rac下如何重建控制文件？

winkexp

这里我可能说错了一点，重建控制文件的时候可能只能建一个thread的logfile group，thread2的logfile要thread 1将数据库起来后再创建的吧~

你看下下面的语法，试着看看能否在重建控制文件的时候能把 1 和 2的online redolog一起建起来~我没试过~

SQL> alter database add logfile thread 2
2 group 4 ('/u01/oracle/oradata/racdb/redo04.log') size 10240K,
3 group 5 ('/u01/oracle/oradata/racdb/redo05.log') size 10240K,
4 group 6 ('/u01/oracle/oradata/racdb/redo06.log') size 10240k;

warehouse

最初由 winkexp 发布
[B]这里我可能说错了一点，重建控制文件的时候可能只能建一个thread的logfile group，thread2的logfile要thread 1将数据库起来后再创建的吧~

你看下下面的语法，试着看看能否在重建控制文件的时候能把 1 和 2的online redolog一起建起来~我没试过~

SQL> alter database add logfile thread 2
2 group 4 ('/u01/oracle/oradata/racdb/redo04.log') size 10240K,
3 group 5 ('/u01/oracle/oradata/racdb/redo05.log') size 10240K,
4 group 6 ('/u01/oracle/oradata/racdb/redo06.log') size 10240k; [/B]

好的，找时间我会试验的，不过一起建起来估计不行，试验之后通知你！

logzgh

举个例子：
STARTUP NOMOUNT
CREATE CONTROLFILE REUSE DATABASE "V816" RESETLOGS ARCHIVELOG
    MAXLOGFILES 12
    MAXLOGMEMBERS 5
    MAXDATAFILES 500
    MAXINSTANCES 3
    MAXLOGHISTORY 10210
LOGFILE
  GROUP 1 '/dev/vx/rdsk/usupportdg/V816_redo1_01.log'  SIZE 9M,
  GROUP 2 '/dev/vx/rdsk/usupportdg/V816_redo1_02.log'  SIZE 9M
DATAFILE
  '/dev/vx/rdsk/usupportdg/V816_system_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_rbs_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_data_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_index_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_temp_01.dbf'
;
ALTER DATABASE ADD LOGFILE THREAD 2
  GROUP 3 '/dev/vx/rdsk/usupportdg/V816_redo2_01.log' SIZE 9M,
  GROUP 4 '/dev/vx/rdsk/usupportdg/V816_redo2_02.log' SIZE 9M;
RECOVER DATABASE USING BACKUP CONTROLFILE UNTIL CANCEL;   ---这一步也不能省
CANCEL;
ALTER DATABASE OPEN RESETLOGS;
ALTER DATABASE ENABLE PUBLIC THREAD 2;

warehouse

最初由 logzgh 发布
[B]举个例子：
STARTUP NOMOUNT
CREATE CONTROLFILE REUSE DATABASE "V816" RESETLOGS ARCHIVELOG
    MAXLOGFILES 12
    MAXLOGMEMBERS 5
    MAXDATAFILES 500
    MAXINSTANCES 3
    MAXLOGHISTORY 10210
LOGFILE
  GROUP 1 '/dev/vx/rdsk/usupportdg/V816_redo1_01.log'  SIZE 9M,
  GROUP 2 '/dev/vx/rdsk/usupportdg/V816_redo1_02.log'  SIZE 9M
DATAFILE
  '/dev/vx/rdsk/usupportdg/V816_system_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_rbs_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_data_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_index_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_temp_01.dbf'
;
ALTER DATABASE ADD LOGFILE THREAD 2
  GROUP 3 '/dev/vx/rdsk/usupportdg/V816_redo2_01.log' SIZE 9M,
  GROUP 4 '/dev/vx/rdsk/usupportdg/V816_redo2_02.log' SIZE 9M;
RECOVER DATABASE USING BACKUP CONTROLFILE UNTIL CANCEL;   ---这一步也不能省
CANCEL;
ALTER DATABASE OPEN RESETLOGS;
ALTER DATABASE ENABLE PUBLIC THREAD 2; [/B]

谢谢，思路很清楚了，哥们，懂中医啊：）

logzgh

医术七窍通了六窍了。。。^_^

warehouse

最初由 wing hong 发布
[B]in  this case, you probably need to do an incomplete media recovery. just re-create controlfile probably won't help.

The steps to re-create controlfile in RAC :
1. shutdown all instances.
2. in the instance of thread 1 ,
startup nomount
create controlfile .....,  but here only list the redo log of the thread 1.
3. alter database add logfile thread 2 group .....
here you add the other thread
4. recover database until cancel using backup controlfile
cancel now or apply archived log , depends on your situation.
5. alter database open resetlogs;
6. alter database enable public thread 2 ;  here enable all the other thread. [/B]

这样做的话，ocm , gsd 还需要起来吗？

warehouse

最初由 logzgh 发布
[B]举个例子：
STARTUP NOMOUNT
CREATE CONTROLFILE REUSE DATABASE "V816" RESETLOGS ARCHIVELOG
    MAXLOGFILES 12
    MAXLOGMEMBERS 5
    MAXDATAFILES 500
    MAXINSTANCES 3
    MAXLOGHISTORY 10210
LOGFILE
  GROUP 1 '/dev/vx/rdsk/usupportdg/V816_redo1_01.log'  SIZE 9M,
  GROUP 2 '/dev/vx/rdsk/usupportdg/V816_redo1_02.log'  SIZE 9M
DATAFILE
  '/dev/vx/rdsk/usupportdg/V816_system_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_rbs_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_data_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_index_01.dbf',
  '/dev/vx/rdsk/usupportdg/V816_temp_01.dbf'
;
ALTER DATABASE ADD LOGFILE THREAD 2
  GROUP 3 '/dev/vx/rdsk/usupportdg/V816_redo2_01.log' SIZE 9M,
  GROUP 4 '/dev/vx/rdsk/usupportdg/V816_redo2_02.log' SIZE 9M;
RECOVER DATABASE USING BACKUP CONTROLFILE UNTIL CANCEL;   ---这一步也不能省
CANCEL;
ALTER DATABASE OPEN RESETLOGS;
ALTER DATABASE ENABLE PUBLIC THREAD 2; [/B]

错误如图：

wing hong

最初由 warehouse 发布
[B]

这样做的话，ocm , gsd 还需要起来吗？ [/B]

ocm is an prerequisive part of RAC, so it must be started and running first.

gsd is not an integral part of RAC, so it doesn't matter it is running or not

wing hong

最初由 warehouse 发布
[B]

错误如图： [/B]

please refer to my first post.

you may think about this in this way,

forget about the RAC, just in the single instance case of Oracle,

if your current and active redo log is corrupt , what should you do ?  

can you solve the problem by just re-creating the control file ?

logzgh

最初由 warehouse 发布
[B]

错误如图： [/B]

我给的例子只是创建控制文件的例子，无法解决你这个问题。

按在线日志损坏来处理吧。