[rrd-users] rrdresize on munin-created rrd files

[Tobias Oetiker]

Hi Jacob,

Today Jacob Alexandersson wrote:

> Hello all,
>
> I posted the message below on the munin userlist a couple of days back but I
> haven't gotten any response. Hopefully someone can help out here?
>
> //Jacob.
>
> --------------
>
> Hello all,
>
> I've been digging around a bit on how munin creates its rrd
> files. I don't claim to understand much on the inner workings on
> rrd, but this is what I've managed to dig up:
>
> >From my understanding, "rrdtool info <filename>" tells me that the first
> three rra's save one pdp per row, meaning that they don't average
> out anything. We have 576 rows, where the "step" is 300 seconds
> which works out to be 48 hours.
>
> So we have 48 hours of fully accurate data stored (nothing averaged out).
>
> I want "fully accurate" data to be stored much longer. Say 90 days.
>
> 90 days = 7776000 seconds. /300 gives me 25920 rows. So I need to
> add 25344 rows to each rra. This can be achieved by:
>
> $ rrdtool resize filename.rrd 0 GROW 25344
>
> ...where "0" points out the first rra.
>
> I could do this for the first three rra's, but what about the
> other ones? There is a set of rra's that store data for 9 days
> over 30min averages if my calculations are correct. This clearly
> falls within the span of my 90 days above. Will these rra's ever
> be used? Or should I modify these accordingly to something
> appropriate after my change above?

when rrd fetch (via rrd graph) pulls data from an rrd file it will
pull the data that BEST matches the resolution required to draw the
graph. This means given a weekly graph it would probably pull the
data from that 9 day RRA ... OTOH if you drew a graph over two
weeks it might fall back to your new 90 days RRA since this
provides better coverage.
>
> I guess what I'm really asking is:
> Is this the right way of going about this?

I would structure the RRD as follows:

RRA 90 Days at 1 step AVERAGE
RRA 90 Days at 6 step AVERAGE
RRA 90 Days at 6 step MAX
RRA 90 Days at 6 step MIN

using MIN and MAX at 1 step makes little sense

cheers
tobi


>
> What I am looking for is to store "fully accurate" data for a
> much longer time than the (munin) default of 48 hours.


> //J
>
> ---------------
>

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi at oetiker.ch ++41 62 775 9902 / sb: -9900