AWR中DB CPU/event 占%dbtime很低？

311ybb

大神，请教下关于AWR中db time问题，DB time=db cpu+noidle wait event+wait on cpu queue ，下面这个AWR截图的系统是个白天运行的系统，晚上基本不用，这时间的CPU是很空闲的，为什么db cpu+noidle wait event占的dbtime%这么低，14%左右，如果CPU不繁忙，没有出现cpu run queue较高的情况，那剩下的db time去哪了？

Maclean Liu(刘相兵

给出 Awr 原件

311ybb

AWR Report for DB_ ESS34, Inst_ ess341, Snaps_ 36544-36545.html (727.3 KB, 下载次数: 690)

2016-3-8 09:14 上传

点击文件名下载附件

AWR Report for DB_ ESS34, Inst_ ess342, Snaps_ 36544-36545.html (756.5 KB, 下载次数: 684)

2016-3-8 09:14 上传

点击文件名下载附件

Maclean Liu(刘相兵

Bug 13846587  AIX: Unrealistic CPU Time / DB time with 11.2.0.3 for IBM AIX

This note gives a brief overview of bug 13846587.
The content was last updated on: 29-JAN-2015
Click here for details of each of the sections below.
Affects:

Product (Component)        Oracle Server (Rdbms)
Range of versions believed to be affected        Versions >= 11.2.0.3 but BELOW 12.1
Versions confirmed as being affected       
11.2.0.3
Platforms affected       
IBM AIX Power Based Systems (64bit)

It is believed to be a regression in default behaviour thus:
   Regression introduced in 11.2.0.3
Fixed:

The fix for 13846587 is first included in       
12.1.0.1 (Base Release)
11.2.0.4 (Server Patch Set)

Interim patches may be available for earlier versions - click here to check.
Symptoms:

Related To:

Excessive CPU Usage
Diagnostic Output Problem / Improvement
Wrong Results
Workload repository / reporting
Performance Monitoring
Description

This problem is introduced in 11.2.0.3 on AIX systems by the change from

This is an issue where the value of a CPU time related statistics such as
the DB time in an AWR or the Execute CPU time in a TKPROF shows an
enlarged value in the scale of E11 seconds.

Note:
There is also an AIX problem in that thread_cputime() may return an incorrect value in some
   cases, leading to incorrect/unexpected DB CPU figures in Oracle - see Note:1631769.1.



11.2.0.3 + AIX  存在bug ，说明AIX  下 thread_cputime() 线程计算CPU的时间可能不准确，虽然认为在11.2.0.4 (Server Patch Set)中修复了， 但你仍可以认为在AIX 上这样计算出来的DB CPU 可能是不准的。

Maclean Liu(刘相兵

Following is complete messahe Customer  given regarding to that :
===================================================================
We had a ticket and a call with IBM on Friday. IBM confirms that vac.cfg.61
file delivered in their C++ compiler which is a separate package from the
runtime compiler that comes by default with the OS upgrade. Our OS runtime
compiler is  11.1.0.2 that is the latest and the greatest delivered with IBM
AIX 6.1 version.

Therefore I confirm that we will not have or not able to get vac.cfg.61 file
because (1) having compiler on the production server is against Bank of
America IT policies (2) We do not have license for the compiler.

What am I expecting from Oracle? I am expecting Oracle to resolve this issue
where patch installs on our database w/o any installation issues. This is SEV
1 issue and I will really appreciate if this issue is resolved ASAP.

RELEASE NOTES:
]]If the value of a cpu time related statistics such as the DB time in an AWR o
]]r the Execute cpu time in a TKPROF shows an enlarged value in the scale of E1
]]1 seconds, then apply this patch to correct the problem.


REDISCOVERY INFORMATION:
An AWR report shows enlarged (E11 secs) "DB time" values.
A TKPROF report shows enlarged (E11 secs) Execute cpu time.
WORKAROUND:
None

311ybb

非常感谢！