欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  IT编程

[20190226]测试使用bbed恢复索引.txt

程序员文章站 2022-08-31 13:53:45
[20190226]测试使用bbed恢复索引.txt--//上午做tab$删除恢复测试时发现,tab$的索引i_tab1很小.可以尝试使用bbed解决这个问题.--//首先在普通表上做一个测试看看.1.环境:SCOTT@book> @ ver1PORT_STRING VERSION BANNER x ......

[20190226]测试使用bbed恢复索引.txt

--//上午做tab$删除恢复测试时发现,tab$的索引i_tab1很小.可以尝试使用bbed解决这个问题.
--//首先在普通表上做一个测试看看.

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> create table empy as select * from emp ;
table created.

scott@book> create index i_empy_empno on empy(empno);
index created.

scott@book> select rowid,empy.* from empy where rownum=1;
rowid                   empno ename      job              mgr hiredate                   sal       comm     deptno
------------------ ---------- ---------- --------- ---------- ------------------- ---------- ---------- ----------
aaawpzaaeaaaailaaa       7369 smith      clerk           7902 1980-12-17 00:00:00        800                    20

scott@book> @ rowid aaawpzaaeaaaailaaa
    object       file      block        row rowid_dba            dba                  text
---------- ---------- ---------- ---------- -------------------- -------------------- ----------------------------------------
     91097          4        523          0  0x100020b           4,523                alter system dump datafile 4 block 523 ;

scott@book> select header_file,header_block,segment_name from dba_segments where owner=user and segment_name='i_empy_empno';
header_file header_block segment_name
----------- ------------ --------------------
          4          530 i_empy_empno
--//索引的roor节点在dba=4,531.

2.删除记录看看:

scott@book> delete from empy where empno not in (7369);
13 rows deleted.

scott@book> commit ;
commit complete.

--//仅仅保留1条.并且这条是empy表最小的empno号.

scott@book> alter system checkpoint ;
system altered.

3.先恢复表:
bbed> set dba 4,523
        dba             0x0100020b (16777739 4,523)


bbed> x /rnccntnnn *kdbr[0]
rowdata[529]                                @8150
------------
flag@8150: 0x2c (kdrhfl, kdrhff, kdrhfh)
lock@8151: 0x00
cols@8152:    8

col    0[3] @8153: 7369
col    1[5] @8157: smith
col    2[5] @8163: clerk
col    3[3] @8169: 7902
col    4[7] @8173: 1980-12-17 00:00:00
col    5[2] @8181: 800
col    6[0] @8184: *null*
col    7[2] @8185: 20
--//这条记录保留.

bbed> x /rnccntnnn *kdbr[1]
rowdata[486]                                @8107
------------
flag@8107: 0x3c (kdrhfl, kdrhff, kdrhfd, kdrhfh)
lock@8108: 0x02
cols@8109:    0

--//执行如下生成bbed脚本:
$ seq 1 13 | xargs -i{} echo 'x /rnccntnnn dba 4,523 *kdbr[{}]' | rlbbed | grep -b1 "^lock@" | grep "^flag@.*=0x3c" | cut -d: -f1| cut -d"@" -f2 | xargs -i{} echo assign dba 4,523 offset {}=0x2c
assign dba 4,523 offset 8107=0x2c
assign dba 4,523 offset 8064=0x2c
assign dba 4,523 offset 8023=0x2c
assign dba 4,523 offset 7978=0x2c
assign dba 4,523 offset 7937=0x2c
assign dba 4,523 offset 7896=0x2c
assign dba 4,523 offset 7856=0x2c
assign dba 4,523 offset 7818=0x2c
assign dba 4,523 offset 7775=0x2c
assign dba 4,523 offset 7737=0x2c
assign dba 4,523 offset 7699=0x2c
assign dba 4,523 offset 7660=0x2c
assign dba 4,523 offset 7621=0x2c

--//执行如上脚本,去除删除标识.

bbed> sum apply
check value for file 4, block 523:
current = 0x8c05, required = 0x8c05

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 523

block checking: dba = 16777739, block type = ktb-managed data block
data header at 0x7f6d56b9427c
kdbchk: the amount of space used is not equal to block size
        used=613 fsc=503 avsp=7451 dtl=8064
block 523 failed with check code 6110

--//dtl-used-fsc = avsp

bbed> p ktbbh.ktbbhitl[1]._ktbitun._ktbitfsc
sb2 _ktbitfsc                               @86       503

bbed> assign ktbbh.ktbbhitl[1]._ktbitun._ktbitfsc=0
sb2 _ktbitfsc                               @86       0

bbed> sum apply
check value for file 4, block 523:
current = 0x8df2, required = 0x8df2

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 523

block checking: dba = 16777739, block type = ktb-managed data block
data header at 0xe9427c
kdbchk: space available on commit is incorrect
        tosp=7980 fsc=0 stb=0 avsp=7451
block 523 failed with check code 6111

--//avsp+fsc+stb=tops.

bbed> assign kdbh.kdbhtosp=kdbh.kdbhavsp
sb2 kdbhtosp                                @136      7451

bbed> sum apply
check value for file 4, block 523:
current = 0x8fc5, required = 0x8fc5

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 523

--//ok, 现在对应块已经修复.

scott@book> alter system flush buffer_cache;
system altered.

scott@book> select * from empy ;
     empno ename      job              mgr hiredate                   sal       comm     deptno
---------- ---------- --------- ---------- ------------------- ---------- ---------- ----------
      7369 smith      clerk           7902 1980-12-17 00:00:00        800                    20
      7499 allen      salesman        7698 1981-02-20 00:00:00       1600        300         30
      7521 ward       salesman        7698 1981-02-22 00:00:00       1250        500         30
      7566 jones      manager         7839 1981-04-02 00:00:00       2975                    20
      7654 martin     salesman        7698 1981-09-28 00:00:00       1250       1400         30
      7698 blake      manager         7839 1981-05-01 00:00:00       2850                    30
      7782 clark      manager         7839 1981-06-09 00:00:00       2450                    10
      7788 scott      analyst         7566 1987-04-19 00:00:00       3000                    20
      7839 king       president            1981-11-17 00:00:00       5000                    10
      7844 turner     salesman        7698 1981-09-08 00:00:00       1500          0         30
      7876 adams      clerk           7788 1987-05-23 00:00:00       1100                    20
      7900 james      clerk           7698 1981-12-03 00:00:00        950                    30
      7902 ford       analyst         7566 1981-12-03 00:00:00       3000                    20
      7934 miller     clerk           7782 1982-01-23 00:00:00       1300                    10
14 rows selected.

scott@book> select * from empy where empno in(7499,7369);
     empno ename      job              mgr hiredate                   sal       comm     deptno
---------- ---------- --------- ---------- ------------------- ---------- ---------- ----------
      7369 smith      clerk           7902 1980-12-17 00:00:00        800                    20

--//可以发现走索引仅仅找到1条.

scott@book> analyze table empy validate structure cascade;
analyze table empy validate structure cascade
*
error at line 1:
ora-01499: table/index cross reference failure - see trace file
--//一般普通表的索引不用恢复,根据表重建就ok了.但是对于sys.tab$的索引不行,这也是做这个测试的原因.

4.恢复索引看看.
bbed> set dba 4,531
        dba             0x01000213 (16777747 4,531)

bbed> p kd_off
sb2 kd_off[0]                               @132      8032
sb2 kd_off[1]                               @134      0
sb2 kd_off[2]                               @136      8019
sb2 kd_off[3]                               @138      8006
sb2 kd_off[4]                               @140      7993
sb2 kd_off[5]                               @142      7980
sb2 kd_off[6]                               @144      7967
sb2 kd_off[7]                               @146      7954
sb2 kd_off[8]                               @148      7941
sb2 kd_off[9]                               @150      7928
sb2 kd_off[10]                              @152      7915
sb2 kd_off[11]                              @154      7902
sb2 kd_off[12]                              @156      7889
sb2 kd_off[13]                              @158      7877

--//可以发现bbed查看索引有问题,kd_off[0],kd_off[1]看到的偏移是不对的.实际上kd_off偏移从136开始kd_off[2].

bbed> x /rnx *kd_off[2]
rowdata[172]                                @8119
------------
flag@8119:     0x00 (none)
lock@8120:     0x00
data key:
col    0[3] @8122: 7369
col    1[6] @8126:  0x01  0x00  0x02  0x0b  0x00  0x00

--//这个表的第一条记录(最小值),我没有删除.empno=7369.

bbed> x /rnx *kd_off[3]
rowdata[159]                                @8106
------------
flag@8106:     0x01 (kdxrdel)
lock@8107:     0x02
data key:
col    0[3] @8109: 7499
col    1[6] @8113:  0x01  0x00  0x02  0x0b  0x00  0x01
--//对比删除与不删除的可以看出.flag不同,删除的flag=0x01.

$ seq 2 13 | xargs -i{} echo 'x /rnx dba 4,531 *kd_off[{}]' | rlbbed | grep -b1 "^lock@.*:*0x02$" | grep "flag@.*0x01" | cut -d: -f1 | cut -d@ -f2 |xargs -i{} echo assign dba 4,531 offset {}=0x00
assign dba 4,531 offset 8106=0x00
assign dba 4,531 offset 8093=0x00
assign dba 4,531 offset 8080=0x00
assign dba 4,531 offset 8067=0x00
assign dba 4,531 offset 8054=0x00
assign dba 4,531 offset 8041=0x00
assign dba 4,531 offset 8028=0x00
assign dba 4,531 offset 8015=0x00
assign dba 4,531 offset 8002=0x00
assign dba 4,531 offset 7989=0x00
assign dba 4,531 offset 7977=0x00
--//仅仅修改11条,还有2条需要修改标识.

bbed> dump  offset 160 count 4
 file: /mnt/ramdisk/book/users01.dbf (4)
 block: 531             offsets:  160 to  163   dba:0x01000213
---------------------------------------------------------------
 b81eab1e

 <64 bytes per line>

--// 说明: kd_off[13]的偏移量记录在偏移158,下面2条记录的偏移记录从160算起.
--// b81e 颠倒过来就是  1eb8=7864 , 记录相对偏移从kdxle算起(当前是100), 7864+100 = 7964
--// ab1e 颠倒过来就是  1eab=7851 , 记录相对偏移从kdxle算起(当前是100), 7851+100  = 7951

bbed> x /rnx offset 7964
rowdata[17]                                 @7964
-----------
flag@7964:     0x01 (kdxrdel)
lock@7965:     0x02
data key:
col    0[3] @7967: 7902
col    1[6] @7971:  0x01  0x00  0x02  0x0b  0x00  0x0c

bbed> x /rnx offset 7951
rowdata[4]                                  @7951
----------
flag@7951:     0x01 (kdxrdel)
lock@7952:     0x02
data key:
col    0[3] @7954: 7934
col    1[6] @7958:  0x01  0x00  0x02  0x0b  0x00  0x0d

--//也就是要补充执行如下:
assign dba 4,531 offset 7964=0x00
assign dba 4,531 offset 7951=0x00

--//执行如下脚本:
--//补充如果你仔细看就可以发现键值是按照行目录排序的.如果有记录插入,oracle是通过2分法定位插入行目录的位置,
--//你可以发现后面的偏移不是有序的,这也就是索引的块内无序,块间有序,但是如果索引分裂,oracle会重新排序.

assign dba 4,531 offset 8106=0x00
assign dba 4,531 offset 8093=0x00
assign dba 4,531 offset 8080=0x00
assign dba 4,531 offset 8067=0x00
assign dba 4,531 offset 8054=0x00
assign dba 4,531 offset 8041=0x00
assign dba 4,531 offset 8028=0x00
assign dba 4,531 offset 8015=0x00
assign dba 4,531 offset 8002=0x00
assign dba 4,531 offset 7989=0x00
assign dba 4,531 offset 7977=0x00
assign dba 4,531 offset 7964=0x00
assign dba 4,531 offset 7951=0x00

bbed> sum apply
check value for file 4, block 531:
current = 0x91b5, required = 0x91b5

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 531

block checking: dba = 16777747, block type = ktb-managed data block
**** actual free space = 7593 < kdxcoavs = 7787
**** actual rows marked deleted = 0 != kdxlende = 13
---- end index block validation
block 531 failed with check code 6401


bbed> p kdxle.kdxlende
sb2 kdxlende                                @118      13

bbed> assign  kdxle.kdxlende=0
sb2 kdxlende                                @118      0

bbed> sum apply
check value for file 4, block 531:
current = 0x91b8, required = 0x91b8

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 531

block checking: dba = 16777747, block type = ktb-managed data block
**** actual free space = 7593 < kdxcoavs = 7787
---- end index block validation
block 531 failed with check code 6401

bbed> p kdxle.kdxlexco.kdxcoavs
sb2 kdxcoavs                                @114      7787

bbed> assign kdxle.kdxlexco.kdxcoavs=7593
sb2 kdxcoavs                                @114      7593

bbed> sum apply
check value for file 4, block 531:
current = 0x927a, required = 0x927a

bbed> verify
dbverify - verification starting
file = /mnt/ramdisk/book/users01.dbf
block = 531

--//ok,现在完全修复了.

3.验证看看:
scott@book> alter system flush buffer_cache;
system altered.

scott@book> select * from empy where empno in(7499,7369);
     empno ename      job              mgr hiredate                   sal       comm     deptno
---------- ---------- --------- ---------- ------------------- ---------- ---------- ----------
      7369 smith      clerk           7902 1980-12-17 00:00:00        800                    20
      7499 allen      salesman        7698 1981-02-20 00:00:00       1600        300         30

--//通过索引能定位行记录.
scott@book> analyze table empy validate structure cascade;
table analyzed.

--//analyze table empy validate structure cascade;一切ok,证明修复没有任何问题.

总结:
--//使用bbed修复索引有点繁琐,主要kd_off记录的偏移不对,从kd_off[2]算起.并且遗漏2条键值记录.