[20201218]数据文件OS头的修复.txt

--//昨天看了http://www.killdb.com/2020/12/10/数据文件头部位图block损坏怎么办,我自己以前也做过类似测试。
--//我把以前的测试看了一遍，自己感觉当时的测试非常混乱,概念也很混乱。决定重新学习，重复测试看看。

1.环境：

SCOTT@book> @ ver1
PORT_STRING                    VERSION        BANNER
------------------------------ -------------- --------------------------------------------------------------------------------
x86_64/Linux 2.4.xx            11.2.0.4.0     Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production

SCOTT@book> alter system dump datafile 4 block 2;
System altered.

SCOTT@book> alter system dump datafile 4 block 3;
System altered.

SCOTT@book> alter system dump datafile 4 block 127;
System altered.

--//首先简单说明一下，数据文件第0块就是os块，如果建立数据文件指定的大小10M，而实际上生产文件大小是10+db_block_size，一般
--//选择db_block_size＝8192, 这样实际文件大小是10*1024*1024+8192.
--//数据文件第1块就是数据文件头，更新检查点就是写这块信息，比如执行alter system checkpoint;命令。
--//第2块为type: 0x1d=KTFB Bitmapped File Space Header
--//第3块为type: 0x1e=KTFB Bitmapped File Space Bitmap
...
--//第127块为type: 0x1e=KTFB Bitmapped File Space Bitmap

--//说明几点：10g与11g的情况有点不同，实际上10g如果如果数据文件很大，可能在尾部还会出现位图区。参考链接：
http://blog.itpub.net/267265/viewspace-2148389/ =>[20171206]位图区一定在数据文件开头吗.txt

--//另外我没有考虑建立小数据文件的情况，因为生产系统不会建立很小的数据文件，比如你可以尝试建立88k的大小，这种情况不在考
--//虑以及测试范围。
--//可以参考http://blog.itpub.net/267265/viewspace-2127936/=>[20161107]关于数据文件位图区.txt
--//另外我个人认为这种破坏1，2个块的情况很少出现，除非遇到oracle的bug，最大的可能前面1M的区域被覆盖。
--//当然测试恢复给几种情况单独非常测试以及恢复。

2.本测试主要解决数据文件OS头的修复:

--//说明一点，os都，数据文件头是不能使用 alter system dump datafile转储的，执行成功并不能看到任何信息。
SCOTT@book> alter system dump datafile 4 block 0;
System altered.

SCOTT@book> alter system dump datafile 4 block 1;
System altered.

*** 2020-12-18 09:56:43.069
End dump data blocks tsn: 4 file#: 4 minblk 2 maxblk 0
*** 2020-12-18 09:57:36.533
Start dump data blocks tsn: 4 file#:4 minblk 1 maxblk 1
Block 1 (file header) not dumped:use dump file header command
End dump data blocks tsn: 4 file#: 4 minblk 2 maxblk 1


SCOTT@book> @ bbvi 4 0
BVI_COMMAND
----------------------------------------------------------------------------------------------------
bvi -b 0 -s 8192 /mnt/ramdisk/book/users01.dbf
xxd -c16 -g 2 -s 0 -l 8192 /mnt/ramdisk/book/users01.dbf
dd if=/mnt/ramdisk/book/users01.dbf bs=8192 skip=0 count=1 of=4_0.dd conv=notrunc 2>/dev/null
od -j 0 -N 8192 -t x1 -v /mnt/ramdisk/book/users01.dbf
hexdump -s 0 -n 8192 -C -v /mnt/ramdisk/book/users01.dbf
alter system dump datafile '/mnt/ramdisk/book/users01.dbf' block 0;

alter session set events 'immediate trace name set_tsn_p1 level 5';
alter session set events 'immediate trace name buffer level 16777216';
9 rows selected.

$ od  -j 0 -N 8192 -t x1  /mnt/ramdisk/book/users01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000020 66 32 00 00 00 20 00 00 00 c8 00 00 7d 7c 7b 7a
0000040 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000060 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0020000


$ ls -l /mnt/ramdisk/book/users01.dbf
-rw-r----- 1 oracle oinstall 419438592 2020-12-18 10:00:50 /mnt/ramdisk/book/users01.dbf

--//419438592-8192 = 419430400
--//419430400/1024/1024 = 400,数据文件大小400M。

419430400/8192 = 51200
51200 = 0xc800

--//对比另外的文件：
$ ls -l /mnt/ramdisk/book/tea01.dbf
-rw-r----- 1 oracle oinstall 41951232 2020-12-18 08:46:21 /mnt/ramdisk/book/tea01.dbf
--//40*1024*1024+8192 = 41951232,建立数据文件大小40M。

$ od  -j 0 -N 8192 -t x1  /mnt/ramdisk/book/tea01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000020 66 ee 00 00 00 20 00 00 00 14 00 00 7d 7c 7b 7a
                                ~~~~~~~~~~~
0000040 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000060 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0020000

$ od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/tea01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000016 66 ee 00 00 00 20 00 00 00 14 00 00 7d 7c 7b 7a
0000032 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0008192

--//40*1024*1024/8192 = 5120
--//5120 = 0x1400
--//对比大概就可以知道主要差异就是在这里，偏移在下划线位置0x24-0x27.
--// 8192 = 0x2000,可以猜测0x20-0x21
--//0x16-0x17是校验和。
--//0x28-0x31是用来标识数据文件的特殊串，我自己这样认为的。

--//实际上这些并不重要，如果os文件头损坏，很简单你只要建立相同大小的数据文件块大小选择一致，使用它覆盖原来的os头就ok了。

3.尝试建立一个数据文件看看。
--//抽取数据文件的定义。
CREATE TABLESPACE TEA DATAFILE
  '/mnt/ramdisk/book/tea01.dbf' SIZE 40M AUTOEXTEND ON NEXT 1M MAXSIZE UNLIMITED
LOGGING
ONLINE
EXTENT MANAGEMENT LOCAL AUTOALLOCATE
BLOCKSIZE 8K
SEGMENT SPACE MANAGEMENT MANUAL
FLASHBACK ON;

--//建立一个类似的数据文件大小一样就ok了。我特意建立大小一样块大小一样，其它参数不一致的情况。
CREATE TABLESPACE sugar DATAFILE
  '/mnt/ramdisk/book/sugar01.dbf' SIZE 40M AUTOEXTEND ON NEXT 2M MAXSIZE 80M
LOGGING
EXTENT MANAGEMENT LOCAL UNIFORM SIZE 1M
BLOCKSIZE 8K
SEGMENT SPACE MANAGEMENT AUTO
FLASHBACK ON;

SCOTT@book> select * from v$dbfile;
     FILE# NAME
---------- ----------------------------------------
         6 /mnt/ramdisk/book/tea01.dbf
         5 /mnt/ramdisk/book/example01.dbf
         4 /mnt/ramdisk/book/users01.dbf
         3 /mnt/ramdisk/book/undotbs01.dbf
         2 /mnt/ramdisk/book/sysaux01.dbf
         1 /mnt/ramdisk/book/system01.dbf
         7 /mnt/ramdisk/book/sugar01.dbf
7 rows selected.

$ od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/sugar01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000016 66 ee 00 00 00 20 00 00 00 14 00 00 7d 7c 7b 7a
0000032 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0008192

$ od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/tea01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000016 66 ee 00 00 00 20 00 00 00 14 00 00 7d 7c 7b 7a
0000032 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0008192

--//对比完成一致。

$ diff -Nur <(od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/tea01.dbf )  <(od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/sugar01.dbf )
--//diff对比就可以说明问题。

4.这样修复os头非常简单，这样使用新建立的数据文件os头覆盖原来的位置就ok了。

--//测试：
$ dd if=/dev/zero of=/mnt/ramdisk/book/users01.dbf count=1 bs=8192  conv=notrunc
1+0 records in
1+0 records out
8192 bytes (8.2 kB) copied, 3.8125e-05 seconds, 215 MB/s

--//再次说明，测试使用dd要小心小心再小心，输入输出文件不要搞错，另外特别注意要加上 conv=notrunc，不然
--//mnt/ramdisk/book/users01.dbf文件会被截断。
--//另外注意我并没有选择/mnt/ramdisk/book/tea01.dbf作为测试对象!!这样主要目的是为了验证不同与原来的os头会出现什么情况。

SYS@book> shutdown abort
ORACLE instance shut down.

SYS@book> startup
ORACLE instance started.

Total System Global Area  643084288 bytes
Fixed Size                  2255872 bytes
Variable Size             205521920 bytes
Database Buffers          427819008 bytes
Redo Buffers                7487488 bytes
Database mounted.
ORA-01157: cannot identify/lock data file 4 - see DBWR trace file
ORA-01110: data file 4: '/mnt/ramdisk/book/users01.dbf'

--//查看对应转储文件：
DDE: Problem Key 'ORA 1110' was flood controlled (0x1) (no incident)
ORA-01110: data file 4: '/mnt/ramdisk/book/users01.dbf'
ORA-01157: cannot identify/lock data file 4 - see DBWR trace file
ORA-01110: data file 4: '/mnt/ramdisk/book/users01.dbf'

$ dbv file=/mnt/ramdisk/book/users01.dbf
DBVERIFY: Release 11.2.0.4.0 - Production on Fri Dec 18 10:50:14 2020
Copyright (c) 1982, 2011, Oracle and/or its affiliates.  All rights reserved.
DBV-00107: Unknown header format (0) (0)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

SYS@book> alter database datafile 4 offline ;
Database altered.

$ dd if=/mnt/ramdisk/book/sugar01.dbf of=/mnt/ramdisk/book/users01.dbf count=1 bs=8192  conv=notrunc
1+0 records in
1+0 records out
8192 bytes (8.2 kB) copied, 3.989e-05 seconds, 205 MB/s
--//再次提醒使用dd要小心。注意文件大小不一致。

$ dbv file=/mnt/ramdisk/book/users01.dbf
DBVERIFY: Release 11.2.0.4.0 - Production on Fri Dec 18 10:52:39 2020
Copyright (c) 1982, 2011, Oracle and/or its affiliates.  All rights reserved.
DBVERIFY - Verification starting : FILE = /mnt/ramdisk/book/users01.dbf
DBVERIFY - Verification complete
Total Pages Examined         : 5120
Total Pages Processed (Data) : 3782
Total Pages Failing   (Data) : 0
Total Pages Processed (Index): 739
Total Pages Failing   (Index): 0
Total Pages Processed (Other): 582
Total Pages Processed (Seg)  : 0
Total Pages Failing   (Seg)  : 0
Total Pages Empty            : 17
Total Pages Marked Corrupt   : 0
Total Pages Influx           : 0
Total Pages Encrypted        : 0
Highest block SCN            : 423619826 (3.423619826)

SYS@book> alter database open ;
alter database open
*
ERROR at line 1:
ORA-01113: file 4 needs media recovery
ORA-01110: data file 4: '/mnt/ramdisk/book/users01.dbf'

SYS@book> recover datafile 4;
Media recovery complete.

SYS@book> alter database open ;
Database altered.

SYS@book> select * from scott.dept where rownum=1;
    DEPTNO DNAME          LOC
---------- -------------- -------------
        10 ACCOUNTING     NEW YORK
--//可以发现并没有出现问题，注意现在os头看到的大小是40M。

$ od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/users01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000016 66 ee 00 00 00 20 00 00 00 14 00 00 7d 7c 7b 7a
0000032 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0008192

SYS@book> alter database datafile '/mnt/ramdisk/book/users01.dbf' resize 300m;
Database altered.

$ od  -A d -j 0 -N 8192 -t x1  /mnt/ramdisk/book/users01.dbf
0000000 00 a2 00 00 00 00 c0 ff 00 00 00 00 00 00 00 00
0000016 66 6c 00 00 00 20 00 00 00 96 00 00 7d 7c 7b 7a
0000032 a0 81 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0000048 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
*
0008192

--//300*1024*1024/8192 = 38400
--//38400 = 0x9600,大小可以对上。

5.总结：
--//OS头恢复相对容易，就是使用dd要小心，如果害怕错误，使用bvi直接修改就不错的选择，因为改动并不是很多。