[rrd-users] Graph discontinuity or jump

[Aleksandr Pasechnik]

Hello all!

I have an rrd collecting meteorological data. It has a step of 300 seconds and data is fairly reliably entered every 150 seconds. There are times when my computer is off and no data is entered. Explicit U's would be entered if there was a problem getting the data, but I haven't encountered that yet. The strange thing is that I have been getting these big, unexplained jumps in the graphs (with corresponding strange numbers in the rrd dump) after blank periods. The parameter in question (it happens to the others sometimes, but for some reason this last time it only manifested in this one) is the atmospheric pressure. It is stored as a GAUGE with a heartbeat of 600, a min of 0 and a max of U. I've got some RRA's, namely Average, Min, and Max at xff 0.1, 1 data point, and 7340 rows (30 days of data i think); and xff 0.5, 12 data points, and 44040 rows (5 years of data). I'm currently plotting the last 24 hours of data with a resolution of one pixel per data point. Before the blank occurred, my script was writing 1015.8 into the pressure. Once the computer came back up, it was entering 1012.8 and 1012.4 (a bit of sensor noise between measurements i suppose). At no point did i write a 1008.9 or whatever its plotting there. Here is the offending graph output:

http://russkey.tumblr.com/post/4971101013/jumping-rrdtool

Any thoughts why this is happening? Anything I can do to stop it? I thought it might have something to do with the xff, so i changed it from 0.5 to 0.1 for the 1 data point per row RRA's but that doesn't seem to have helped. Though now that I think about it, perhaps it would make more sense to have just a single RRA for the 1 data point per row section, and have it a Last instead of an Average? Any help at all would be appreciated!

Al