Xymon Mailing List Archive search

bb-iostat.sh?

4 messages in this thread

list David W Gore · Sun, 15 May 2011 21:34:31 +0000 ·
We use bb-iostat.sh to monitor some performance stats but I think it may
be way outdated and not very accurate for multiple cpu, SAN disk
attached Solaris machines.  Does anyone have any opinions or a better
tool to monitor iostat, swapping etc?  I am fairly ignorant about
performance measurements and tuning btw except for the basics.

 
~David
list Roland Soderstrom · Mon, 16 May 2011 07:59:55 +1000 ·
Hi

I'm very interested in the same.
There is some code in the current Xymon but it is broken.
Iostat data is sent (by default) to the server but it is not taken care of properly at the server end.
Some code is there but it lacks the code to put it on pages. If I remember right it is the same xxx.c that add disks.
Someone with some C skills need to make it work.

Search the archive and you find some posts from me that explains more details.
I had a very lame go at the code but couldn't spend the time to learn C again...

- Roland

quoted from David W GoreOn 16/05/11 07:34 AM, Gore, David W wrote:

We use bb-iostat.sh to monitor some performance stats but I think it may be way outdated and not very accurate for multiple cpu, SAN disk attached Solaris machines. Does anyone have any opinions or a better tool to monitor iostat, swapping etc? I am fairly ignorant about performance measurements and tuning btw except for the basics.


~David






Xymon@xymon.com

list Vernon Everett · Mon, 16 May 2011 12:42:47 +0800 ·
For memory, I use http://xymonton.org/doku.php/monitors:memstat
For disk stats, I use http://xymonton.org/doku.php/monitors:diskstat.ksh

Works for me. YMMV.

Regards
     Vernon


On Mon, May 16, 2011 at 5:59 AM, Roland Soderstrom
quoted from Roland Soderstrom
<user-0cec9512a49f@xymon.invalid> wrote:
Hi

I'm very interested in the same.
There is some code in the current Xymon but it is broken.
Iostat data is sent (by default) to the server but it is not taken care of
properly at the server end.
Some code is there but it lacks the code to put it on pages. If I remember
right it is the same xxx.c that add disks.
Someone with some C skills need to make it work.

Search the archive and you find some posts from me that explains more
details.
I had a very lame go at the code but couldn't spend the time to learn C
again...

- Roland

On 16/05/11 07:34 AM, Gore, David W wrote:

We use bb-iostat.sh to monitor some performance stats but I think it may be
way outdated and not very accurate for multiple cpu, SAN disk attached
Solaris machines.  Does anyone have any opinions or a better tool to monitor
iostat, swapping etc?  I am fairly ignorant about performance measurements
and tuning btw except for the basics.


~David

list Vernon Everett · Mon, 16 May 2011 17:37:15 +0800 ·
Then either make the sample time longer, or increase the sample frequency.
The former will generate more "average" data.
The latter will require some serious modding to your Xymon environment
and rrd setup.

It's really there only to give an indication of status, for which a
short sample will do.
A 10 second spike, is not really a problem. Three consecutive spikes
would seem to indicate that there is a sustained peak.
Possibly a worry? I don't know - depends on your environment.
Something you might expect at backup time perhaps?

Like all sample tests, it's a sample, which we use to create a model of reality.
What is there has been "good enough" for my needs, and has allowed me
to do some very important diagnostics and tuning.
Feel free to adjust your sample set to suit your needs.
Disclaimer - YMMV :-)
quoted from Vernon Everett

Regards
     Vernon


On Mon, May 16, 2011 at 2:52 PM, W.J.M. Nelis <user-6956df205d63@xymon.invalid> wrote:
Vernon Everett wrote:
For memory, I use http://xymonton.org/doku.php/monitors:memstat
For disk stats, I use http://xymonton.org/doku.php/monitors:diskstat.ksh
My problem with the latter is that it takes a sample, thus the measurements
cover only a fraction of the time. In this case it will cover 10 out of
every 300 seconds.

Regards,
 Wim Nelis.


*******************************************************************************************************
The NLR disclaimer (http://www.nlr.nl/emaildisclaimer) is valid for NLR
e-mail messages.
*******************************************************************************************************