背景

网络异常、主机异常等情况下强制切主容易发生集群分裂,如下图

本文分享如何处理集群分裂异常。
需要注意的是，本文通过重新初始化异常库的方式解决问题，是高危操作，仅供参考。

处理步骤

1.在正常的主库导出数据，把导出的文件拷贝到异常库

BACKUP DATABASE BACKUPSET '/opt/dmdbms/data/DAMENG/bak/BACKUP_FILE_3';

2.停止异常库,停止watcher 和 server

/data01/dmdbms/bin/DmServiceDMSERVER stop
/data01/dmdbms/bin/DmWatcherServiceWatcher stop

3.异常库恢复数据

/data01/dmdbms/bin/dmrman CTLSTMT="RESTORE DATABASE '/data01/dmdata/DAMENG/dm.ini' FROM BACKUPSET '/data01/dm_bak/BACKUP_FILE_3'"

/data01/dmdbms/bin/dmrman CTLSTMT="RECOVER DATABASE '/data01/dmdata/DAMENG/dm.ini' FROM BACKUPSET '/data01/dm_bak/BACKUP_FILE_3'"

/data01/dmdbms/bin/dmrman CTLSTMT="RECOVER DATABASE '/data01/dmdata/DAMENG/dm.ini' UPDATE DB_MAGIC"

4.删除 dmwatcher.ctl

cd /data01/dmdata/DAMENG
mv dmwatcher.ctl dmwatcher.ctl_bak

5.启动异常库

/data01/dmdbms/bin/DmServiceDMSERVER start

6.异常库执行SQL，让异常库处于standby 状态

ALTER DATABASE STANDBY;

7.异常库启动watcher

/data01/dmdbms/bin/DmServiceDMSERVER start
/data01/dmdbms/bin/DmWatcherServiceWatcher start