[rrd-users] Adding aberrant detection to existing RRD files

Eduardo M. Bragatto eduardo at bragatto.com
Wed Nov 28 01:59:28 CET 2007

Livio Zanol Puppim wrote:

> Your calc is wrong... After subtracting the old step from the new one, 
> rrdtool makes a division with the step number used in the "rrdtool 
> create" command, to get the bps(bits per second) rating... This is used 
> if the COUNTER DS type was used in the create step, so after getting the 
> "OID counter" (74335473981897) you need to multiply by 300 to get the 
> value that you want. To get RRDTool to not make the division you need to 
> use ABSOLUTE, but then, you will have only BITS as the unit 

After Bitman gave me the tip for the rrd_hwreapply I compared my code to 
that and realized the value recorded by COUNTER should be multiplied by 
step, which made a lot of sense and made me feel dumb. :-)

BTW, rrd_hwreapply only worked for me because I wanted only the RRA with 
1 step resolution. Be aware that any other RRA with more than 1 data 
point per row (in my case, 5min step sample), will be stripped out after 
"conversion". That script works in a completely different way than MRTG 
when converting from rateup to rrd (which preserves all RRAs).

The idea behind rrd_hwreapply is to allow you to experiment different 
parameters on the same data sample, which helps a lot when tunning the 
aberrant behavior detection for certain graphs.

Now, my new question is: is there a way to insert values directly into 
specific RRAs that have more than 1 PDP per row? In better english: is 
there a way to insert a value directly into the "weekly resolution" RRA?

I said in the beginning I did not need to preserve data that old, but 
I'd like to it if I could. ;-)


More information about the rrd-users mailing list