Configuring JMX data collection is not user friendly. Currently one must:
connect to JVM,
browse mbean server until target for collection is found,
copy-and-paste the name into the vi editor,
configure the RRD stuff,
then restart opennms.
Really? Does it have to be that cumbersome?
Browsing an mbean server on a discovered JMX server is a trivial task with many examples to be found.
Re-writing XML files is trivial.
Reading XML files is trivial.
Reconfiguring runtime config data beans with new information
— This one is part of my research into parts of the code base I don't normally look at. I'm hoping to find some kind of life-cycle management on configuration data sources. If not, this could get very interesting....
I think ease of use is key to adoption rates. Making the act of configuring JVM performance monitoring a point-and-click operation is likely to make for a compelling feature.
Once it's built, it might even be re-usable for SNMP data collection configuration (think: SNMP MIB browser).