dgbroker配置Fast-Start Failover

环境:
OS:Centos 7
DB:11.2.0.4
架构:1主2从
slnngk->slavea
slnngk->slaveb

1.主库开启flashback on(若已经开启了，则省略)
SQL> select flashback_on from v$database;

FLASHBACK_ON
------------------
NO

SQL> alter database flashback on;

Database altered.

2.备库开启flashback on(备库slavea和slaveb都做同样的操作,若已经开启了，则省略)
这里在主库上登录dgmgrl
备库slavea:
在备库开启flashback之前，要先将备库的intended state的状态更改为apply-off.

DGMGRL> edit database slavea set state='apply-off';
Succeeded.

DGMGRL> show database slavea;

Database - slavea

Role: PHYSICAL STANDBY
Intended State: APPLY-OFF
Transport Lag: 0 seconds (computed 1 second ago)
Apply Lag: 0 seconds (computed 1 second ago)
Apply Rate: (unknown)
Real Time Query: OFF
Instance(s):
slavea

Database Status:
SUCCESS

sqlplus登录到备库slavea执行如下命令:
SQL> alter database flashback on;

Database altered.

打开flashback之后,要将standby数据库的intended state的状态更改为apply-on

DGMGRL> edit database slavea set state='apply-on';
Succeeded.

DGMGRL> show database slavea;

Database - slavea

Role: PHYSICAL STANDBY
Intended State: APPLY-ON
Transport Lag: 0 seconds (computed 1 second ago)
Apply Lag: 0 seconds (computed 1 second ago)
Apply Rate: (unknown)
Real Time Query: ON
Instance(s):
slavea

Database Status:
SUCCESS

备库slaveb
DGMGRL>edit database slaveb set state='apply-off';
SQL> alter database flashback on;
DGMGRL> edit database slaveb set state='apply-on';

3.配置和启用observer
官方建议observer应该与primary和standby数据库隔离,不能在一台机器上,我这里资源有限,将observer部署在主库上面
[oracle@dbmaster ~]$ pwd
/home/oracle
[oracle@dbmaster ~]$ dgmgrl -logfile /tmp/observer.log sys/oracle@tnsslnngk "start observer" &

[oracle@dbmaster ~]$ ps -ef|grep start
oracle 27400 4049 0 03:59 pts/0 00:00:00 dgmgrl start observer file=FSFO.dat

查看启动日志
[oracle@dbmaster ~]$ more fsfo.log
nohup: ignoring input
DGMGRL for Linux: Version 11.2.0.4.0 - 64bit Production

Welcome to DGMGRL, type "help" for information.
Connected.
Observer started

或是写成脚本使用nohup后台运行

[oracle@dbmaster scripts]$ more start_ob.sh
#!/bin/bash
. /etc/profile
. ~/.bash_profile
dgmgrl -logfile /tmp/observer.log sys/oracle@tnsslnngk "start observer" &

4.启用failover(主库上登录)
查看fast_start failover的状态
[oracle@dbmaster ~]$ dgmgrl
DGMGRL> connect sys/oracle

DGMGRL> show fast_start failover

Fast-Start Failover: DISABLED

  Threshold:          30 seconds
  Target:             (none)
  Observer:           dbmaster
  Lag Limit:          30 seconds
  Shutdown Primary:   TRUE
  Auto-reinstate:     TRUE
  Observer Reconnect: (none)
  Observer Override:  FALSE

Configurable Failover Conditions
  Health Conditions:
    Corrupted Controlfile          YES
    Corrupted Dictionary           YES
    Inaccessible Logfile            NO
    Stuck Archiver                  NO
    Datafile Offline               YES

  Oracle Error Conditions:
    (none)

启用
DGMGRL> enable fast_start failover
Error: ORA-16651: requirements not met for enabling fast-start failover

Failed.

解决办法:配置同步模式(若已经是 ASYNC 模式就不需要修改了)
EDIT DATABASE slnngk SET PROPERTY LogXptMode='async';
EDIT DATABASE slavea SET PROPERTY LogXptMode='async';
EDIT DATABASE slaveb SET PROPERTY LogXptMode='async';
如果主备的设置为最大高可用保护模式,则需要设置LogXptMode为sync
如果主备的设置为最大性能保护模式,则需要设置LogXptMode为async
我这里是最大性能模式

指定故障转移目的,我们这里将slavea作为故障目的,这里只能设置一个目的,两个的话好像不行FastStartFailoverTarget=slavea,slaveb
DGMGRL> edit database slnngk set property FastStartFailoverTarget=slavea;
DGMGRL> enable fast_start failover
Enabled.

查看

DGMGRL> show fast_start failover

Fast-Start Failover: ENABLED

  Threshold:          30 seconds
  Target:             slavea
  Observer:           dbmaster
  Lag Limit:          30 seconds
  Shutdown Primary:   TRUE
  Auto-reinstate:     TRUE
  Observer Reconnect: (none)
  Observer Override:  FALSE

Configurable Failover Conditions
  Health Conditions:
    Corrupted Controlfile          YES
    Corrupted Dictionary           YES
    Inaccessible Logfile            NO
    Stuck Archiver                  NO
    Datafile Offline               YES

  Oracle Error Conditions:
    (none)

5.要是observer没有启动的话,会提示如下的错误信息

DGMGRL> show configuration;

Configuration - slnngktest

  Protection Mode: MaxPerformance
  Databases:
    slnngk - Primary database
      Error: ORA-16820: fast-start failover observer is no longer observing this database

    slavea - (*) Physical standby database
      Error: ORA-16820: fast-start failover observer is no longer observing this database

    slaveb - Physical standby database

Fast-Start Failover: ENABLED

Configuration Status:
ERROR

后台启动observer
[oracle@dbmaster ~]$ dgmgrl -logfile /tmp/observer.log sys/oracle@tnsslnngk "start observer" &

可以写个脚本做成自启动的方式

再次查看数据库状态

DGMGRL> show configuration;

Configuration - slnngktest

  Protection Mode: MaxPerformance
  Databases:
    slnngk - Primary database
    slavea - (*) Physical standby database
    slaveb - Physical standby database

Fast-Start Failover: ENABLED

Configuration Status:
SUCCESS

可以做成脚本开机自动执行:

[oracle@dbmaster scripts]$ more start_ob.sh
#!/bin/bash
dgmgrl -logfile /tmp/observer.log sys/oracle@tnsslnngk "start observer" &

后台运行
nohup ./start_ob.sh > start_ob.out 2>&1 &

或是将脚本添加到启动文件,先启动监听,然后启动数据库,最后启动observer
[root@dbmaster ~]# more /etc/rc.d/rc.local
su - oracle -lc "/u01/app/oracle/product/11.2.0.4/db_1/bin/lsnrctl start"
su - oracle -lc "/u01/app/oracle/product/11.2.0.4/db_1/bin/dbstart"
su - oracle -c "/home/oracle/scripts/start_ob.sh"

遇到的问题:
[oracle@dbmaster scripts]$ more fsfo.log
nohup: ignoring input
DGMGRL for Linux: Version 11.2.0.4.0 - 64bit Production

Welcome to DGMGRL, type "help" for information.
Connected.
Error: ORA-16647: could not start more than one observer

解决办法:
stop observer