[mrtg] totals do not add up when graphing multiple targets

[McDonald, Dan]

On Thu, 2009-02-12 at 08:03 -0800, pvh wrote:
> Thanks for your reply Dan.
> 
> I don't really understand how I could be experiencing counter roll-overs?
> This would be an issue when a 32 bit counter wraps around in less than 5
> minutes, which cannot be happening on my fast ethernet interfaces (which btw
> don't even go above 10% utilization).

No, it occurs when any of the interfaces rolls over once, because the
aggregate traffic can mask the issue.

> surely MRTG (well, Perl
> I guess) is capable of handling this properly and is not just adding up my
> 8x32 bit counters into a single 32bit counter?

Nope.  all it does is add them up, it does not add up the distilled
numbers.  What if you wanted to graph an OID as both a counter and a
gauge in different instances?


> Besides, wouldn't counter wrap result in _lower_ values than expected? 

No, merely random.  What if two of the devices roll over at the same
time? Then what?

> Anyway if that's what happening then I don't understand the point of being
> able to add/subtract/divide targets in MRTG if it doesn't work?

It doesn't work reliably with counters.  It works reasonably well with
absolute and guage types.  For low-bandwidth connections where there are
only a couple of interfaces, it can probably survive a single roll-over.
For 8 high-speed interfaces where multiple single rollovers are likely,
it's completely unreliable.

Patches welcome.

-- 
Daniel J McDonald, CCIE #2495, CISSP #78281, CNX
Austin Energy
http://www.austinenergy.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://lists.oetiker.ch/pipermail/mrtg/attachments/20090212/8ef8eb12/attachment.bin