[mrtg] Strange MRTG problem

[Simon Westlake]

Steve,

Completely agreed that there would be a difference but the 700-800Mbps
usage is pretty sustained. I don't think it is simply the averaging over
the interval that is causing the huge discrepancy.

I misspoke in the earlier email - it is running on 5 minutes in cron.
However, I've been watching it today and it is taking significantly longer
than 5 minutes to complete. So I believe this is a likely candidate - the
interval is set 5 minutes but it is actually only completing a run once
every 15-20 minutes. I'm going to try changing the interval to a much
larger value tonight, dumping the RRDs and setting cron to a value that it
is actually able to complete.


On Wed, May 15, 2013 at 3:12 PM, Steve Shipway <s.shipway at auckland.ac.nz>wrote:

>  >(Cacti is running on 1 minute, MRTG is running on 10 minutes.)
>
> And here we have the explanation.
>
> If MRTG is only polling every 10min (and you have the wrong Interval in
> the cfg file and in your RRD file for this) then it will be, in effect,
> giving you a 10min average speed for the port.
>
> Cacti, on the other hand, is polling every 1min and so is giving you 1min
> averages.
>
> Therefore, the MRTG graph will be much more smoothed out and flatter, and
> Cacti's graph will be more spiky, showing higher peaks for bursts of
> traffic that are relevant in a 1min interval but not over a 10min
> interval.  This is why you see much higher values in Cacti.  The same
> effect happens in MRTG if you compare the peaks on a 5min average graph
> (daily) and on the 2hour average graph (monthly).
>
> Consider this set of 1-minute interval data:
>
> 10
> 1
> 1
> 1
> 1
> 1
> 1
> 1
> 1
> 1
>
> This is a 10min sample; for MRTG it would see a total of 19, IE 1.9/min.
> Cacti, however, would see them all, and show a graph with a spike up to
> 10.  Therefore, very different graphs.
>
> If you want them to match, you will need to change the MRTG Interval to 1
> minute, recreate your RRD files, and poll every minute - and change your
> PHP graphing script to take account of the higher granularity of data
> available.
>
> Steve
>
>  *Steve Shipway*
> University of Auckland ITS
> *UNIX Systems Design Lead*
> s.shipway at auckland.ac.nz
> Ph: +64 9 373 7599 ext 86487
> **Voice: (414) 810-6392
>



-- 
- Simon Westlake
[ digitalgunfire.com ]
AIM: ShirowDG
ICQ: 2227213
Voice: (414) 810-6392
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/mrtg/attachments/20130515/13ce1827/attachment.htm