Oracle数据库坏块的恢复

模拟数据块坏块：

对于发生数据块不一致的数据块，如果当前数据库有备份且处于归档模式，那么就可以利用rman工具数据块恢复功能 对数据块进行恢复，这种方法最简单有效，而且可以在数据文件在线时进行，不会发生数据丢失。对于被有备份的数据库 发生数据块损坏，可能会发生数据的丢失或数据不丢失，这要根据发生坏块的所在的对象决定的，如索引块发生损坏，数据 就不会丢失，重建索引就可以了，发生数据丢失的多发生在表或分区表数据块上。

1.不丢数据的恢复方法

---使用rman工具的

blockrecover 

blockrecover datafile xx block xx;--修复单个坏块
blockrecover corruption list;--修复全部坏块

SQL> select * from livan.test;
select * from livan.test
                    *
ERROR at line :
ORA-: ORACLE data block corrupted (file # , block # )
ORA-: data file : '/u02/app/oradata/PSDB/livan_tbs01.dbf'

[oracle@std u02]$ rman target /

Recovery Manager: Release 10.2.0.4. - Production on Thu Feb  :: 

Copyright (c) , , Oracle.  All rights reserved.

connected to target database: PSDB (DBID=)

RMAN> blockrecover datafile  block ;

Starting blockrecover at -FEB-
using target database control file instead of recovery catalog
allocated channel: ORA_DISK_1
channel ORA_DISK_1: sid= devtype=DISK

channel ORA_DISK_1: restoring block(s)
channel ORA_DISK_1: specifying block(s) to restore from backup set
restoring blocks of datafile
channel ORA_DISK_1: reading from backup piece /u02/PSDB_BACKUP/full_PSDB_870868610
channel ORA_DISK_1: restored block(s) from backup piece
piece handle=/u02/PSDB_BACKUP/full_PSDB_870868610 tag=TAG20150205T115650
channel ORA_DISK_1: block restore complete, elapsed time: ::

starting media recovery
media recovery complete, elapsed time: ::

Finished blockrecover at -FEB-

[oracle@std u02]$ sqlplus '/as sysdba'

SQL*Plus: Release 10.2.0.4. - Production on Thu Feb  :: 

Copyright (c) , , Oracle.  All Rights Reserved.

Connected to:
Oracle Database 10g Enterprise Edition Release 10.2.0.4. - Production
With the Partitioning, OLAP, Data Mining and Real Application Testing options

SQL> select * from livan.test;

        ID NAME
---------- ------------------------------
          beijing
          shanghai
          shandong

如果坏块上的表最近都没有更新，还可以利用bbed的copy命令来从一个最近的备份中copy过来一个数据块恢复，具体不演示。

2.有可能存在数据丢失的恢复（在没有备份没有归档的情况下）

---用户表数据损坏

<1>正常情况下数据条目数

SQL> select count(*) from test;

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

<2>制作一个坏块

select rowid,
dbms_rowid.rowid_relative_fno(rowid) rel_fno,
dbms_rowid.rowid_block_number(rowid) blockno,
dbms_rowid.rowid_row_number(rowid) rowno
from livan.test;   

BBED> set dba ,
        DBA             0x0180004c ( ,)

BBED> d /v dba , offset
 File: /u02/app/oradata/PSDB/livan_tbs01.dbf ()
 Block:       Offsets:     to    Dba:0x0180004c
-------------------------------------------------------
 06a20000 4c008001 d3220800  l .?.L...?......
 b8510000  ddce0000 b4220800 l 窺......菸..?..
    ffff0000 l .......A.......
     l ................
 b4220800    l ?..............
     l ................
     l ................
     l ................

 < bytes per line>

BBED> modify /x  dba , offset
 File: /u02/app/oradata/PSDB/livan_tbs01.dbf ()
 Block:                Offsets:     to             Dba:0x0180004c
------------------------------------------------------------------------
  4c008001 d3220800  b8510000  ddce0000 b4220800
    ffff0000
 b4220800        

 < bytes per line>

BBED> sum play
BBED-: invalid parameter (play)

BBED> sum apply
Check value for File , Block :
current = 0xd0fa, required = 0xd0fa

SQL> alter system flush buffer_cache;

System altered.

SQL> select count(*) from test;
select count(*) from test
*
ERROR at line :
ORA-: ORACLE data block corrupted (file # , block # )
ORA-: data file : '/u02/app/oradata/PSDB/livan_tbs01.dbf'

发现我们第6个文件第76号数据块损坏,报ORA-0178错误，我们知掉只要数据库报ORA-01578错误,
说明该数据块已经被标识为:"software corrupt"

<3>确认坏块的类型

SQL> select segment_name,partition_name,segment_type,owner,tablespace_name
    from sys.dba_extents
    where file_id=&AFN
    and &bad_block_id between block_id and block_id + blocks-;
Enter value for afn:
old   : where file_id=&AFN
new   : where file_id=
Enter value for bad_block_id:
old   : and &bad_block_id between block_id and block_id + blocks-
new   : and  between block_id and block_id + blocks-

SEGMENT_NAME    PARTITION_NAME       SEGMENT_TYPE       OWNER      TABLESPACE_NAME
--------------- -------------------- ------------------ ---------- ------------------------------
TEST                                 TABLE              LIVAN      LIVAN_TBS

经查我们的数据损坏坏位于我们的用户表上,无备份数据会丢失。

<4>标记坏块为"software corrupt"

在第2步的时候全表扫描时已经报ORA-01578错误,说明该数据块已经被标识为:"software corrupt"， 正常情况下可以跳过这步。 我们使用dbms_repair包演示标记坏块为"software corrupt"

使用dbms_repair包可参考：http://blog.itpub.net/8494287/viewspace-1357457/

--利用dbms_repair包必须先创建repair table两个表：

SQL> begin
    dbms_repair.admin_tables(
    table_name=>'REPAIR_TABLE',
    table_type=>dbms_repair.repair_table,
    action=>dbms_repair.create_action,
    tablespace=>'LIVAN_TBS');
    end;
    /

PL/SQL procedure successfully completed.

SQL>  col object_name for a20
SQL> select owner,object_name,object_type
    from dba_objects
    where object_name like '%REPAIR_TABLE%';

OWNER                          OBJECT_NAME          OBJECT_TYPE
------------------------------ -------------------- -------------------
SYS                            REPAIR_TABLE         TABLE
SYS                            DBA_REPAIR_TABLE     VIEW

--再创建orphan key table

SQL> begin
    dbms_repair.admin_tables(
    table_type=>dbms_repair.orphan_table,
    action=>dbms_repair.create_action,
    tablespace=>'LIVAN_TBS');
    end;
    /

PL/SQL procedure successfully completed.

SQL> select owner,object_name,object_type
    from dba_objects
    where object_name like '%ORPHAN_KEY_TABLE%';

OWNER                          OBJECT_NAME          OBJECT_TYPE
------------------------------ -------------------- -------------------
SYS                            ORPHAN_KEY_TABLE     TABLE
SYS                            DBA_ORPHAN_KEY_TABLE VIEW

--检查对象，检查结果会放到我们之前创建的repair_table中

SQL> set serveroutput on
SQL> declare
    rpr_count int;
    begin
    rpr_count:=;
    dbms_repair.check_object(
    schema_name=>'LIVAN',
    object_name=>'TEST',
    repair_table_name=>'REPAIR_TABLE',
    corrupt_count=>rpr_count);
   dbms_output.put_line('repair count:'||to_char(rpr_count));
   end;
   /
repair count:

PL/SQL procedure successfully completed.

检查出有1个坏块

--检查校验的坏块结果

SQL> select object_name,block_id,corrupt_type,marked_corrupt,
    corrupt_description,repair_description
    from repair_table;

OBJECT_NAM   BLOCK_ID CORRUPT_TYPE MARKED_COR CORRUPT_DESCRIP REPAIR_DESCRIPTION
---------- ---------- ------------ ---------- --------------- ------------------------------
TEST                         TRUE                       mark block software corrupt

我们知道当marked_corrupt为TRUE时，标识这个数据块已经被标识过software corrupt

---标识坏块为software corrupt（重新演示一下）

SQL> declare
    fix_count int;
    begin
    fix_count:=;
    dbms_repair.fix_corrupt_blocks(
    schema_name=>'LIVAN',
    object_name=>'TEST',
    object_type=>dbms_repair.table_object,
    repair_table_name=>'REPAIR_TABLE',
   fix_count=>fix_count);
   dbms_output.put_line('fix count:'||to_char(fix_count));
   end;
   /
fix count:

PL/SQL procedure successfully completed.

--再次检查，因为已经被标志为software corrupt，所以在此标志也没什么变化

SQL> select object_name,block_id,corrupt_type,marked_corrupt,
    corrupt_description,repair_description
    from repair_table;

OBJECT_NAM   BLOCK_ID CORRUPT_TYPE MARKED_COR CORRUPT_DESCRIP REPAIR_DESCRIPTION
---------- ---------- ------------ ---------- --------------- ------------------------------
TEST                         TRUE                       mark block software corrupt

未被标志为oftware corrupt ，marked_corrupt列会显示FALSE

<5>检查其他关联对象

检查有多少个索引项指向了坏块的记录

SQL> select index_name from dba_indexes
    where table_name in (select distinct object_name from repair_table);

INDEX_NAME
------------------------------
INDEX_TEST

查询发现有一个索引指向这个坏块

--检查都有多少索引数据

SQL> set serveroutput on
SQL> declare
    key_count int;
    begin
    key_count:=;
    dbms_repair.dump_orphan_keys(
    schema_name=>'LIVAN',
    object_name=>'INDEX_TEST',
    object_type=>dbms_repair.index_object,
    repair_table_name=>'REPAIR_TABLE',
   orphan_table_name=>'ORPHAN_KEY_TABLE',
   key_count=>key_count);
   dbms_output.put_line('orphan key count:'||to_char(key_count));
   end;
   /
orphan key count:

PL/SQL procedure successfully completed.

SQL> select index_name,count(*) from orphan_key_table
    group by index_name;

INDEX_NAME                       COUNT(*)
------------------------------ ----------
INDEX_TEST                            

可以看到有146条数据指向坏块

<6>使用dbms_repair.skip_corrupt_blocks或10231事件方式跳过坏块

SQL> select count(*) from livan.test;
select count(*) from livan.test
                           *
ERROR at line :
ORA-: ORACLE data block corrupted (file # , block # )
ORA-: data file : '/u02/app/oradata/PSDB/livan_tbs01.dbf'

SQL> begin
    dbms_repair.skip_corrupt_blocks(
    schema_name=>'LIVAN',
    object_name=>'TEST',
    object_type=>dbms_repair.table_object,
    flags=>dbms_repair.skip_flag);
    end;
    /

PL/SQL procedure successfully completed.

SQL> select count(*) from livan.test;

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

可以看到当执行完dbms_repair.skip_corrupt_blocks数据可以正常访问了，只是统计出来的数据比原先
少了146条(50604-50458),也就是我们坏块上的数据没有统计，被跳过了。

---使用10231事件跳过

SQL> select count(*) from livan.test;

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

SQL> begin
    dbms_repair.skip_corrupt_blocks(
    schema_name=>'LIVAN',
    object_name=>'TEST',
    object_type=>dbms_repair.table_object,
    flags=>dbms_repair.noskip_flag);
    end;
    /

PL/SQL procedure successfully completed.

SQL> select count(*) from livan.test;
select count(*) from livan.test
                           *
ERROR at line :
ORA-: ORACLE data block corrupted (file # , block # )
ORA-: data file : '/u02/app/oradata/PSDB/livan_tbs01.dbf'

SQL> alter session set events '10231 trace name context forever,level 10'; 

Session altered.

SQL> select count(*) from livan.test;

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

SQL> alter session set events '10231 trace name context off';

Session altered.  

<7>使用CTAS方式重建表及索引

SQL> create table test_bak as select * from test;

Table created.

SQL> create index idx_test_bak on test_bak(object_id);

Index created.

--重建索引语句
SQL> alter index index_test rebuild online;

Index altered.

<8>使用重建对象的freelists方式修复原表

使用这种方式防止坏块以后被加入到freelists中 注意这个方法只适用于段空间手动管理的表空间(SEGMENT SPACE MANAGEMENT MANUAL), 否则会报ORA-10614: Operation not allowed on this segment 错误

SQL> begin
    dbms_repair.rebuild_freelists(
    schema_name=>'LIVAN',
    object_name=>'TEST',
    object_type=>dbms_repair.table_object);
    end;
    /
begin
*
ERROR at line :
ORA-: Operation not allowed on this segment
ORA-: at "SYS.DBMS_REPAIR", line
ORA-: at line 

<9>坏块中的数据

如果坏块中的数据不可丢失，只能尝试其他方法从恢复坏块内容，这其中也有一些第三方付费工具可使用， 也可进行以下尝试：

*尝试从索引内容中恢复出索引列的内容

*尝试使用logminer，从日志中挖掘

*联系Oracle Support，会有些工具解释数据块中的内容。