[rrd-users] A reason there isn't a CF SUM function?

Tobias Oetiker tobi at oetiker.ch
Wed Aug 22 09:51:44 CEST 2012

Today Simon Hobson wrote:

> Tobias Oetiker wrote:
> >you can always 'convert' the unknowns to zero first using an
> >appropriate CDEF
> Does that work when the data has been consolidated ?
> Eg, suppose you have 10, 10, 10, ... unknown, 10, and consolidate 10
> values (9x 10, 1x unknown) into one CDP. Do you get 10 (average of
> the 9 known values) or 9 (average of the known values and assuming 0
> for the missing one) ?

you get 10

> The problem is that by the time you come to graph the data, you only
> have the consolidated average to work with, and 10 would give the
> wrong answer - as could 9, it depends on your requirements.
> If the missing data is because we've had the power off, and so had a
> step period where nothing happens, we'd expect to have done/moved 90
> <somethings> during the consolidation period, not 100.

if you are not updating the rrd file, the data is simply unknown,
there is no helping it ...

if you can update the rrdfile, you could add a special DS which you
update with 1 when you got data and with 0 when you got no data,
then this DS will always reflect the 'availability' even through
consolidation ...


Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi at oetiker.ch ++41 62 775 9902 / sb: -9900

More information about the rrd-users mailing list