RMI server crashing the Virtual Machine

Question

I am writing an enterprise RMI service at work and it has been crashing at seemingly random points and in-frequently. It is not actually crashing as such, because I have added a shutdown hook and that is getting called before it exits. I know for sure that it is not a system.exit() call because I have prevented all system.exit() calls in the security manager. I also have put in very tight memory management and an extended heap so I don't think its a memory issue either. Frequent calls to freeMemory() show that there is no memory leakage and that everything seems to be being cleaned up correctly.
I am using the IBM 1.3 virtual machine on a linux box for the server and thus the calls to the garbage collector are working nicely and when requested.
The crashing is happening at seemingly random points and I have covered all my public methods with try....catch blocks that catch Throwable, and it doesn't seem to be an exception. Code that 19 out of 20 times will run fine, crashes on the 20th run etc.
What I am wondering is - has anyone had similar problems and do you know how to fix it?
My other thought is that perhaps it is a problem with the IBM virtual machine, and I am thus going to try running things through the Sun 1.3 virtual machine to see if I get any different results.
Any insights or help from people developing similar services would be greatly appreciated (this is the first RMI service I have implemented).

Answer 1

YES!!! I finally solved it..........
I put the monitoring code into the startup script - incidentally, on linux the shell variable is $? as opposed to $status.
Then I let it run and this morning it exited again and the error code was 130 - or Control-C.... This made no sense to me as the service was running on a linux box which I was monitoring through a telnet window on my Windows PC. And I know for sure that I wasn't pressing Control-C.
After talking to someone else at work I found out that they had heaps of problems running a resin server in a similar way (through a telnet window), and in the end it came down to the telnet window was sending things it shouldn't to the resin server and causing all sorts of problems. As soon as they ran it in the background, the problems disappeared.
So I ran it in the background and the problem has disappeared. Instead, to monitor the service I just did a tail -f on the log file and every half hour or so the tail program running through the telnet window was just dying mysteriously...... ha ha ha ha
Thanks a lot for your help EJP, I wouldn't have been able to sort it out without your suggestions. And I have learnt a lot more about RMI in tracking this bug down. Only problem is that the clients were here for a demo last week and had it exit once while they were running the client app......arg!!!