- 最后登录
- 2015-1-16
- 在线时间
- 15 小时
- 威望
- 0
- 金钱
- 11
- 注册时间
- 2013-6-4
- 阅读权限
- 10
- 帖子
- 19
- 精华
- 0
- 积分
- 0
- UID
- 1121
|
1#
发表于 2013-6-18 17:51:57
|
查看: 4028 |
回复: 8
各位大牛好:
awr中top5排首为library cache mutex x;
Top 5 Timed Foreground Events
Event Waits Time(s) Avg wait (ms) % DB time Wait Class
library cache: mutex X 38,001,653 1,097,433 29 66.95 Concurrency
db file sequential read 10,093,326 352,657 35 21.51 User I/O
DB CPU 115,600 7.05
latch: cache buffers chains 800,279 34,512 43 2.11 Concurrency
db file scattered read 226,327 6,741 30 0.41 User I/O
---硬解析很低;
User calls: 56,660.4 93.8
Parses: 1,881.1 3.1
--version_count也不高,最高才80个
Version Count Executions SQL Id SQL Module SQL Text
80 81,143 ayvfafpr53d34 insert into CP_TMS.TMS_MAIL_PO...
63 382,695 g61pdxvgp6wgt update tms_mail_monitor_info s...
35 1,308,456 2mn1hhpgw9n6g select * from ( select p.*, w....
35 1,308,456 2mn1hhpgw9n6g select * from ( select p.*, w....
30 182,654 4a4kr636txhw4 insert into CP_TMS.TMS_MAIL_MO...
23 677 cxfspv17xb542 insert /*+ append */ into vw_e...
22 5 adfcd02jcf4xc DBMS_SCHEDULER INSERT /*+append*/ INTO SE_TMP...
--Library Cache Activity,reload和invalidations有些高吧
Namespace Get Requests Pct Miss Pin Requests Pct Miss Reloads Invali- dations
SQL AREA 5,752,404 0.04 77,676,677 0.15 21,945 12,119
---如上皆不明显,是否为BUG呢?
https://support.oracle.com/epmos/faces/ui/km/DocumentDisplay.jspx?_afrLoop=488376014159452&id=727400.1&_afrWindowMode=0&_adf.ctrl-state=1d92goqdrn_333
我根据如上的MOS文档727400.1好像也不是BUG;
再看TOP 5第二个等待事件:
db file sequential read
此事件产生原因:
1,过小的buffer cache,导致过量的IO
2,选取了过差的索引或不合理的索引,导致IO增加
3,索引的CLUSTERING FACTOR过高;导致IO增加
4,行链接或迁移产生过量多余的IO
5,统计信息不准导致CBO选用了INDEX SCAN,IO增加
查看AWR中的IO统计部分,在此我仅摘举几条,属于表空间tms和tms_idx的数据文件平均读取时间大大超过了20ms;
Tablespace Filename Reads Av Reads/s Av Rd(ms) Av Blks/Rd Writes Av Writes/s Buffer Waits Av Buf Wt(ms
File IO Stats
TMS +DATA2/tmsdb/datafile/tms.280.814312597 9,215 3 55.62 20.00 24,603 7 10 81.00
TMS_IDX +DATAIDX/tmsdb/datafile/tms_idx.260.813374823 63,947 18 40.25 1.00 71,569 20 128 21.72
另外:
Load Profile
Per Second Per Transaction Per Exec Per Cal
Block changes: 95,415.6 157.9
基本每秒为700m的数据量变化,数据库蛮BUSY的
--如下可知主机CPU很闲;压力很低;而ORACLE占用的CPU很低,仅25.9%
Host CPU (CPUs: 128 Cores: 64 Sockets: 8)
Load Average Begin Load Average End %User %System %WIO %Idle
116.51 70.94 23.3 2.3 31.5 73.1
Instance CPU
%Total CPU %Busy CPU %DB time waiting for CPU (Resource Manager)
25.9 96.2 0.0
我想请问的是:
基于这个awr,
问题:
1,诊断思路是如何的呢,我的思路对吗?
2,大家谈谈思路与方法 |
提供针对ORACLE初学者及进阶者培训,ORACLE各项RAC,DATA GUARD安装,部署,调优及SQL优化,http://blog.itpub.net/9240380/,欢迎来电咨询:18201115468
|
|