[rrd-users] rrd at the ragged edge (updates every second)
Rick Jones
rick.jones2 at hp.com
Tue Apr 19 20:30:28 CEST 2011
On Tue, 2011-04-19 at 18:41 +0100, Simon Hobson wrote:
> Rick Jones wrote:
>
> > > AIUI, if you use "N" (now) for the update time, then internally RRD
> >> uses sub-second timing and you can update more frequently. It does
> >> mean that your update "happens" when you submit the call rather than
> >> when you take the timestamp, but for your setup I doubt that's a huge
> > > issue.
>
> > > you can input time with millisecond precision, then this will not
> >> happen
> >>
> >> 1303164367.002
> > > 1303164367.393
>
> >is there any additional subtle differences between the two I should know
> >about before picking one or the other?
>
> "N" means the time you make the call, nnn.nnn means a specific time
> which may be different to when you make the call. It's particularly
> important if the script that collects data and prepares the update
> takes a long time between collection and update (perhaps some complex
> conversions). If the script takes very little time to run, the two
> options are near enough the same in effect.
>
> When I wrote the above I wasn't aware you could use decimals in an update time.
No worries - at this point I have to go back and see more precisely how
ntop is using/manipulating time.
BTW, how long as "N" been supported and how long have milliseconds been
supported. If either/both go back "far enough" in time I may not have
to add checks in ntop's configure script and #ifdef the code.
thanks,
rick jones
More information about the rrd-users
mailing list