逻辑读产生Cache Buffer Chain(简称CBC) Latch的解析

测试环境:版本11gR2
SQL> select * from v$version where rownum=1;

BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - 64bit Production

 

一、逻辑读的过程
  1、Oracle以每个块的文件号、块号和类型做HASH运算,得到HASH值。根据HASH值,到HASH表中取出指定块的内存地址
  2、获取CBC Latch(实验的重点测试部分)
  3、根据HASH值,搜索CBC链表
  4、根据DBA找到BH(Buffer Header)加Buffer Pin
  5、加完Buffer Pin马上释放CBC Latch
  6、访问Buffer开始fetch数据
  7、获取CBC Latch
  8、释放Buffer Pin
  9、释放CBC Latch

 逻辑读产生Cache Buffer Chain(简称CBC) Latch的解析_第1张图片

二、取T1表的第一行数据及ROWID,根据dbms_rowid包查出这行数据的文件号、块号
SQL> select rowid,dbms_rowid.rowid_relative_fno(rowid) file#,dbms_rowid.rowid_block_number(rowid) block#,id,name from gyj.t1 where rownum=1;

ROWID                   FILE#     BLOCK#         ID NAME
------------------ ---------- ---------- ---------- ----------
AAASP9AAGAAAACHAAA          6       135          1 aaaaa

这里的DBA(Data Block Address)就是由6号文件和135号块组成

 

三、根据文件号块号获取CBC Latch的地址
SQL> select hladdr from x$bh where file#=6 and dbablk=135;
  HLADDR
----------------
00000003A4282A90

 

四、根据CBC Latch的地址可以查出这个CBC Latch被获得的次数
1.获取CBC Latch的次数为245
SQL> select addr,name,gets from v$latch_children  where addr='00000003A4282A90';
ADDR             NAME                                                                   GETS
---------------- ---------------------------------------------------------------- ----------
00000003B09F7C58 cache buffers chains                                                    245

2.再次读取T1表的第一数据,即产生一次逻辑读
SQL>select id,name from gyj.t1 where rowid='AAASP9AAGAAAACHAAA';
   ID      NAME
   -------- ------------
    1     gyj1

3.CBC Latch的次数变为247,说明一次逻辑读产生两次CBC Latch
SQL> select addr,name,gets from v$latch_children  where addr='00000003A4282A90';
ADDR             NAME                                                                   GETS
---------------- ---------------------------------------------------------------- ----------
00000003A4282A90 cache buffers chains                                                    247

 

五、查本会话下的进程号
SQL> select spid from v$session s,v$process b where s.paddr=b.addr and s.sid in(select sid from v$mystat where rownum=1);

SPID
------------------------
2416

 

六、利用Dtrace工具跟踪
1.编写个简单的dtrace脚本
vi dtrace.d

#!/usr/sbin/dtrace -s -n
dtrace:::BEGIN
{
        i=1;
}

pid$1:::entry
{
        printf("i=%d PID::entry:==%s:%s:%s:%s %x %x %x %d %x %x",i, probeprov, probemod, probefunc, probename,arg0,arg1,arg2,arg3,arg4,arg5);
        i=i+1;
}

2.授权限
chmod 755 dtrace.d

3.执行dtrace命令生成跟踪日志
./dtrace.d -x switchrate=1hz -b 32m 2416 > dtracecbc.log --用这条命令加大了缓存,避免丢失日志

4.显示主要dtrace部分日志,在dtracecbc.log中搜索latch地址:3A4282A90

  CPU     ID                    FUNCTION:NAME
  --逻辑读开始
  1  63919                    kcbgtcr:entry i=592 PID::entry:==pid2416:oracle:kcbgtcr:entry fffffd7ffc98bc00 0 2fe 0 23dede0 fffffd7fffdfa7a0
  --获取CBC Latch
  1 128720                  sskgslcas:entry i=593 PID::entry:==pid2416:oracle:sskgslcas:entry 3a4282a90 0 1 0 0 3a4c953d0          
  1  60714                     ktrexf:entry i=594 PID::entry:==pid2416:oracle:ktrexf:entry fffffd7fffdfa7a0 d7fcd60 0 15392235540 1fe8 23dc9f0
  1  64185                     kcbzgs:entry i=595 PID::entry:==pid2416:oracle:kcbzgs:entry 1 d7fcd60 0 0 0 23dc9f0
  1  53939           kssadf_numa_intl:entry i=596 PID::entry:==pid2416:oracle:kssadf_numa_intl:entry 23 3a4145900 3a487d9e8 0 0 1d
  --根据DBA找到BH(Buffer Header)加Buffer Pin
  1 128720                  sskgslcas:entry i=597 PID::entry:==pid2416:oracle:sskgslcas:entry 3957f6280 0 1 0 3957f6238 3957f6238
  --加完Buffer Pin马上释放CBC Latch
  1 128723                 sskgsldecr:entry i=598 PID::entry:==pid2416:oracle:sskgsldecr:entry 3a4282a90 1 1 0 1 2000000000000000          
  1  63951                     kcbcge:entry i=599 PID::entry:==pid2416:oracle:kcbcge:entry fffffd7ffc98bc00 0 77 0 fffffd7ffc98b7dc 0
  1  64116                  kcbds2pbh:entry i=600 PID::entry:==pid2416:oracle:kcbds2pbh:entry fffffd7ffc98bc00 0 77 0 fffffd7ffc98b7dc 0
  1  60179                     ktcckv:entry i=601 PID::entry:==pid2416:oracle:ktcckv:entry fffffd7ffc98bc00 fffffd7ffc98b7d4 77 0 3a3c935f0 0
  1 104774                     kafger:entry i=602 PID::entry:==pid2416:oracle:kafger:entry 39572e064 fffffd7fffdfb200 39d84e4f0 2 1 1
  1  83919                     kpofcr:entry i=603 PID::entry:==pid2416:oracle:kpofcr:entry fffffd7fffdfb4b0 1 39d84e4f0 2 4c30d40 3a4fd3790
  1 124023                     ttcrxh:entry i=604 PID::entry:==pid2416:oracle:ttcrxh:entry d8055e8 d7f58f0 fffffd7fffdfafe0 48 214 1
  1 209006                     memcpy:entry i=605 PID::entry:==pid2416:libc.so.1:memcpy:entry ddfa329 fffffd7fffdfafe0 30 48 ddfa359 d7f5a00
  1 209006                     memcpy:entry i=606 PID::entry:==pid2416:libc.so.1:memcpy:entry ddfa35b 39572fff5 2 2 2 d7f5a00
  1 123964                     ttcc2u:entry i=607 PID::entry:==pid2416:oracle:ttcc2u:entry d8055e8 d7f58f0 39572fff8 4 1 1
  1 209006                     memcpy:entry i=608 PID::entry:==pid2416:libc.so.1:memcpy:entry ddfa35e 39572fff8 4 852 d7f5a00 d7f5a00
  1  61112            ktsmg_max_query:entry i=609 PID::entry:==pid2416:oracle:ktsmg_max_query:entry 0 0 39572fffc 0 d805250 2
  1 211309                  gethrtime:entry i=610 PID::entry:==pid2416:libc.so.1:gethrtime:entry 92d 1a 39572fffc 8 3a4447d80 39de97dc8
  1 128314                    slcpums:entry i=611 PID::entry:==pid2416:oracle:slcpums:entry d805218 1a 494b3f016 8 3a443a460 39c253698
  1 211401                      times:entry i=612 PID::entry:==pid2416:libc.so.1:times:entry fffffd7fffdfb240 1a 494b3f016 8 3a443a460 39c253698
  1 131260     kglHandleInvalidations:entry i=613 PID::entry:==pid2416:oracle:kglHandleInvalidations:entry d8055e8 39df04bb0 0 8 d7f1a10 39c2d1140
  1 131261             kglHandleLoads:entry i=614 PID::entry:==pid2416:oracle:kglHandleLoads:entry d8055e8 39df04bb0 0 8 d7f1a10 39c2d1140
  1  91756                kksGetStats:entry i=615 PID::entry:==pid2416:oracle:kksGetStats:entry 39d738198 39df04bb0 0 8 d7f1a10 39c2d1140
  1  72794            kews_sqlcol_end:entry i=616 PID::entry:==pid2416:oracle:kews_sqlcol_end:entry 3a4fd2210 fffffd7ffc9e1ad8 39d737e88 70 1 1
  1 110094                  qecrlssub:entry i=617 PID::entry:==pid2416:oracle:qecrlssub:entry 39d84ed10 fffffd7ffc9e1ad8 9 0 fffffd7ffc98ac58 0
  1 108487               qertbRelease:entry i=618 PID::entry:==pid2416:oracle:qertbRelease:entry 39d84dea8 fffffd7ffc98bae0 9 0 4c0 dbb56b0
  1 104779                     kafcpy:entry i=619 PID::entry:==pid2416:oracle:kafcpy:entry 39d84e4f0 2 9 0 fffffd7ffc987430 20
  --访问Buffer开始fetch数据
  1 209006                     memcpy:entry i=620 PID::entry:==pid2416:libc.so.1:memcpy:entry fffffd7ffc9f49f8 39572fff8 4 0 30 20              
  1 209006                     memcpy:entry i=621 PID::entry:==pid2416:libc.so.1:memcpy:entry fffffd7ffc9f49e0 39572fff5 2 1 30 20            
  1  63927                   kcbipnns:entry i=622 PID::entry:==pid2416:oracle:kcbipnns:entry fffffd7ffc98bc00 39572fff5 39572fff7 0 fffffd7ffc98bbe8 2c1

--释放Buffer Pin,调用函数sskgslcas(目的是获取CBC Latch) 

1  63940                     kcbrls:entry i=623 PID::entry:==pid2416:oracle:kcbrls:entry fffffd7ffc98bc00 39572fff5 39572fff7 0 3a3c935f0 2c1
  --获取CBC Latch
  1 128720                  sskgslcas:entry i=624 PID::entry:==pid2416:oracle:sskgslcas:entry 3a4282a90 0 200000000000001d 9 3a4c95288 2000000000000000
  1  64198                     kcbzar:entry i=625 PID::entry:==pid2416:oracle:kcbzar:entry 3a3c93670 3a4282a90 108000 -2748781186960 0 10
   --释放CBC Latch
  1 128723                 sskgsldecr:entry i=626 PID::entry:==pid2416:oracle:sskgsldecr:entry 3a4282a90 200000000000001d fffffffd 1 2000000000000000 1d


      
5.利用oradebug工具,逻辑读获取CBC Latch
SQL> oradebug setmypid
Statement processed.
SQL> oradebug peek 0x3a4282a90 4          --查0x3a4282a90地址开始的4字节信息的值为0
[3A4282A90, 3A4282A94) = 00000000
SQL> oradebug poke 0x3a4282a90 4 1     --修改0x3a4282a90地址开始的4字节信息的值为1,相当于获取了Latch
BEFORE: [3A4282A90, 3A4282A94) = 00000000 --修改前的值
AFTER:  [3A4282A90, 3A4282A94) = 00000001 --修改后的值
SQL> oradebug peek 0x3a4282a90 4       --查0x3a4282a90地址开始的4字节信息的值为1  
[3A4282A90, 3A4282A94) = 00000001

 

6.再开一个会话,会话号为13:
select sid from v$mystat where rownum=1;
 SID
----------
        13

 

7.在13会话下再查询T1表的第一行,这时已被阻塞
select id,name from gyj.t1 where rowid='AAASP9AAGAAAACHAAA';

 

8.再开一个会话,查看会话号是13的等待事件,产生了CBC Latch的等待事件
select sid,event,p1raw,p2raw,p3raw from v$session where sid=13;
  SID EVENT                                                            P1RAW            P2RAW            P3RAW
---------- ---------------------------------------------------------------- ---------------- ---------------- ----------------
        13 latch: cache buffers chains                                      00000003A4282A90 0000000000000096 00


这样跟踪一下,就很清楚为什么一次逻辑读要加两次CBC Latch,一次为了加Buffer Pin,一次为了释放Buffer Pin!





**********本博客所有内容均为原创,如有转载请注明作者和出处!!!**********
Name:    guoyJoe

QQ:        252803295

Email:    [email protected]

Blog:      http://blog.csdn.net/guoyJoe

ITPUB:   http://www.itpub.net/space-uid-28460966.html

OCM:     http://education.oracle.com/education/otn/YGuo.HTM
 _____________________________________________________________
加群验证问题:哪些SGA结构是必需的,哪些是可选的?否则拒绝申请!!!

答案在:http://blog.csdn.net/guoyjoe/article/details/8624392

Oracle@Paradise  总群:127149411

Oracle@Paradise No.1群:177089463(已满)

Oracle@Paradise No.2群:121341761

Oracle@Paradise No.3群:140856036



你可能感兴趣的:(逻辑读产生Cache Buffer Chain(简称CBC) Latch的解析)