[mrtg] Re: graph spike problem - stumped.

[Alex van den Bogaerdt]

On Tue, Sep 03, 2002 at 02:04:22PM -0500, Greg.Volk at edwardjones.com wrote:

> This sort of sounds like spikes introduced by a counter RRD data type, 
> although you mentioned that you don't have counter wrap problems, so 
> this may not be the case. Nonetheless, it may be worth your time to 
> experiment with changing the data type to DERIVE instead of counter to 
> see if it helps at all. I've been considering a blanket change from 
> COUNTER to DERIVE to avoid getting 'reboot-spikes' and discussed it on 
> the list in the following post:
> 
> http://www.ee.ethz.ch/~slist/mrtg/msg18908.html

Don't forget to mention that you *need* to set the minimum
allowed value to zero.  This may be the default for MRTG
derived setups but it may not be so for others.

For those of you who don't know this yet:

DERIVE can result in negative values.  You don't want negative
values so you shouldn't allow negative values.  This parameter
can be tuned by using "rrdtool tune x.rrd --minimum ds0:0"
(and of course the same for the other DS).


There's another problem with DERIVE:  each time a counter wraps,
DERIVE results in a negative rate.  This should be, as discussed
above, rejected and turned into an unknown.  If you use a 32-bit
counter for a 100 Mbps interface, the counter will wrap frequently.
This means that when using this trick you will get plenty of NaN
values and little real values.


I still think that using DERIVE is a cool trick but one has to
know when it is safe to use it.

> As always, back-up your data before doing something like this.

s/ before doing something like this/ frequently/

cheers,
Alex

--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi