[rrd-users] Re: Bug with "MAX" RRA ?

[Alex van den Bogaerdt]

On Tue, Dec 30, 2003 at 02:58:48PM +0100, Paul Courbis wrote:

> I'm trying to setup an RRD database to kkep a track f "MAX" values of a 
> counter. My problem is that the values are transformed depending on the 
> timeslot alignement. That is, if I setup a RRD database receiving a data 
> every minute, the value is used unchanged if it's store on the beginning 
> of the minute, while it is transformed (average with previous value) if 
> it's entered some second laters.

That's the normalization process kicking in.

> here's a little shell script that produces 2 curves, one with data 
> stored on the beginning of a minute, the other 30 second laters. As I 
> keep "MAX" values, curves should be the same (in my mind at least) but 
> it's not the case.... Run this script, it'll create a "demo.gif" graph 
> showing the problem.

MAX has to do with RRA.  Rates are going to the RRA _after_ normalizing
is done.

Try your script after changing the following:

- change "step" from 60 to 1
- for each RRA, multiply the pdp_per_cdp value by 60

(hint:
 1*60 == 60*1 ,
 normalizing 1 second is different from normalizing 60 seconds)

> If anyone can explain...

I did, countless times.  Please find them on the web, archives, google

HTH
Alex

--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi