On Fri, 2007-04-20 at 05:51 -0500, Volk,Gregory B wrote: > I'm using mrtg to plot the number of users associated with wireless > access point over time. Even though the SNMP var is an integer, I've > noticed non-integers in the data when looking at the output from > "rrdtool dump", and "rrdtool fetch". It appears that there is some > averaging going on inside the RRD file
Correct, rrdtool always regularizes the data to n-minute norms, assuming linear regression. The goal is to determine what the data would have been at precisely 06:15:00, rather than at 06:16:23. The benefit of this regularization is that your polling times can be a little off (maybe you poll at 06:20:13, but then a problem with another host brings in the next value at 06:28:15) but the stored data is predictable as to the timestamp on the data. > even though the GAUGE variable type was specified when the RRD file > was created. It really doesn't make sense to do that in this case and > fractional numbers of users (for the Peak and Current values embedded > in the graphs) are causing the users who look at the data to call me > and ask how 5.8 users can be attached to an AP. > > Is there anything I can do to disable the gauge averaging in this > case? I don't believe so, but I'm not an rrdtool expert. > I don't quite understand why it happens - gauge data doesn't really > seem average-able. I set up a quick & dirty script (outside of MRTG) > to verify what was happening and this is what I'm seeing. I don't know > if it has always been this way or if this is the result of the version > 3 rrd format. After five years of doing this you'd think I would have > noticed it before now. ;) mrtg/rrdtool has had this behavior for at least the past 8 years. There used to be a great FAQ on the topic. -- Daniel J McDonald, CCIE # 2495, CISSP # 78281, CNX Austin Energy http://www.austinenergy.com _______________________________________________ mrtg mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
