使用 DBMS_REPAIR 修复坏块

对于Oracle数据块物理损坏的情形，在我们有备份的情况下可以直接使用备份来恢复。对于通过备份恢复，Oracel为我们提供了很多种方式，冷备，基于用户管理方式，RMAN方式等等。对于这几种方式我们需要实现基于数据库以及文件级别的恢复。RMAN同时也提供了基于块介质方式的恢复。也就是说我们根本不需要还原数据文件，而是直接从备份文件基于块来提取以实现联机恢复。可参考基于RMAN实现坏块介质恢复(blockrecover) 。这是比较理想的情形。如果没有任何备份怎么办？我们可以使用Oracle自带的DBMS_REPAIR包来实现修复。注意，不要被文章题目有所误导。这里的修复是有损修复也就是说将受损的数据块标记为坏块，不对其进行访问而已。就好比我们磁盘有坏道，找个磁盘修复工具将坏道标出来不使用，同理。那受损的数据岂不是无力回天啦，呜呜......要记得随时备份阿。。

1、DBMS_REPAIR包所含的过程
  Procedure_Name       Description
  -----------------    ------------------------------------
  ADMIN_TABLES         Provides administrative functions (create, drop, purge) for repair or orphan key tables.
                       Note: These tables are always created in the SYS schema.
  CHECK_OBJECT         Detects and reports corruptions in a table or index
  DUMP_ORPHAN_KEYS     Reports on index entries that point to rows in corrupt data blocks
  FIX_CORRUPT_BLOCKS   Marks blocks as software corrupt that have been previously identified as corrupt by the CHECK_OBJECT procedure
  REBUILD_FREELISTS    Rebuilds the free lists of the object
  SEGMENT_FIX_STATUS   Provides the capability to fix the corrupted state of a bitmap entry when segment space management is AUTO
  SKIP_CORRUPT_BLOCKS When used, ignores blocks marked corrupt during table and index scans.
                       If not used, you get error ORA-01578 when encountering blocks marked corrupt.

2、DBMS_REPAIR的一些局限性
    Tables with LOB data types, nested tables, and varrays are supported, but the out-of-line columns are ignored.
  Clusters are supported in the SKIP_CORRUPT_BLOCKS and REBUILD_FREELISTS procedures, but not in the CHECK_OBJECT procedure.
  Index-organized tables and LOB indexes are not supported.
  The DUMP_ORPHAN_KEYS procedure does not operate on bitmap indexes or function-based indexes.
  The DUMP_ORPHAN_KEYS procedure processes keys that are no more than 3,950 bytes long.

3、创建演示环境

--当前环境
sys@USBO> select * from v$version where rownum<2;

BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production

--创建表空间
sys@USBO> create tablespace tbs_tmp datafile '/u02/database/usbo/oradata/tbs_tmp.dbf' size 10m autoextend on;

--创建表对象tb_obj及其索引
sys@USBO> create table tb_obj tablespace tbs_tmp as select * from dba_objects;

sys@USBO> create index i_tb_obj on tb_obj(object_id);

--表段上的相关信息
sys@USBO> select segment_name , header_file , header_block,blocks
  2  from dba_segments where segment_name ='TB_OBJ';

SEGMENT_NAME                   HEADER_FILE HEADER_BLOCK     BLOCKS
------------------------------ ----------- ------------ ----------
TB_OBJ                                   6          130       1152

--使用linux自带的dd命令来损坏数据块
[oracle@linux1 ~]$ dd of=/u02/database/usbo/oradata/tbs_tmp.dbf bs=8192 conv=notrunc seek=131 <<EOF
> Corrupt me!
> EOF
0+1 records in
0+1 records out
12 bytes (12 B) copied, 0.000209854 seconds, 57.2 kB/s
[oracle@linux1 ~]$ dd of=/u02/database/usbo/oradata/tbs_tmp.dbf bs=8192 conv=notrunc seek=141 <<EOF 
> Corrupt me!
> EOF
0+1 records in
0+1 records out
12 bytes (12 B) copied, 0.00019939 seconds, 60.2 kB/s
[oracle@linux1 ~]$ dd of=/u02/database/usbo/oradata/tbs_tmp.dbf bs=8192 conv=notrunc seek=151 <<EOF 
> Corrupt me!
> EOF
0+1 records in
0+1 records out
12 bytes (12 B) copied, 2.1672e-05 seconds, 554 kB/s

sys@USBO> alter system flush buffer_cache;

--下面的查询收到了错误提示
sys@USBO> select count(*) from tb_obj;
select count(*) from tb_obj
                     *
ERROR at line 1:
ORA-01578: ORACLE data block corrupted (file # 6, block # 131)
ORA-01110: data file 6: '/u02/database/usbo/oradata/tbs_tmp.dbf'

4、使用DBMS_REPAIR修复坏块

Step a 创建相应的表对象
--使用DBMS_REPAIR.ADMIN_TABLES过程创建一个表对象，用于记录需要被修复的表
sys@USBO> BEGIN
  2    DBMS_REPAIR.ADMIN_TABLES (
  3       TABLE_NAME => 'REPAIR_TABLE',
  4       TABLE_TYPE => dbms_repair.repair_table,
  5       ACTION     => dbms_repair.create_action,
  6       TABLESPACE => 'USERS');
  7  END;
  8  /

PL/SQL procedure successfully completed.

--使用DBMS_REPAIR.ADMIN_TABLES过程创建一个表对象，用于记录在表块损坏后那些孤立索引，也就是指向坏块的那些索引                     
sys@USBO> BEGIN
  2    DBMS_REPAIR.ADMIN_TABLES 
  3    (
  4       TABLE_NAME => 'ORPHAN_KEY_TABLE',
  5       TABLE_TYPE => DBMS_REPAIR.ORPHAN_TABLE,
  6       ACTION     => DBMS_REPAIR.CREATE_ACTION,
  7       TABLESPACE => 'USERS'
  8    );
  9  END;
 10  /

PL/SQL procedure successfully completed.


Step b 校验受损的对象
--使用DBMS_REPAIR.CHECK_OBJECT来检测对象上受损的情形，并返回受损块数
sys@USBO> SET SERVEROUTPUT ON
sys@USBO> DECLARE num_corrupt INT;
  2  BEGIN
  3   num_corrupt := 0;
  4   DBMS_REPAIR.CHECK_OBJECT (
  5       SCHEMA_NAME => 'SYS',
  6       OBJECT_NAME => 'TB_OBJ',
  7       REPAIR_TABLE_NAME => 'REPAIR_TABLE',
  8       CORRUPT_COUNT =>  num_corrupt);
  9   DBMS_OUTPUT.PUT_LINE('number corrupt: ' || TO_CHAR (num_corrupt));
 10  END;
 11  /
number corrupt: 3

PL/SQL procedure successfully completed.

--下面我们可以从repair_table查询到受损的块
--从下面的查询中可以看出列marked_corrupt全部为true，表明我们在CHECK_OBJECT已经标注了坏块
--有一点不太清楚的是为什么块131在查询中被列出3次？
sys@USBO> COLUMN object_name FORMAT a10
sys@USBO> COLUMN repair_description FORMAT a28
sys@USBO> SET LINES 10000
sys@USBO> SELECT object_name, block_id, corrupt_type,marked_corrupt,repair_description  FROM repair_table;

OBJECT_NAM   BLOCK_ID CORRUPT_TYPE MARKED_COR REPAIR_DESCRIPTION
---------- ---------- ------------ ---------- ----------------------------
TB_OBJ            131         6148 TRUE       mark block software corrupt
TB_OBJ            131         6148 TRUE       mark block software corrupt
TB_OBJ            131         6148 TRUE       mark block software corrupt
TB_OBJ            141         6148 TRUE       mark block software corrupt
TB_OBJ            151         6148 TRUE       mark block software corrupt


Step c 标记坏块
--过程FIX_CORRUPT_BLOCKS用于标记坏块，在这个演示中，我们在CHECK_OBJECT已经被标注了，如没有执行下面的过程
--由于上一步已经标注，所以下面的输出为0
sys@USBO> SET SERVEROUTPUT ON
sys@USBO> DECLARE num_fix INT;
  2  BEGIN 
  3   num_fix := 0;
  4   DBMS_REPAIR.FIX_CORRUPT_BLOCKS (
  5       SCHEMA_NAME => 'SYS',
  6       OBJECT_NAME=> 'TB_OBJ',
  7       OBJECT_TYPE => dbms_repair.table_object,
  8       REPAIR_TABLE_NAME => 'REPAIR_TABLE',
  9       FIX_COUNT=> num_fix);
 10   DBMS_OUTPUT.PUT_LINE('num fix: ' || TO_CHAR(num_fix));
 11  END;
 12  /
num fix: 0

-- Author   : Robinson Cheng
-- Blog     : http://blog.csdn.net/robinson_0612
-- DB forum : http://bbs.dbsupport.cn

PL/SQL procedure successfully completed.


Step d DUMP孤立的索引键值
--使用DUMP_ORPHAN_KEYS过程将那些指向坏块的索引键值填充到ORPHAN_KEY_TABLE
sys@USBO> SET SERVEROUTPUT ON
sys@USBO> DECLARE num_orphans INT;
  2  BEGIN
  3   num_orphans := 0;
  4   DBMS_REPAIR.DUMP_ORPHAN_KEYS (
  5       SCHEMA_NAME => 'SYS',
  6       OBJECT_NAME => 'I_TB_OBJ',
  7       OBJECT_TYPE => dbms_repair.index_object,
  8       REPAIR_TABLE_NAME => 'REPAIR_TABLE',
  9       ORPHAN_TABLE_NAME=> 'ORPHAN_KEY_TABLE',
 10       KEY_COUNT => num_orphans);
 11   DBMS_OUTPUT.PUT_LINE('orphan key count: ' || TO_CHAR(num_orphans));
 12  END;
 13  /
orphan key count: 242

PL/SQL procedure successfully completed.

--下面的查询可以看到正好等于上面返回的数量也就是242条记录
sys@USBO> select count(*) from orphan_key_table;

  COUNT(*)
----------
       242

--验证对象是否可以查询，下面的结果显示依旧无法查询       
sys@USBO> select count(*) from tb_obj;
select count(*) from tb_obj
                     *
ERROR at line 1:
ORA-01578: ORACLE data block corrupted (file # 6, block # 131)
ORA-01110: data file 6: '/u02/database/usbo/oradata/tbs_tmp.dbf'


Step e 跳过坏块
--使用SKIP_CORRUPT_BLOCKS来告知Oracle哪些坏块需要被跳过
sys@USBO> BEGIN
  2    DBMS_REPAIR.SKIP_CORRUPT_BLOCKS (
  3       SCHEMA_NAME => 'SYS',
  4       OBJECT_NAME => 'TB_OBJ',
  5       OBJECT_TYPE => dbms_repair.table_object,
  6       FLAGS => dbms_repair.skip_flag);
  7  END;
  8  /

PL/SQL procedure successfully completed.

--由于索引键上存在孤立索引，因此我们重建索引
sys@USBO> alter index i_tb_obj rebuild;

Index altered.

--验证结果
sys@USBO> select count(*) from tb_obj;

  COUNT(*)
----------
     72211

5、后记
a、再次提醒，备份重于一切，因此无论何时应保留可用的备份。
b、DBMS_REPAIR包并不是真正意思上的坏块修复，而是标记坏块，损坏的这部分数据被丢失。
c、DBMS_REPAIR包的几个步骤，先创建相应的表用于存储修复表及孤立索引，其次CHECK_OBJECT，FIX_CORRUPT_BLOCKS，DUMP_ORPHAN_KEYS，SKIP_CORRUPT_BLOCKS。
d、完整DBMS_REPAIR上面提到的几个步骤后，建议重建索引。
e、注，如果受损表对象被其他对象参照，建议先disable这些约束，那些在子表上孤立的记录可根据情形决定后再enable约束。

专业数据库论坛：http://bbs.dbsupport.cn/