFO brokerOutApi_thread_3 - Registering current broker > > to name server completed. TargetHost=192.168.105.103:9876 > > 2023-06-08 09:26:16 INFO brokerOutApi_thread_1 - Registering current broker > > to name server completed. TargetHost=192.168.105.101:9876 > > 2023-06-08 09:26:16 INFO brokerOutApi_thread_2 - Registering current broker > > to name server completed. TargetHost=192.168.105.102:9876 > > 2023-06-08 09:26:19 INFO ReplicasManager_ScheduledService_3 - Update > > controller leader address to 192.168.105.112:9877 > > 2023-06-08 09:26:29 INFO ReplicasManager_ScheduledService_1 - Update > > controller leader address to 192.168.105.112:9877 > > 2023-06-08 09:26:39 INFO ReplicasManager_ScheduledService_2 - Update > > controller leader address to 192.168.105.112:9877 > > controller集群状态也是好的，执行命令后如下： > > > sh mqadmin getControllerMetaData -a localhost:9877 > > #ControllerGroup group1 > > #ControllerLeaderId n1 > > #ControllerLeaderAddress 192.168.105.112:9877 > > #Peer: n0:192.168.105.111:9877 > > #Peer: n1:192.168.105.112:9877 > > #Peer: n2:192.168.105.113:9877 如果你需要一主一备，则需要使得两个broker配置文件中的brokerName是相同的。但从上面的描述看，两个broker的brokerName并不相同，因此会选出两个主来 GitHub link: https://github.com/apache/rocketmq/discussions/6869#discussioncomment-6144084 ---- This is an automatically sent email for dev@rocketmq.apache.org. To unsubscribe, please send an email to: dev-unsubscr...@rocketmq.apache.org
