[rrd-users] Optimize RRD update

Fabien Wernli wernli at in2p3.fr
Fri Feb 1 10:13:59 CET 2008


On Fri, Feb 01, 2008 at 09:43:31AM +0100, Gwenael Lahay wrote:
> More over, is it possible to replace (by default) NaN values with zero for
> the period you don't update a rrd.

Well, you could use a very large heartbeat.

> (especially when there is no data and that i need top
> update all the files with zero).

why do you bother?

> If i could use zero instead of NaN, it will allow me to only update the rdd
> files with data. It will take less time.

I don't understand why you'd want to do this. I think you have a design
issue in your code or misunderstand something. It is preferrable to have
"Unknown" i.e. "nan" because it corresponds to reality: no data. How
would you tell a real, measured "0" value from an "Unknown" value if
all Nan were 0? (especially when your minimum value is 0):

> DS:${name}:${type}:300:0:U

Don't bother with this in your update script, handle the nans in your
display/graph/export script. You can set the Nans to 0 when graphing for
instance if you really want.

More information about the rrd-users mailing list