I'm currently attempting to instrument a standalone Eclipse RCP application. I can attach the Java agent no problem, I can put in a PBD file to capture metrics on certain classes and methods, and I can see things come out in the EM through the Investigator. Heap, CPU, File System, class metrics, all good.
But I would really kill for is the ability to get a transaction trace out of it. I appreciate that it's not a transaction in a more traditional sense, but the code hotspots and timings would be incredibly useful for tracking an intermittent performance issue, and I have a gut feel that CA APM should work for this, if I knew how.
I've attempted to force the agent to start tracing (all agents > 5 milliseconds for 30 minutes), but nothing gets generated and comes out. Closing the application obviously disconnects the agent.
Any suggestions as to what I can do to force a transactional trace snapshot to come out? If need be, I have access to modify the source code. But there's some points we're trying to investigate, and I'm suspicious that there's something in one of the pathways, so we'd really like to try and capture as much info as we can.
Thanks for any suggestions that might be had.