[mrtg] Log File all 0-values

Alex van den Bogaerdt alex at ergens.op.het.net
Wed Mar 21 19:16:08 CET 2007

On Wed, Mar 21, 2007 at 01:58:53PM -0400, Brian A. Seklecki wrote:
> Yea perminute seemed to be the option argument that fixed things
> So what's the rule here? If the couner increments less than 300 ticks (at 
> 5 minute intervals), then switch to "perminute" ?

If you understand the problem, you can make an informed decision.

The problem is that MRTG doesn't write fractions in its log. It truncates
numbers to an integer.

Normally a rate is computed by dividing delta(value) through delta(time).

PerMinute and PerHour modify this process. Rates are 60 or 3600 times
as high as they would normally be.

299/300 is less than 1, thus truncated to 0.
60*299/300 is truncated to 59.
3600*299/300 is truncated to 3588.

There is no rule.  You can decide what to do.  You could also multiply
the input by 1,000,000 and compensate for this -in your mind when viewing
the graphs-.

Or you could switch to RRDtool as backend, select a nice front end and
receive numbers like 900m events (milli, not mega!).


More information about the mrtg mailing list