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
--创建表空间
SQL> show user;
USER is "SCOTT"
SQL> create tablespace tbs_tmp datafile '/u01/app/oracle/oradata/DBdb/tbs_tmp.dbf' size 10m autoextend on;
Tablespace created.
--创建表对象tb_obj及其索引
SQL> create table tb_obj tablespace tbs_tmp as select * from dba_objects;
Table created.
SQL> create index i_tb_obj on tb_obj(object_id);
Index created.
--表段上的相关信息
SQL> select segment_name , header_file , header_block,blocks from dba_segments where segment_name ='TB_OBJ';
SEGMENT_NAME HEADER_FILE HEADER_BLOCK BLOCKS
--------------------------------------------------------------------------------- ----------- ------------ ----------
TB_OBJ 6 130 1280
--使用linux自带的dd命令来损坏数据块
[oracle@wang ~]$ dd of=/u01/app/oracle/oradata/DBdb/tbs_tmp.dbf bs=8192 conv=notrunc seek=131 <<eof
> Corrupt me!
> EOF
0+1 records in
0+1 records out
13 bytes (13 B) copied, 0.000117639 s, 111 kB/s
[oracle@wang ~]$
--下面的查询收到了错误提示
SQL> 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: '/u01/app/oracle/oradata/DBdb/tbs_tmp.dbf'
SQL>
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
SQL>BEGIN
DBMS_REPAIR.ADMIN_TABLES (
TABLE_NAME => 'REPAIR_TABLE',
TABLE_TYPE => dbms_repair.repair_table,
ACTION => dbms_repair.create_action,
TABLESPACE => 'USERS');
END;
/
PL/SQL procedure successfully completed.
SQL>
--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
SQL> BEGIN
DBMS_REPAIR.ADMIN_TABLES
(
TABLE_NAME => 'ORPHAN_KEY_TABLE',
TABLE_TYPE => DBMS_REPAIR.ORPHAN_TABLE,
ACTION => DBMS_REPAIR.CREATE_ACTION,
TABLESPACE => 'USERS'
);
END;
/
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
SQL> SET SERVEROUTPUT ON
SQL> DECLARE num_corrupt INT;
BEGIN
num_corrupt := 0;
DBMS_REPAIR.CHECK_OBJECT (
SCHEMA_NAME => 'SCOTT',
OBJECT_NAME => 'TB_OBJ',
REPAIR_TABLE_NAME => 'REPAIR_TABLE',
CORRUPT_COUNT => num_corrupt);
DBMS_OUTPUT.PUT_LINE('number corrupt: ' || TO_CHAR (num_corrupt));
END;
/
number corrupt: 1
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
COLUMN object_name FORMAT a10
COLUMN repair_description FORMAT a28
SET LINES 10000
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
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
SQL> SET SERVEROUTPUT ON
SQL> DECLARE num_fix INT;
BEGIN
num_fix := 0;
DBMS_REPAIR.FIX_CORRUPT_BLOCKS (
SCHEMA_NAME => 'SCOTT',
OBJECT_NAME=> 'TB_OBJ',
OBJECT_TYPE => dbms_repair.table_object,
REPAIR_TABLE_NAME => 'REPAIR_TABLE',
FIX_COUNT=> num_fix);
DBMS_OUTPUT.PUT_LINE('num fix: ' || TO_CHAR(num_fix));
END;
/
num fix: 0
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，I_TB_OBJ索引名
SQL> SET SERVEROUTPUT ON
SQL> DECLARE num_orphans INT;
BEGIN
num_orphans := 0;
DBMS_REPAIR.DUMP_ORPHAN_KEYS (
SCHEMA_NAME => 'SCOTT',
OBJECT_NAME => 'I_TB_OBJ',
OBJECT_TYPE => dbms_repair.index_object,
REPAIR_TABLE_NAME => 'REPAIR_TABLE',
ORPHAN_TABLE_NAME=> 'ORPHAN_KEY_TABLE',
KEY_COUNT => num_orphans);
DBMS_OUTPUT.PUT_LINE('orphan key count: ' || TO_CHAR(num_orphans));
END;
/
orphan key count: 88
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88条记录
SQL> select count(*) from orphan_key_table;
COUNT(*)
----------
88
--验证对象是否可以查询，下面的结果显示依旧无法查询
SQL> show user
USER is "SCOTT"
SQL> 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: '/u01/app/oracle/oradata/DBdb/tbs_tmp.dbf'
SQL>
Step e 跳过坏块
--使用SKIP_CORRUPT_BLOCKS来告知Oracle哪些坏块需要被跳过
SQL> BEGIN
DBMS_REPAIR.SKIP_CORRUPT_BLOCKS (
SCHEMA_NAME => 'SCOTT',
OBJECT_NAME => 'TB_OBJ',
OBJECT_TYPE => dbms_repair.table_object,
FLAGS => dbms_repair.skip_flag);
END;
/
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
SQL> alter index scott.i_tb_obj rebuild;
Index altered.
--验证结果
SQL> show user
USER is "SCOTT"
SQL>
SQL> select count(*) from tb_obj;
COUNT(*)
----------
86952
SQL>
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、完成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。
来自 “ ITPUB博客 ” ，链接：https://blog.itpub.net/31397003/viewspace-2147833/ ，如需转载，请注明出处，否则将追究法律责任。
