Hi Roland
A good comment, and one I did consider.
My concern with a long sample, is that it becomes "too averaged".
If I put one hand in the fire, and the other in liquid nitrogen, on average,
I am doing fine, right? :-)
I think it's one of those things that will need to be flavoured to taste,
depending on your requirements, which is why I changed it to be a soft
parameter that could be set at server-side.
Ultimately, there is no difference in the results of running iostat 300 13
every hour, and iostat 300 2 every 5 minutes for one hour.
And will save a lot of coding :-)
Cheers
Vernon
On Fri, Sep 10, 2010 at 5:56 AM, Roland Soderstrom <
user-0cec9512a49f@xymon.invalid> wrote:
Vernon, I have an comment.
• -
- 100 dollar bill
• -
- If you have a 10sec iostat which you run every 5min, as you set the test
to run, you get huge gaps. -
- Say that your server always have a lot of IO every 3min for 1
min. -
- You will probably miss all this IO, set it to the same as the
clientlaunch time 5min = 300 in the script. -
- iostat need to run constantly to make accurate measurements.
• - That's why you see the default do 300
sec. -
• -
To get better accuracy you should probably run "iostat dxn 300 13" and run
the script from clientlaunch every hour.
A bit more coding to take care of the 11 "new" iostat though.
- Roland
On 9/09/10 06:56 PM, Vernon Everett wrote:
Just updated the code. Minor change.
After discussion with a colleague we thought that a 2 second iostat sample
might not be adequate.
We settled on 10, but with the update, I have made it a parameter, which
can be set at server side.
Cheers
Vernon
On Thu, Sep 9, 2010 at 1:56 PM, Vernon Everett <user-b3f8dacb72c8@xymon.invalid>wrote:
Hi all
Just posted an iostat add-on, with graphing goodness to
http://xymonton.trantor.org/doku.php/monitors:diskstat.ksh