[rrd-users] 95% percentile value - varies with width of graph - how can I get an accurate value consistently?

[John Roberts]

Hi All

We use icinga as a monitoring tool.

We display graphs of switch/router/firewall interface data with the 95% percentile drawn as a line on the graph and the 95% percentile displayed in the legend. The graphs are generated by using pnp4nagios, which if I understand correctly, is using the rrdtool to extract the data and manipulate it.

Recently I found that 95% percentiles were showing on the graph as lower than we expected - especially when I increased the custom time span to more than a certain time period.

I believe that the cause of this is that "If the resolution of the data is higher than the resolution of the graph, the data will be further consolidated (http://oss.oetiker.ch/rrdtool/doc/rrdgraph_data.en.html).

I have found out that if I had the following "&w=X" to the url of the pnp4nagios graph, the 95% percentile changes. The 95% percentile continues to increase until the width is 6000 for the particular data in question for a 5 week period i.e. I assume that at this point the graph is wide enough for there to be no need for data consolidation.

Do you think that I am correct in my assumptions?

And if that is correct, can I use the rrdtool command line to get the 95% percentile value without any need of generating the graph? i.e. I could be sure that no data consolidation has happened.

Many thanks in advance for your help and please excuse any deficiencies in my knowledge of rrd/pnp4nagios.

Thank you.

John
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.oetiker.ch/pipermail/rrd-users/attachments/20160422/8c2f8a77/attachment.html>