Well, I've got our local install of mrtg running with the non-cifs bits of filer-mrtg (I'm using an SGI for this monitoring task and already had an mrtg setup in place...so filer-mrtg was bit narrow-focus for my taste).
Basically I lifted the first three parts of the filer-mrtg config file and run mrtg out of cron with them, for each of the filers I look after.
I'm looking at the whole MIB for other useful things but my understanding of SNMP is still maturing. mrtg doesn't seem to be the right tool for some of the things I want to do though, along with the fact that MIB doesn't quite seem as simple as it could be.
I know people will want me to justify that last statement, so here goes:
There are raw figures for the number of, say, NFSPROC_READ calls since the stats were zeroed, AND there are percentages for the number of NFSPROC_READ calls against total NFS ops since stts began. This seems to be unuseful as it can be worked out from the raw figures, and SNMP was always meant to provide the raw data with the least processing at the agent side...
What it practically means to me is that the MIB is large and I'm not very good at reading it. SNMP's umpteen different representations and translations are also causing me grief.
If anyone's done more than rudimentary things with SNMP and mrtg, or anyother free snmp stuff, and feels like trumpeting their genius I promise to be suitably impressed when you show us how to do it properly! 8)
Last time I counted there are 533 basic data (datums would convey the meaning better, but isn't a real word ;) in the MIB and some are only interesting when you perform some arithmetic on them and graph the results, and mrtg is confusing me as to how to get there from here. Probably my fault.
Hey ho, back to the grindstone.