At a first glance this seems to be an artefact of normalization in combination with missing data.
Can you try a step size of 1 second, still updating every 5 seconds, heartbeat as desired? ----- Original Message ----- From: "Matt Elias" <[email protected]> To: <[email protected]> Sent: Tuesday, April 08, 2014 3:23 PM Subject: [rrd-users] RRD Graphs with bad averaging of missing data? > Hi, I have an Ubuntu server (v12.10) running Perl (v5.14.2) scripts with > the librrds-perl (v1.4.7-2) package. The Perl scripts connect to an > Arduino and get temperature data from DS18B20 sensors. The shortest time > frame I'm graphing is 1hr, so for a decent resolution, I'm updating the > rrd database every 5 secs. Whenever a reading times out or the Arduino > hangs, after the updates resume I see the anomalies shown on my image at > 9:56-ish (see google driv link). I run other temperature rrd databases > that have a step of 2mins and a heartbeat of 4mins that just show a gap > when there's unknown data. For this rrd with a step of 5s, I've tried a > heartbeat of 5s, 8s, 10s, 15s, 120s & 240s and they all have this issue. > My Perl script also saves the Arduino data in a plain text file and when > I compare it to the graph, the plain text file shows no erratic numbers > that would cause the low spike, the data/time stamp just shows missing > data. So I think it's either the graphing or the rrd database itself > that is averaging an unknown value with a legit value and > storing/displaying an incorrect value. Has anyone seem this behavior > before? > > https://drive.google.com/file/d/0B-KBMnHT5sWjdEltYlM0eG5rZ2c/edit?usp=sharing > > Matt > > > _______________________________________________ > rrd-users mailing list > [email protected] > https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users > _______________________________________________ rrd-users mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
