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

[pvh]

Hello,

I've recently had to install MRTG on Windows with the latest ActivePerl at
work to monitor a server.
What I'm doing is graphing all active interfaces (as per the default
cfgmaker configuration), and I added a totals graph which is simply
something like that:
1:community at ip + 2:community at ip + 3:community at ip .. 

This works perfectly in normal circumstances. However at night when some
backups run all of a sudden the totals no longer add up - even just looking
at the scale (Y axis legend) of the graphs shows a problem: the total
graph's legend goes up to about 80 Mbps where as if you tally up the legends
of all the other graphs you barely get to 30Mbps. The individual interface
graphs make perfect sense, it's the total computed by MRTG that seems way
off.

I can also confirm that by looking at the max values - I did a quick grep of
say "maxin d" on all my individual interface HTML files to get the daily max
octets in value, if I add them all up I'm way below the "maxin d" value I
can find in the totals HTML file. The same test during the day using the
"cuin d" value (current value) adds up perfectly again.

So the issue is clearly with MRTG as the individual graphs are correct and
it's the MRTG generated total that is inconsistent, and this apparently only
happens when there's a larger numbers involved due to the backups.

I'm working with fast ethernet interfaces (all 32 bit counters) here.
One issue I should mention is that the latest ActivePerl package caused an
error related to MRTG's Pod module, which I had to delete (I used the one
provided by ActivePerl instead, and had no problems untim this thing). Could
that be my problem?

What else could I check?

Thanks & Rgds,

Pvh
-- 
View this message in context: http://n2.nabble.com/totals-do-not-add-up-when-graphing-multiple-targets-tp2308575p2308575.html
Sent from the MRTG Users Mailinglist mailing list archive at Nabble.com.