[rrd-users] Question concerning Archives and creating graphs

Simon Hobson linux at thehobsons.co.uk
Wed Oct 24 22:53:30 CEST 2007


Patrik Grill wrote:

>I've created a rrd database with one datasource and two archives, 
>both of the consolidation function AVERAGE, but with different steps 
>and rows, like this:
>
>DS:input:GAUGE:600:0:U
>RRA:AVERAGE:0.5:6:700
>RRA:AVERAGE:0.5:2:300
>
>Now I want to generate a graph for each archive, but with:
>
>DEF:in=traffic.rrd:input:AVERAGE
>
>I always get the first archive plotted (RRA:AVERAGE:0.5:6:700)--- 
>how can I plot the second archive (RRA:AVERAGE:0.5:2:300) ?
>
>Or am I complete wrong in my understanding of the RRDTool ?

You have the wrong understanding.


The tools will use the best fit RRA that is available. So if you plot 
a graph that exceeds the time range encompassed in the second RRA, 
the first will be used. Also (not applicable in this case), if the 
start/end times fit one better than another then this will have an 
effect.

You should specify the start and end times as exact times which are 
multiples of your aggregate period rather than just "now" and "now-x" 
- eg "now - now % step" as end and "end - n x step" as start. If the 
steps weren't multiples of each other - eg you had used 2 and 5, then 
there would be times where the x5 periods would give a better match 
to a time than the x2 periods - eg from 5 to 15s would be exactly 
2x5, but be 1/2x2 plus 4x2 plus 1/2x2.


So taking your figures, always make sure your start and end times are 
multiples of 20 minutes (600x2 s), and end-start is less than 100 
hours (600x2x300 s) - then RRD tools should use the higher resolution 
RRA.



More information about the rrd-users mailing list