[rrd-users] Question about RRDTool ..

[Jean-Yves Avenard]

Hi

Thank you for your answer !

On Fri, Mar 21, 2008 at 7:42 PM, Tobias Oetiker <tobi at oetiker.ch> wrote:
>  yes sure, it depends on the definition of exact though ... what you
>  seem to be looking for is the 'step' maximum. So if you keep an rra
>  with consolidation method MAX and use that as a basis for
>  the max value in the graph, you will get what you expect ...

Doesn't seem so ; I have an RRA defined with the maximum definition
(discrete value collected every minute for one year).
rrdtool create modem.rrd -s 60 \
DS:maxup:GAUGE:600:0:22000 \
DS:maxdown:GAUGE:600:0:22000 \
RRA:AVERAGE:0.5:1:527040 \
RRA:AVERAGE:0.5:60:744 \
RRA:AVERAGE:0.5:10080:4 \
RRA:AVERAGE:0.5:44640:12 \
RRA:MIN:0.5:1:527040 \
RRA:MIN:0.5:60:744 \
RRA:MIN:0.5:10080:4 \
RRA:MIN:0.5:44640:12 \
RRA:MAX:0.5:1:527040 \
RRA:MAX:0.5:60:744 \
RRA:MAX:0.5:10080:4 \
RRA:MAX:0.5:44640:12


However, when I ask for the max value in a graph where one pixel isn't
equal to the step (here it's 400 pixels = 24 hours), it is
"normalised" even the min and max value and this doesn't correspond to
the data recorded. To get the non normalised data I must have one
pixel = one step (or a multiple) to get the correct maximum and
minimum

IMO, from a mathematical point of view, it makes no sense to normalise
the data when calculating the minimum or maximum.

JY