[rrd-users] the usage of CF

Alex van den Bogaerdt alex at vandenbogaerdt.nl
Thu Sep 24 07:37:29 CEST 2009

>>> I have question using CF MAX and LAST. I have 80000 data point for each
>>> second.
>>> And I plot a chart with the step size 1. However, I found the chart 
>>> using
>>> MAX CF and LAST are different.
>>> Does anyone know the reason?
>> Step size 1 means 1 second.  You claim to have 80,000 datapoints per
>> second,
>> I'm only going to show 5 per second. This is what happens when 5
>> datapoints
>> are consolidated into one:
>> LAST(10,20,30,20,10)==10
>> MAX(10,20,30,20,10)==30
> I am sorry. 1 data point is for one second. That means there are total 
> 80000
> seconds data.

And how many of those are you showing in your chart, and how wide is that 
Consolidation may still happen, at graph time.

> As the step size is 1, I expect the MAX and LAST should be the same.

No. See my previous remark. Or it is normalization which kicks in. This 
will, with step==1, only happen if you do not use timestamps to update, like 
so: rrdtool update your.rrd N:12345
(I am assuming that not only step==1 but also steps==1, 1 step per RRA row)

In such a case updates may not be exactly on the second, it may for instance 
be 07:30:00.05

See my site, http://www.vandenbogaerdt.nl/rrdtool/ , item Rates, normalizing 
and consolidating

First make sure you understand what is written there.  If this still does 
not help, write a small script which:

1: creates a small and simple database
2: updates a couple of times
3: creates a small and simple graph.

This script should show the behaviour which you are experiencing, there's 
really no need for many DS's, a gazillion updates and a fancy graph.


More information about the rrd-users mailing list