Hi, I am seeing some anomalies in my graphs for weekly vs daily display of temperature measurements . I have measurements coming in on a per minute basis from the sensor, and the daily graph reports different max values when spiking than the weekly graph, which does show a spike but not nearly as high.I have an alert set upon reaching high temperatures and the daily graph seems to be displaying the correct values.
I am looking for clues towards this behaviour. Anyone?
Persistence strategy:
Strategies {
// for rrd charts, we need a cron strategy
everyMinute : “0 * * * * ?”
}
Items {
Temperature*, Weather_Chart*, Weather*, VCC_Chart*, PIR_Chart*, itm_uber7*, itm_node2* : strategy = everyMinute, everyUpdate, restoreOnStartup
}
rrd4j is a fixed sized database that “compresses” data as it gets older by culling the data and using averages of points thrown away as the new value. So as your data gets older the actual data gets replaced with sparser and less accurate data. That is probably what you are seeing here.
For relatively smooth data in a plot this wouldn’t be a problem. But because you have one point that is way higher than the rest the average causes a large impact.
Unfortunately the only solution is to use a different persistence engine such as db4o or InfluxDB.
