[mrtg] Sane handling of missing data and parse and access errors
resident at telcontar.net
Fri Sep 28 23:18:54 CEST 2018
One thing that has frustrated me for years -- when there is no data, MRTG has no single correct facility for handling this properly.
The options are:
a) Show zero -- no use if extended zero periods are acceptable, but fine for graphs like RAM/CPU/SNR that always have values
b) Hold the value -- fine for bandwidth where it should never settle, but useless for CPU/RAM that can remain fixed
I've figured, that it should be possible for the graphs to show something like:
a) Orange background for no data (e.g. read timeout)
b) Chequered background for read error (e.g. trying to read timeticks as a string, or invalid OID for that device)
(Or other colours/patterns as appropriate)
I realise that this can be set on a per-metric basis (I went through all our metrics the other day in the config generator and set them to zero or hold to better indicate this, but even so, I still think it would be clearer to see this presented on the graph).
I also feel that the HTML generator should note any parse/read/access errors on the HTML, instead of expecting you to notice that the absence of the uptime means that the host is down. (There is logging, but you shouldn't have to rummage around in a log to be able to read off an error message about access/processing.)
Am I mad, or does that make any kind of sense?
More information about the mrtg