> Hi everyone;
> I have a data source that occasionally provides a NaN.
> This is OK for us, it's expected about 20 times a day or so.
> What we'd like to do is to eliminate "gaps" in the graph data, by
> using a technique like drawing a similar value as the last good poll.
> Is there a way we can do this?
> Is there a selector that can tell rrdgraph how to interpret NaN ? Or
> would the mechanism be in the creation of the rrd?
Can't you just increase the heartbeat to permit a longer time without
data before a NaN is recorded? Or is the Nan's dou to other reasons than
too long time between data samples?
I am very much convinced that a CDEF will allow me to graph a NaN datapoint using a number from the last-known-good value that we have seen (as defined by a range, say <50000 ).
I am , however, not an expert on RPN (Reverse Polish Notation).
This seems like a CDEF that many people would like. Sometimes counters do not rollover at predictable 32-bit or 64-bit positions, and in my case, a counter can rollover at any random spot, several times in a day.
I would appreciate some help from the community on coming up with a CDEF how to present NaN data as a continuation of the same last-known-good datapoint.
Tobi-wan, if you're out there, I could really use your help right now!