> I am trying to monitor some 622Mbit/sec lines, but the graphs start to look 
> strange when the load reaches ~115 Mbit/sec. Polling interval 5 minutes.
> 
> I took the calculator and did (2^32) / (5*60) = 14316558 bytes/s, multiply by 8 
> gives 114,5 Mbit/sec. Is this a "hard" limit in MRTG, or can I do anything to 
> make it work.
> 
I think you're suffering from modulo 32 calculations. Even on a 64-bit
counter this will produce 32 bit counter wraps.

For the short term you could just call mrtg more often, say once every
two minutes or so. This won't avoid the problem, however it will make
it less visible. The problem is in the way mrtg handles counter wraps [1]
and by calling it more often, you get more good samples. Make sure that
your server can handle the extra load and that the job finishes before
the next one starts.

I don't know what happens if you try the combination of mrtg and rrdtool,
it is very much possible that it handles the wraps better but it is also
possible that the problem is in mrtg itself, not rateup. Look in the
contrib directory of the most recent mrtg distribution for 14all.

If this doesn't help, leave mrtg and go for rrdtool by itself or with
another frontend.

Regards,
Alex
[1] Whenever the sample taken this time is lower than the previous one,
    the value is considered wrong and in stead the last rate is used.
    This clips the peaks from your graph. If you enable an option in the 
    config you get zero, not 114.5 mbps but it is still not right.
    Above happens when the data is near but below 114.5 mbps.
    If the new sample is just higher than the last one then it may very
    well be possible that the real increase was 2^32 + the graphed value.

    You didn't mention how the graphs look (just strange) but I think
    they will show a pattern that have both features; first clipping,
    then showing near zero values.

    Please note that I didn't check the most recent rateup but this was
    the behaviour for a long time and I didn't hear anything about it
    being changed.

--
* To unsubscribe from the mrtg mailing list, send a message with the
  subject: unsubscribe to mrtg-request@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg