RMAN备份与恢复之UNDO表空间丢失

一 UNDO表空间讲解

      

        在上一篇文章（RMAN备份与恢复之可脱机数据文件丢失）中，我们讲到可脱机数据文件丢失怎么处理，这篇文章我们讲解UNDO表空间丢失的解决办法。

 

        UNDO表空间用于存放UNDO数据，当执行DML操作（INSERT、UPDATE、DELETE）的时候，ORACLE会将这些操作的旧数据写入到UNDO段。UNDO数据也称为回滚数据，用于确保数据的一致性。作用包括：回退事、读一致性、事务恢复、闪回查询。9i开始，管理UNDO数据可以使用UNDO表空间，也可以使用回滚段。10g开始，ORACLE已经放弃使用回滚段。提到UNDO表空间，不得不提UNDO段。UNDO Segment分为两个部分，一个是UNDO Segment Head，还有一个是UNDO Segment Block（也称为事务槽)。UNDO Segment Head中包含了这个回滚段的事务信息，而且有一个指针指向Undo Segment Block。UNDO表空间是非常重要的，如果丢失，会出现无法对数据进行更新。平时的数据库管理中应该注意UNDO表空间的空间是否足够，采用自动扩展还是限制大小，undo_retention值的设定等等。

 

二 备份与恢复UNDO表空间讲解

 

        备份与恢复UNDO表空间，首先要有备份。使用RMAN备份完成后，我们模拟UNDO表空间丢失。此时做更新操作仍然成功，因为shared pool和buffer cache存放了更新的信息。如果我们刷新shared pool和buffer cache，再做连接用户或者更新操作，会提示数据文件找不到。因为UNDO表空间丢失，并且UNDO表空间不可脱机，所以我们不能在数据库运行状态下对UNDO表空间进行恢复。这就要求我们关闭数据库进行恢复操作。如果在真实环境中进行操作，务必在业务低峰期或者测试库进行操作。我们使用一致性关闭数据库会失败，只有强制关闭。此时参数文件、控制文件正常，只是数据文件不正常，所以我们能把数据库启动到MOUNT状态。启动到MOUNT状态后，我们需要使UNDO表空间数据文件离线，注意此时的数据文件编号。然后登录到RMAN中，还原UNDO表空间数据文件，实际上做了一个拷贝的操作，从备份文件中拷贝UNDO表空间数据文件到数据目录，待拷贝完成后，我们需要对UNDO表空间数据文件进行恢复。恢复完成后，再使UNDO表空间数据文件在线，此时的数据库是MOUNT状态，我们需要打开数据库。如果所有的操作都成功，就可以对数据进行更新。

 

三 模拟

Step 1，RMAN中备份全库

RMAN> BACKUP DATABASE;Starting backup at 12-DEC-13using channel ORA_DISK_1channel ORA_DISK_1: starting full datafile backup setchannel ORA_DISK_1: specifying datafile(s) in backup setinput datafile file number=00001 name=/u01/oracle/oradata/justdb/system01.dbfinput datafile file number=00002 name=/u01/oracle/oradata/justdb/sysaux01.dbfinput datafile file number=00003 name=/u01/oracle/oradata/justdb/undotbs01.dbfinput datafile file number=00004 name=/u01/oracle/oradata/justdb/users01.dbfchannel ORA_DISK_1: starting piece 1 at 12-DEC-13channel ORA_DISK_1: finished piece 1 at 12-DEC-13piece handle=/u01/oracle/fast_recovery_area/JUSTDB/backupset/2013_12_12/o1_mf_nnndf_TAG20131212T095816_9bl61rrn_.bkp tag=TAG20131212T095816 comment=NONEchannel ORA_DISK_1: backup set complete, elapsed time: 00:00:25channel ORA_DISK_1: starting full datafile backup setchannel ORA_DISK_1: specifying datafile(s) in backup setincluding current control file in backup setincluding current SPFILE in backup setchannel ORA_DISK_1: starting piece 1 at 12-DEC-13channel ORA_DISK_1: finished piece 1 at 12-DEC-13piece handle=/u01/oracle/fast_recovery_area/JUSTDB/backupset/2013_12_12/o1_mf_ncsnf_TAG20131212T095816_9bl62lw2_.bkp tag=TAG20131212T095816 comment=NONEchannel ORA_DISK_1: backup set complete, elapsed time: 00:00:01Finished backup at 12-DEC-13

Step 2，模拟UNDO表空间丢失

SQL> CONN / AS SYSDBAConnected.SQL> HO mv /u01/oracle/oradata/justdb/undotbs01.dbf /opt/learn/

Step 3，SQL Plus中连接到sys用户，刷新shared pool和buffer cache

SQL> CONN / AS SYSDBAConnected.SQL> CONN / AS SYSDBAConnected.SQL> ALTER SYSTEM FLUSH shared_pool;System altered.SQL> ALTER SYSTEM FLUSH buffer_cache;System altered.

Step 4，SQL Plus连接到scoot用户，发现报ORA-01110错误，数据文件不能找到

SQL> CONN SCOTT/tiger;ERROR:ORA-00604: error occurred at recursive SQL level 1ORA-01116: error in opening database file 3ORA-01110: data file 3: '/u01/oracle/oradata/justdb/undotbs01.dbf'ORA-27041: unable to open fileLinux-x86_64 Error: 2: No such file or directoryAdditional information: 3Warning: You are no longer connected to ORACLE.

Step 5，SQL Plus一致性关闭数据库，失败，只有强制关闭数据库

SQL> CONN / AS SYSDBACONN / AS SYSDBAConnected.SQL> SHUTDOWN IMMEDIATE;ORA-01116: error in opening database file 3ORA-01110: data file 3: '/u01/oracle/oradata/justdb/undotbs01.dbf'ORA-27041: unable to open fileLinux-x86_64 Error: 2: No such file or directoryAdditional information: 3SQL> SHUTDOWN ABORT;ORACLE instance shut down.

Step 6，再次登录到SQL Plus，启动数据库到MOUNT状态

[oracle@orcl ~]$ sqlplus [uniread] Loaded history (157 lines)SQL*Plus: Release 11.2.0.3.0 Production on Thu Dec 12 10:37:52 2013Copyright (c) 1982, 2011, Oracle.  All rights reserved.Connected to an idle instance.SQL> STARTUP MOUNT;ORACLE instance started.Total System Global Area 1269366784 bytesFixed Size        2227984 bytesVariable Size     754974960 bytesDatabase Buffers    503316480 bytesRedo Buffers        8847360 bytesDatabase mounted.

Step 7，SQL Plus中使3号文件（UNDO表空间）离线

SQL> ALTER DATABASE DATAFILE 3 OFFLINE;Database altered.

Step 8，使用sys用户登录到RMAN

[oracle@orcl ~]$ uniread rman target /[uniread] Loaded history (96 lines)Recovery Manager: Release 11.2.0.3.0 - Production on Thu Dec 12 10:38:26 2013Copyright (c) 1982, 2011, Oracle and/or its affiliates.  All rights reserved.connected to target database: JUSTDB (DBID=57321598, not open)RMAN> 

Step 9，RMAN中还原3号文件

RMAN> RESTORE DATAFILE 3;Starting restore at 12-DEC-13using target database control file instead of recovery catalogallocated channel: ORA_DISK_1channel ORA_DISK_1: SID=20 device type=DISKchannel ORA_DISK_1: starting datafile backup set restorechannel ORA_DISK_1: specifying datafile(s) to restore from backup setchannel ORA_DISK_1: restoring datafile 00003 to /u01/oracle/oradata/justdb/undotbs01.dbfchannel ORA_DISK_1: reading from backup piece /u01/oracle/fast_recovery_area/JUSTDB/backupset/2013_12_12/o1_mf_nnndf_TAG20131212T095816_9bl61rrn_.bkpchannel ORA_DISK_1: piece handle=/u01/oracle/fast_recovery_area/JUSTDB/backupset/2013_12_12/o1_mf_nnndf_TAG20131212T095816_9bl61rrn_.bkp tag=TAG20131212T095816channel ORA_DISK_1: restored backup piece 1channel ORA_DISK_1: restore complete, elapsed time: 00:00:01Finished restore at 12-DEC-13

Step 10，RMAN中恢复3号文件

RMAN> RECOVER DATAFILE 3;RECOVER DATAFILE 3;Starting recover at 12-DEC-13using channel ORA_DISK_1starting media recoverymedia recovery complete, elapsed time: 00:00:00Finished recover at 12-DEC-13

Step 11，SQL Plus中使3号数据文件在线

SQL> ALTER DATABASE DATAFILE 3 ONLINE;Database altered.

Step 12，SQL Plus中打开数据库

SQL> ALTER DATABASE OPEN;Database altered.

Step 13，SQL Plus查看数据，插入数据，成功

SQL> SELECT * FROM scott.dept;    DEPTNO DNAME    LOC---------- -------------- -------------  10 ACCOUNTING   NEW YORK  20 RESEARCH   DALLAS  30 SALES    CHICAGO  40 OPERATIONS   BOSTONSQL> INSERT INTO dept VALUES(89,'GZ','DBA');1 row created.SQL> COMMIT;Commit complete.

四 相关文章

• RMAN备份与恢复之控制文件丢失
• RMAN备份与恢复之可脱机数据文件丢失
• RMAN备份与恢复之基于时间点的不完全恢复

  我的邮箱：[e-mail]  新浪微博：@jutdb           微信公众平台：JustOracle（微信号：justoracle）  数据库技术交流群：336882565（加群时验证 From CSDN XXX）  All is well  2014年1月16日  By Larry Wen

 @Wentasy 博文仅供参考，欢迎大家来访。如有错误之处，希望批评指正。原创博文如需转载请注明出处，谢谢 :) [CSDN博客]