[SNMP4J] creating MIBs from (M)Beans

[Torsten Curdt]

> Often the resulting MIB is not quite usable. MBeans have very
> few restrictions regarding their interface and their relations
> whereas SMI is relative restrictive.

True

> Particularly because of
> the lexicographic ordering of SNMP/SMI objects and the limited
> bulk transfer capabilities compared to JMX, choosing the right
> INDEX for SNMP tables is essential.
>
> By automatically generating SNMP MIBs from MBeanInfos, one
> will not get an useful MIB specification in most cases.
> Companies that provide such automatic mapping often conceal
> these "details".

Could you give an example of the shortcomings? Essentially you
should be able to provide similar information as through direct JMX.
...with a few limitations - but I think simple scalars and simple tables
should cover 90% of our needs.

I also thought to "enrich" the MBens and give hints trough annotations
to help creating a better MIB.

The problem I see is that we have a big team of developers. They are using
Spring and exposing a bean via JMX is just a piece of cake
...basically it is just
a given. If they all have to communicate with the person taking care of the
monitoring and defining the MIB it will become a big pain ...keeping the mapping
and the stubs all in sync etc.

So the idea was to create the MIB from the metadata and eliminate at least
some of the manual process. Of course I am talking about non-standard
application level monitoring here.

At the moment I am still failing to see the shortcomings of this approach but
given the fact that you guys know probably more about SNMP and JMX than
I do.... please enlighten me ;-)

cheers
--
Torsten