[rrd-users] RRD Graphs with bad averaging of missing data?

Alex van den Bogaerdt alex at vandenbogaerdt.nl
Wed Apr 9 16:36:15 CEST 2014


----- Original Message ----- 
From: "Matt Elias" <eliaschief at gmail.com>
To: <rrd-users at lists.oetiker.ch>
Sent: Wednesday, April 09, 2014 3:49 PM
Subject: [rrd-users] RRD Graphs with bad averaging of missing data?


> Year;Mon;Day:Hour;Min;Sec:Bin;Sensor;Value:Bin;Sensor;Value:etc
> 2014;04;05:09;55;03:120;1;331:120;2;205:120;3;360:120;4;297:120;5;431:120;6;1306
> 2014;04;05:09;55;07:120;1;331:120;2;205:120;3;360:120;4;297:120;5;431:120;6;1306
> 2014;04;05:09;55;13:120;1;331:120;2;205:120;3;360:120;4;297:120;5;431:120;6;1306
> 2014;04;05:09;56;25:120;1;331:120;2;204:120;3;360:120;4;296:120;5;431:120;6;1306
> 2014;04;05:09;56;30:120;1;331:120;2;205:120;3;360:120;4;296:120;5;430:120;6;1306
> 2014;04;05:09;56;34:120;1;331:120;2;205:120;3;360:120;4;296:120;5;430:120;6;1306

Updates are okay until 09:55:13,  then there's a gap, then the next update 
is 09:56:25.
Given your step and heartbeat settings, I think this should result in:
the interval from 09:55:10 to 09:55:15 is defined, as more than half of it 
has known data.
the interval from 09:55:15 to 09:55:20, as well as from 09:55:20 to 
09:55:25, is undefined because of the heartbeat setting. It should contain 
"NaN" or similar.
09:55:25 to 09:55:30 has data again.

Please check this by dumping your database (rrdtool dump) and examining the 
file.

Also please verify that RRDtool does indeed get its input exactly 
corresponding to what you just showed here. By that I mean that the 
timestamp is computed from those numbers, and no extra updates happen (e.g. 
for the missing update at 09:56:20, inserting zeros).

If this is all well, there could also be an issue when graphing. Are you 
using STACK somehow?  I ask because there seems to be no immediate relation 
between the numbers you show as rates, and the graph. Also the anomalies 
have the same shape, with different positions on the Y-axis.



More information about the rrd-users mailing list