I'm trying to use jboss-profiler-2.0.0.Beta5 to profile an app I have on jboss-4.2.3.GA on SUSE 10.3. I have been able to run the profiler without the app using the default server config for example. However when I try to configure the server config for the real application that I have with the profiler the last line I get is the familiar "JBoss Profiler depends on external communication module" line. There's no "Starting JBoss" line and the VM exits so there's no process to thread dump.
Now if I remove the javaassist.jar file from the bin dir the app will start, but there's no profiler services.I also noticed that profiler 2.0 comes with a newer javassist.jar than the one jboss-4.2.3.GA comes with. Does there need to be another build of the profiler that works with JBoss 4.x?
Another difference between the profiling and non-profiling configs is that the non-profiling config uses the clustering and high availability jars while the profiling config does not. Does anyone know if profiler 2.0 works with the clustering environment?
Has anybody seen this strange early exit behavior from JBoss and solved it?
Message was edited by: John Aronson Upon more consideration ... When the profiler starts up it instruments code. One of the jar files in the fdposmgmt config / app was causing an unexpected error in the JBoss profiler instrumentation code. That was causing JBoss not to start. This explains why we would get a log file with only 2 log lines about instrumentation and no line that says “Starting JBoss”. When javaassist.jar was not available to the profiler it was not trying to instrument any code because it noticed that problem before trying to instrument. I have attached listings of the lib and deploy dirs of the non-working JBoss configuration because I suspect that one of those jars was causing the profiler to fail. Better error reporting during profiler instumentation would be helpful.