又是latch: cache buffers chains惹得祸
前言
一大早，客户给我打电话说:
xx，应用很慢，查询数据总是超时，让我看看。。。
根据多年DBA经验，首当其冲的肯定是去查询数据库在这段时间都在干嘛。
分析
导出awr报告分析
1）. 数据库在此时间段非常繁忙。
2）. 查看Top 5 Timed Events，出现了Concurrency等待事件latch: library cache**
3）. 查看SQL ordered by Gets,不看不知道，一看吓一跳
4）. *50pwxa3bzp7gkSQL语句
select *
from (select d.*, rownum as num
from (SELECT A.BILLNO,
A.BILLCODE,
A.GETDATE,
A.GETUNITCODE,
A.GETCODE,
A.GETORGANCODE,
A.USEORGANCODE,
A.USEDATE,
A.USEUNITCODE,
A.USERCODE,
A.CURRENCYCODE,
A.AMOUNT,
A.NAME,
A.NOTES,
A.STATUSCODE,
A.IFPAGEONHOLE,
A.OPCODE,
A.OPUNITCODE,
A.OPDATE,
A.LOCKTIME,
A.GETAGENTCODE,
(SELECT D.AGENTNAME
FROM SYN_MM_AGENTCODE_TC D
WHERE D.AGENTCODE = A.GETAGENTCODE) AS GETAGENTNAME,
A.USEAGENTCODE,
A.OUTSTATUS,
CASE A.BILLCODE
WHEN 'B2010005' THEN
A.FACTBILLCODE
ELSE
''
END FACTBILLCODE,
A.SALES,
A.FROMDATE,
A.TODATE,
(SELECT BILLNAME
FROM BD_BILLCODE
WHERE BILLCODE = A.BILLCODE) BILLNAME,
(SELECT HANDLERNAME
FROM BD_HANDLER
WHERE HANDLERCODE = A.USERCODE) USERNAME,
(SELECT HANDLERNAME
FROM BD_HANDLER
WHERE HANDLERCODE = A.GETCODE) GETERNAME,
(SELECT NO3
FROM B_BILLDETAIL
WHERE BILLNO = A.BILLNO
AND BILLCODE = A.BILLCODE
AND FACTBILLCODE = A.FACTBILLCODE) ONLINEINVOICENO,
(SELECT NO4
FROM B_BILLDETAIL
WHERE BILLNO = A.BILLNO
AND BILLCODE = A.BILLCODE
AND FACTBILLCODE = A.FACTBILLCODE) ONLINEINVOICECODE
FROM B_BILL A
WHERE 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND TRIM(BILLNO) >= :B1
AND TRIM(BILLNO) <= :B2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
ORDER BY A.BILLNO) d
where rownum <= 1)
where num > 0
执行计划：
Plan hash value: 4085294641
------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 4632 | 9 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL | CHAGENTBASE | 1 | 56 | 6 (0)| 00:00:01 |
| 2 | TABLE ACCESS BY INDEX ROWID | BD_BILLCODE | 1 | 31 | 1 (0)| 00:00:01 |
|* 3 | INDEX UNIQUE SCAN | PK_BD_BILLCODE | 1 | | 0 (0)| 00:00:01 |
| 4 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 5 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
| 6 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 7 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
| 8 | TABLE ACCESS BY INDEX ROWID | B_BILLDETAIL | 1 | 50 | 4 (0)| 00:00:01 |
|* 9 | INDEX RANGE SCAN | PK_B_BILLDETAIL_02 | 1 | | 3 (0)| 00:00:01 |
| 10 | TABLE ACCESS BY INDEX ROWID | B_BILLDETAIL | 1 | 50 | 4 (0)| 00:00:01 |
|* 11 | INDEX RANGE SCAN | PK_B_BILLDETAIL_02 | 1 | | 3 (0)| 00:00:01 |
|* 12 | VIEW | | 1 | 4632 | 9 (0)| 00:00:01 |
|* 13 | COUNT STOPKEY | | | | | |
| 14 | VIEW | | 2 | 9238 | 9 (0)| 00:00:01 |
| 15 | TABLE ACCESS BY INDEX ROWID| B_BILL | 17395 | 3822K| 9 (0)| 00:00:01 |
|* 16 | INDEX FULL SCAN | PK_B_BILL_01 | 2 | | 8 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("D"."AGENTCODE"=:B1)
3 - access("BILLCODE"=:B1)
5 - access("A"."CODE"=:B1)
7 - access("A"."CODE"=:B1)
9 - access("BILLNO"=:B1 AND "BILLCODE"=:B2 AND "FACTBILLCODE"=:B3)
11 - access("BILLNO"=:B1 AND "BILLCODE"=:B2 AND "FACTBILLCODE"=:B3)
12 - filter("NUM">0)
13 - filter(ROWNUM<=1)
16 - filter(TO_NUMBER(TRIM("BILLNO"))>=601710100010 AND
TO_NUMBER(TRIM("BILLNO"))<=601710100010)5）. 对sql语句进行分析
1、SQL语句中有很多标量子查询，我们可以利用left join 对其改写。
2、id = 1 为TABLE ACCESS FULL，表示CHAGENTBASE走的是全表扫描，在标量子查询中，主表返回多少
行，子表也跟着被扫描多少次，所以需要对CHAGENTBASE建索引。
3、SQL语句中出现TRIM(BILLNO) >= :B1 AND TRIM(BILLNO) <= :B2，导致ID = 16 为
INDEX FULL SCAN。对主键进行索引全扫描，这种访问方式是最垃圾的。
优化
1）创建索引
create indexIDX_CHAGENTBASE_TEST on CHAGENTBASE (AGENTCODE);
2）标量改成left join
select *
from (select G.*, rownum as num
from (SELECT A.BILLNO,
A.BILLCODE,
A.GETDATE,
A.GETUNITCODE,
A.GETCODE,
A.GETORGANCODE,
A.USEORGANCODE,
A.USEDATE,
A.USEUNITCODE,
A.USERCODE,
A.CURRENCYCODE,
A.AMOUNT,
A.NAME,
A.NOTES,
A.STATUSCODE,
A.IFPAGEONHOLE,
A.OPCODE,
A.OPUNITCODE,
A.OPDATE,
A.LOCKTIME,
A.GETAGENTCODE,
/* (SELECT D.AGENTNAME
FROM SYN_MM_AGENTCODE_TC D
WHERE D.AGENTCODE = A.GETAGENTCODE) AS GETAGENTNAME*/
D.AGENTNAME AS GETAGENTNAME,
A.USEAGENTCODE,
A.OUTSTATUS,
CASE A.BILLCODE
WHEN 'B2010005' THEN
A.FACTBILLCODE
ELSE
''
END FACTBILLCODE,
A.SALES,
A.FROMDATE,
A.TODATE,
/* (SELECT BILLNAME
FROM BD_BILLCODE
WHERE BILLCODE = A.BILLCODE) BILLNAME,*/
B.BILLNAME,
/* (SELECT HANDLERNAME
FROM BD_HANDLER
WHERE HANDLERCODE = A.USERCODE) USERNAME,*/
C.HANDLERNAME USERNAME,
/* (SELECT HANDLERNAME
FROM BD_HANDLER
WHERE HANDLERCODE = A.GETCODE) GETERNAME,*/
E.HANDLERNAME GETERNAME,
F.no3 ONLINEINVOICENO,
F.no4 ONLINEINVOICECODE
FROM B_BILL A
LEFT JOIN SYN_MM_AGENTCODE_TC D
ON D.AGENTCODE = A.GETAGENTCODE
LEFT JOIN BD_BILLCODE B
ON B.BILLCODE = A.BILLCODE
LEFT JOIN BD_HANDLER C
ON C.HANDLERCODE = A.USERCODE
LEFT JOIN BD_HANDLER E
ON E.HANDLERCODE = A.GETCODE
LEFT JOIN B_BILLDETAIL F
ON F.BILLNO = A.BILLNO
AND F.BILLCODE = A.BILLCODE
AND F.FACTBILLCODE = A.FACTBILLCODE
WHERE 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND trim(A.BILLNO) >= '601710100010'
AND trim(A.BILLNO) <= '601710100010'
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
AND 3 > 2
ORDER BY A.BILLNO)G
where rownum <= 1)
where num > 0;
Plan hash value: 1528527901
------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 4632 | 21 (0)| 00:00:01 |
|* 1 | VIEW | | 1 | 4632 | 21 (0)| 00:00:01 |
|* 2 | COUNT STOPKEY | | | | | |
| 3 | VIEW | | 2 | 9238 | 21 (0)| 00:00:01 |
| 4 | NESTED LOOPS OUTER | | 2 | 832 | 21 (0)| 00:00:01 |
| 5 | NESTED LOOPS OUTER | | 2 | 770 | 19 (0)| 00:00:01 |
| 6 | NESTED LOOPS OUTER | | 2 | 718 | 17 (0)| 00:00:01 |
| 7 | NESTED LOOPS OUTER | | 2 | 614 | 12 (0)| 00:00:01 |
| 8 | NESTED LOOPS OUTER | | 2 | 562 | 10 (0)| 00:00:01 |
| 9 | TABLE ACCESS BY INDEX ROWID| B_BILL | 17395 | 3822K| 9 (0)| 00:00:01 |
|* 10 | INDEX FULL SCAN | PK_B_BILL_01 | 2 | | 8 (0)| 00:00:01 |
| 11 | TABLE ACCESS BY INDEX ROWID| CHAGENTBASE | 1 | 56 | 1 (0)| 00:00:01 |
|* 12 | INDEX RANGE SCAN | IDX_CHAGENTBASE_TEST | 1 | | 0 (0)| 00:00:01 |
| 13 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 14 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
| 15 | TABLE ACCESS BY INDEX ROWID | B_BILLDETAIL | 1 | 52 | 3 (0)| 00:00:01 |
|* 16 | INDEX RANGE SCAN | PK_B_BILLDETAIL_02 | 1 | | 2 (0)| 00:00:01 |
| 17 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 18 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
| 19 | TABLE ACCESS BY INDEX ROWID | BD_BILLCODE | 1 | 31 | 1 (0)| 00:00:01 |
|* 20 | INDEX UNIQUE SCAN | PK_BD_BILLCODE | 1 | | 0 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("NUM">0)
2 - filter(ROWNUM<=1)
10 - filter(TRIM("A"."BILLNO")>='601710100010' AND TRIM("A"."BILLNO")<='601710100010')
12 - access("D"."AGENTCODE"(+)="A"."GETAGENTCODE")
14 - access("A"."CODE"(+)="A"."GETCODE")
16 - access("F"."BILLNO"(+)="A"."BILLNO" AND "F"."BILLCODE"(+)="A"."BILLCODE" AND
"F"."FACTBILLCODE"(+)="A"."FACTBILLCODE")
18 - access("A"."CODE"(+)="A"."USERCODE")
20 - access("B"."BILLCODE"(+)="A"."BILLCODE")
3） 把Trim去掉
Execution Plan
----------------------------------------------------------
Plan hash value: 1229065410
------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 4632 | 13 (0)| 00:00:01 |
|* 1 | VIEW | | 1 | 4632 | 13 (0)| 00:00:01 |
|* 2 | COUNT STOPKEY | | | | | |
| 3 | VIEW | | 1 | 4619 | 13 (0)| 00:00:01 |
| 4 | NESTED LOOPS OUTER | | 1 | 416 | 13 (0)| 00:00:01 |
| 5 | NESTED LOOPS OUTER | | 1 | 390 | 11 (0)| 00:00:01 |
| 6 | NESTED LOOPS OUTER | | 1 | 364 | 9 (0)| 00:00:01 |
| 7 | NESTED LOOPS OUTER | | 1 | 308 | 8 (0)| 00:00:01 |
| 8 | NESTED LOOPS OUTER | | 1 | 277 | 7 (0)| 00:00:01 |
| 9 | TABLE ACCESS BY INDEX ROWID| B_BILL | 1 | 225 | 4 (0)| 00:00:01 |
|* 10 | INDEX RANGE SCAN | PK_B_BILL | 1 | | 3 (0)| 00:00:01 |
| 11 | TABLE ACCESS BY INDEX ROWID| B_BILLDETAIL | 1 | 52 | 3 (0)| 00:00:01 |
|* 12 | INDEX RANGE SCAN | PK_B_BILLDETAIL_02 | 1 | | 2 (0)| 00:00:01 |
| 13 | TABLE ACCESS BY INDEX ROWID | BD_BILLCODE | 1 | 31 | 1 (0)| 00:00:01 |
|* 14 | INDEX UNIQUE SCAN | PK_BD_BILLCODE | 1 | | 0 (0)| 00:00:01 |
| 15 | TABLE ACCESS BY INDEX ROWID | CHAGENTBASE | 1 | 56 | 1 (0)| 00:00:01 |
|* 16 | INDEX RANGE SCAN | IDX_CHAGENTBASE_TEST | 1 | | 0 (0)| 00:00:01 |
| 17 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 18 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
| 19 | TABLE ACCESS BY INDEX ROWID | T_EMPLOYEE_VIEW | 1 | 26 | 2 (0)| 00:00:01 |
|* 20 | INDEX RANGE SCAN | IND_T_EMPLOYEE_VIEW | 1 | | 1 (0)| 00:00:01 |
------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("NUM">0)
2 - filter(ROWNUM<=1)
10 - access("A"."BILLNO"='601710100010')
12 - access("F"."BILLNO"(+)='601710100010' AND "F"."BILLCODE"(+)="A"."BILLCODE" AND
"F"."FACTBILLCODE"(+)="A"."FACTBILLCODE")
filter("F"."BILLNO"(+)="A"."BILLNO")
14 - access("B"."BILLCODE"(+)="A"."BILLCODE")
16 - access("D"."AGENTCODE"(+)="A"."GETAGENTCODE")
18 - access("A"."CODE"(+)="A"."GETCODE")
20 - access("A"."CODE"(+)="A"."USERCODE")
Statistics
----------------------------------------------------------
621 recursive calls
0 db block gets
229 consistent gets
17 physical reads
0 redo size
2937 bytes sent via SQL*Net to client
2086 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
16 sorts (memory)
0 sorts (disk)
1 rows processed优化效果
可以看出消耗的 buffer cache 从之前的
882,856,212.00降到了229，效率提升了N倍.
又是latch: cache buffers chains惹得祸的更多相关文章
- 【转载】latch: cache buffers chains
本文转自惜分飞的博客,博客原文地址:www.xifenfei.com/1109.html,支持原创,分享知识! 当一个数据块读入sga区,相应的buffer header会被放置到hash列表上,我们 ...
- Oracle索引失效问题：WHERE C1='' OR C2 IN(SubQuery)，并发请求时出现大量latch: cache buffers chains等待
问题描述: 项目反馈某功能响应时间很长,高峰期时系统整体响应很慢... 获取相应的AWR,问题确实比较严重,latch: cache buffers chains等待,因为这些会话SQL执行时间太长, ...
- latch: cache buffers chains故障处理总结(转载)
一大早就接到开发商的电话,说数据库的CPU使用率为100%,应用相应迟缓.急匆匆的赶到现场发现进行了基本的检查后发现是latch: cache buffers chains 作祟,处理过程还算顺利,当 ...
- 案例：latch: cache buffers chains event tuning
前两天对oracle数据库(single instance)进行了迁移升级从10.2.0.4 升到11.2.0.3,有一个项目迁完后第二天,cpu负载升到了130更高(16cpus). 向用户询问后使 ...
- latch: cache buffers chains故障处理总结
一大早就接到开发商的电话,说数据库的CPU使用率为100%,应用相应迟缓.急匆匆的赶到现场发现进行了基本的检查后发现是latch: cache buffers chains 作祟,处理过程还算顺利,当 ...
- 关于latch: cache buffers chains的sql优化
前段时间,优化了一些耗buffer比较多的sql,但是CPU使用率还是没下来 . 查看操作系统CPU使用率 查看awr,发现又有一条超级耗性能的sql冒出来了. 该SQL每次执行耗费3e多个buffe ...
- [转帖]深入理解latch: cache buffers chains
深入理解latch: cache buffers chains http://blog.itpub.net/12679300/viewspace-1244578/ 原创 Oracle 作者:wzq60 ...
- latch：cache buffers chains的优化思路
数据块在buffer cache存放是以linked list方式存放的.当一个session想要访问/修改buffer cache的block,首先需要通过hash算法检查该block是否存在于bu ...
- cache buffers chains latch
cache buffers chains latch 从 Oracle 8i Database 开始, 散列锁存器<-------(1:m)------>hash bucket<-- ...
随机推荐
- caj转pdf——包含下载链接
很多人在知网上下载论文后,想转换成PDF格式,本片一站式教学,包含下载链接. 需要工具 1 caj格式的文件,即要转换的文件. 2 cajviewer,可以在知网的官网上面下载,下载地址参考这里. 3 ...
- Luogu P1638 逛画展 【二分答案】
题目描述 博览馆正在展出由世上最佳的 M 位画家所画的图画. wangjy想到博览馆去看这几位大师的作品. 可是,那里的博览馆有一个很奇怪的规定,就是在购买门票时必须说明两个数字, a和b,代表他要看 ...
- Go 连接PostgreSQL数据库
先在PostgreSQL数据库中建一个表,可以使用PostgreSQL官方的pgAdmin来完成: CREATE TABLE userinfo ( uid serial NOT NULL, usern ...
- nginx 配置tp3.2
server { listen 80; server_name 域名; #charset koi8-r; #access_log /var/log/nginx/host.access.log main ...
- 《Windows核心编程系列》十谈谈同步设备IO与异步设备IO之异步IO
同步设备IO与异步设备IO之异步IO介绍 设备IO与cpu速度甚至是内存访问相比较都是比较慢的,而且更不可预测.虽然如此,通过使用异步设备IO我们仍然能够创造出更高效的程序. 同步IO时,发出IO请求 ...
- ROS学习笔记五：创建和使用ROS msg和srv
1 msg和srv简介 1.1 msg文件 msg文件就是一个简单的text文件,其中每行有一个类型和名称,可用的类型如下: int8, int16, int32, int64 (plus uint* ...
- [POJ1721]Cards
Description 剀剀和凡凡有N张牌(依次标号为1,2,--,N)和一台洗牌机.假设N是奇数.洗牌机的功能是进行如下的操作:对所有位置I(1≤I≤N),如果位置I上的牌是J,而且位置J上的牌是K ...
- 题解报告：hdu 2094 产生冠军
题目链接:http://acm.hdu.edu.cn/showproblem.php?pid=2094 Problem Description 有一群人,打乒乓球比赛,两两捉对撕杀,每两个人之间最多打 ...
- 彩色模型 分类： 图像处理 Matlab 2015-01-08 20:43 364人阅读 评论(0) 收藏
彩色模型(又称彩色空间或彩色系统)是描述色彩的一种方法,本质上,彩色模型就是坐标系统和子空间的规范,系统中的每种颜色由单个点来表示.下面介绍两种最常用的彩色模型. 一.RGB彩色模型: RGB模型是最 ...
- Android开机自启动
1.原理 当Android启动时,会发出一个系统广播,内容为ACTION_BOOT_COMPLETED,它的字符串常量表示为 android.intent.action.BOOT_COMPLETED. ...