[rrd-users] Debugging graph smoothing or bad data samples? interpolation bug?

Marc MERLIN marc_rrd at merlins.org
Tue Nov 17 23:00:00 CET 2009


On Tue, Nov 17, 2009 at 10:25:48PM +0100, Tobias Oetiker wrote:
> Hi Marc,
> 
> Today Marc MERLIN wrote:
> 
> > On Tue, Nov 17, 2009 at 06:50:40AM +0100, Tobias Oetiker wrote:
> > > Hi Marc,
> > >
> > > > Does that make any sense to you?
> > >
> > > not entirely, I can only say that irregular data arrival times are
> > > rrdtools forte ... have you tried to write the data into a logfile
> > > (including the timestamp) and then recreate the phaenomenon by hand ?
> >
> > I have all the data in a logfile already, this is where I read it from :)
> > (script writes to logfile, cacti polls logfile from cron and feeds to
> > rrdtool).
> >
> > I have another script that will dump the entire logfile, generate a log list
> > of rrdtool update statements and I can recreate the entire rrd from scratch
> > after erasing it and replaying all the events.
> 
> ah ... in that case it is clear ... sometimes, you have one update
> in the file and sometimes two by the time cacti comes around ...a
> very bad case of jitter ... you should realy make it so that cacti
> can trigger the data accqusition ...

Mmmh, I see what you mean. Yes, now that I think about it, in a bad case
scenario, the cacti poller could hit a race condition and get the next sample
one period early, and see the same sample again.

That must be what happened, thanks for pointing it out to me.

On Tue, Nov 17, 2009 at 04:33:47PM -0500, Kevin P. Foote wrote:
> Alternative in line w/Tobi.. scrap cacti (unless you really use it for
> other things) and move to using your (obvious) scriptability to update
> the RRD on your interval. Then graph with something like drraw which 
> focuses on the graph not the acquisition of the data AND the graph.

So, I do use cacti for being able to generate graphs on the fly and zoom
in/out with a few mouse clicks. It's quite nice for that.
See:
http://marc.merlins.org/perso/solar/2009-11.html#Brand-One-Powermeter_-Solar-and-Power-Monitoring-with-Cacti-and-Real-Time-PG_E-Time-of-Use-price-calculation
All the graphs that you see, you can select a range with the mouse and zoom
in or zoom out with right mouse click.

All that said, cacti has an option not to update the rrd itself and let me
do it myself. This is likely the road I'll take to fix this race condition
problem.

Thanks for the answers,
Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems & security ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/  



More information about the rrd-users mailing list