[rrd-users] Graphing integer data in Cacti/RRDTool

Nick Ellson grimm at nickellson.com
Thu Jun 12 14:27:04 CEST 2008

Hi Tobi!

In this case, with the # of users logged in, I'd like it to display the 
sampling that it has for the 12:00 instant. Then the 12:05 instant for the 
12:05-12:10 window.

So ultimately perhaps I need Cacti's poller to assist in keeping the data 
collection times consistant over the 5 min polling intervals, yet tell 
RRDTool that the data set it is dealing with is integer in nature.

And from what I read, there seems to a a small handful of data sets folks 
are graphing that fit this model (where an interpolation with a decimal 
answer make no sense).

How would you describe the behavior that MRTG without RRD behaved in this 
matter? As it did produce the squarewave like graphs I felt did a great 
job of representing interger gauges such as this.


Nick Ellson
CCIE# 20018
MCSE 2000, Security+, Network+
Network Hobbyist, VFR Private Pilot.

On Thu, 12 Jun 2008, Tobias Oetiker wrote:

> Hi Nick,
> Yesterday Nick Ellson wrote:
>> I have been point to RRD-Tool as the possible cause of an artifact I am
>> seeing in my Cacti graphs of a Cisco ASA VPN Connection Count OID.  If I
>> understand what it happening, the squarewave looking graphs I had in MRTG
>> showing the number of remote users currently logged in, is not possible
>> with RRD Tool, because as the 5 min samples are taken, and an integer
>> value is retrieved but more that 10 secs past the 5 min sampling window
>> then RRD Tool will interpolate the value to what it could have been at the
>> 5 min interval.
>> Hence I have a very analog looking graph of my VPN User count with my
>> current connection count being 3.2 users, with a high today of 9.83.
>> Is this something that can be adjusted in some way to allow this OID to be
>> treated as an absolute interger?
> If you tell rrdtool that you had 5 users from 11:58-12:04 and then
> 3 users from 12:04 - 12:09. What number should rrdtool write into
> the slot from 12:00 - 12:05 ?
> I guess, what you want todo, is to lie to rrdtool about the
> acquisition time of your data. So instead of 12:04 to 12:09 you
> should tell it that the data was for 12:05 to 12:10. But then
> again, maybe you want to take the data as exact as it gets and this
> is unfortunately 'with comma' if you have fixed sampleing times for
> discreet events that happen in analog time.
> cheers
> tobi
> -- 
> Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten
> http://it.oetiker.ch tobi at oetiker.ch ++41 62 213 9902

More information about the rrd-users mailing list