[SNMP4J] snmp vs jmx

[Frank Fock]

Hi Torsten,

Torsten Curdt wrote:
> Hi there
> 
> I am interested in the current state of the project. Is 
> active/stable/dormant?
> 
Active and stable as stable as a 1.0 version can be.
This is statement is true for SNMP4J-Agent. SNMP4J
itself
There will be only bug fixes and enhancements in the
near future. There are no plans to change the API
significantly.

> We need to monitor a couple of machines and I am currently evaluating
> on how to do that best. For many reasons SNMP would be the first
> joice. On the other hand the JVM already exposes many information via
> JMX. How much work is left for the JMX-SNMP bridge?

I am currently developing an add-on API for SNMP4J-Agent
that facilitates integrating JMX with the agent instrumentation.
As a proof of concept I am implementing the JVM-MANAGEMENT-MIB
with that API and without it (thus directly with SNMP4J-Agent
MO* objects and AgenPro generated stubs).

> 
> I did hear thoough that such a bridge is expensive and not really
> desireable. How suitable would it be to expose JVM information and
> log4j configurations via SNMP. Is there code for that already? (I
> think I've seen some log4j stuff in the javadocs)

The log4j debug levels (and the J2SE 1.4 logging) can be
already changed/configured with SNMP4J-AGENT. See the
SNMP4J-LOG-MIB for details. I think this MIB is easy to
use. Comments are welcome. The JVM-MANAGEMENT-MIB is even
easier to use (less complex).

A JMX<->SNMP bridge is not as effective as a direct instrumentation
that directly accesses the MBeans, but I expect not a significant
performance loss.

> 
> Some comments would be great. If it's a viable way we may be able to
> help out here with some engineering time.

I will contact you when the basics of the API are ready.
Currently, I am doing more research than coding...

Cheers,
Frank