[mrtg] Re: mrtg performance

[Alex van den Bogaerdt]

Schuyler Bishop wrote:

> I wonder if this is related to my other issue - getting "artifacts" for
> spikes in my interfaces... it only seems to happen for a single poll, but I
> get "spikes" that are 70-80% of theoretical max.  It's really annoying!
> Granted, I'm polling over the internet...

Spikes are usually generated by counter resets.  There may be other
causes; they are NOT described in this mail so if the following does
not apply to you, I'm sorry:

If the counter is at (say) 268435456 and then wraps to 0, the normal
behaviour of MRTG --->used to be<--- (I don't know the current status)
to ignore the update.  That is, MRTG assume{s|d} a counter reset.

The normal behaviour for RRDtool is to assume a counter overflow.
Overflows may happen quite often.  The overflow detection is quite
elegant:  If the current value is lower than the previous value,
2^32 is added and the resulting number must have been the increment.

counter before:  268,435,456
counter after:            10
Increment:  10 - 268,435,456 ==     -268,435,446
Adjusted:      4,026,531,850
Per second:    4,026,531,850 / 300 == 13,421,772

This is, in this case, obviously wrong.  This is why the min and max
values should be set in RRDtool.  You can do this using rrdtool tune.
The same command can also display the current settings.

The proper way to handle this BTW is to detect a reset from within
the front end.  If a reset occurs, an unknown value should be inserted
into the RRD.

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