Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: off topic: iostat -x in linux

Re: off topic: iostat -x in linux

From: Craig I. Hagan <hagan_at_cih.com>
Date: Wed, 29 Jan 2003 11:43:57 -0800
Message-ID: <F001.0053DA30.20030129114357@fatcity.com>


> In linux, it seems difficult, I have tried linux iostat and sysstat-iostat, non of them seems to work properly with util column in iostat report. Though no activity on the server at all, iostat always report utilization of the partitions 100% Why?
> I also searched google and found the same question asked by others, but nobody replied.
>
> Anyone have the same experience? How did you measure the load on the io subsystem in your linux server then? IO wait is still 0%, but io capacity usage is 100%. Confusing...

I usually take the %util as provided by the os and ignore it. Only the disk itself truly knows how well utilized it is -- you may have other problems (small requests over fibre, etc which causeyou to talk more protocol than data, for example). If you can't pull stats from the disk, you can watch its response time which will tell you when it is truly overutilized (service time tends to increase as a function of load, with a knee in most systems that is rather pronounced, raid5 being *far* more noticeable when faced with writes, especially non-stripe aligned writes).

This appears in iostat as svtcm, which is the number of ms it takes the relevant disk/partition to respond to the io requests just tracked. You can also look at mean io size, which requires a teensy bit of math, compute it like this:

		Read: ((rsec)/2)/rrqm)
		Write: ((wsec)/2)/wrqm)


Also, realize with with modern disk subsystems, you can often take bursts of traffic which far exceed the disks capacity to handle it and not see a major spike due to cache effects, particularly with write-behind caching systems. Often such systems will look fabulous for small spikes but will show good behavior that suddenly falls apart when handed a continous load which exceeds the array's ability to flush cache to the disks. This is part of why the utilization stats (on the os) are fairly meaningless as they are based on complete guesses rather than true knowledge of the underlying subsystems.

          .- ... . -.-. .-. . - -- . ... ... .- --. .

			    Craig I. Hagan
			   hagan(at)cih.com



-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: Craig I. Hagan
  INET: hagan_at_cih.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
Received on Wed Jan 29 2003 - 13:43:57 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US