[mrtg] Re: Antwort: Re: Differences between mrtg.log and html view

[Daniel J McDonald]

On Mon, 2003-07-21 at 09:58, Bernd.Stroessenreuther at spb.de wrote:

> > The values are being interpolated so that they line up on 5-minute
> > boundaries.  This is because you can't poll exactly in 5 minute
> > intervals, so it fiddles with the delta and guesses where you would have
> > been at the precise 5 minute boundaryy.
> 
> This is ok, if you have a always increasing value like an interface couter,
> but it's IMO not ok, if I have a per-calutlated value, that I give to mrtg.
> I tell just this to mrtg by setting the option gauge.

No, it still has to interpolate, even for gauges.  The only palce were
it is really not appropriate is for flags - if 1 is ok, 2 is warning,
and 3 is critical, then a value of 1.5 won't trip thresholds until a
second interval has passed.  Things that are continuously variable have
to be assumed to take straight-line changes.  For example, when I record
temperature, I want it to guess what the temperature would have been at
the time it was supposed to graph so that my graphs are relatively
accurate.
> 
> Take my example: The CPU has had a usage (for usr-processes) of 68% over
> the last 5 minutes. So even if the measurement interval is exactly 4 min 50
> sec or 5 min 10 sec, the usage in the last (exact) 5 min is probably 68%.
> It is probably not 65% and probably not 71%.
> 
But, the question is:  what would the 5 minute rate have been at the
time that the graph has to display?  The graphs display at strict
5-minute intervals - 1:00:00, 1:05:00, 1:10:00, etc, while the data
might have been collected at 1:02:01, 1:07:03, 1:11:57

> I just would like mrtg to graph exactly the value I hand over to it, with
> no interpolation. Is this impossible?
Not in the current code.

> Would it be an idea to introduce an other option - lets say
> "nointerpolation" or so - that tells mrtg to do no interpolation? I think
> there must be a lot of other people who would like to graph e. g. their CPU
> or other pre-calculated values.

I think it's use would be fraught with danger.  The interpolated value
is, in 95% of the usage, what you actually want.  The other five percent
are flag - type issues, which mrtg really isn't the proper tool for.
-- 
Daniel J McDonald, CCIE 2495, CNX
Austin Energy


--
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