J
JStonebracker
Hello,
I'm having a very strange problem in one of my applications. This is
what's happening:
A few JVMs crash without giving any clues as to what is happening. No
error messages, no stack traces, no core dumps. Nothing. Nada. They
just disappear.
I'm running on Windows 2000 and I had the same problem using JDKs
1.3.1_07 and 1.3.1_08. Sometimes I also notice a Windows error dialog
say that the "Application log is full" but I'm not sure it is related
(because when the problem happens and the dialog is NOT there we can't
see any entries in the application log).
The application is a JBoss server process which is part of a cluster.
In the process we have a RMI application running. The nodes that crash
are the nodes where this RMI client/server is deployed. When the app
server process starts we bind the RMI server stub in to the JNDI tree
and this server object communicates with a number of different
applications.
I suspect that this could be related to RMI (distributed gc, maybe?)
or the fact that this service running within the JBoss process is
multi-threaded (relies heavily on waits and notifies). But I don't
really know. The process simply disappears.
We have deployed the RMI service in different machines to rule out the
possibility of hardware failure but the crash happened anyway.
We have a memory monitor that checks the memory usage every 30
seconds. The logs show that the memory usage should not be a problem.
It is possible, though, that we have a memory spike that takes less
than 30 seconds to show up and crash the system in such a dramatic
way.
Has anyone here ever experienced or seen anything like this?
Thanks in advance for any help.
JS
I'm having a very strange problem in one of my applications. This is
what's happening:
A few JVMs crash without giving any clues as to what is happening. No
error messages, no stack traces, no core dumps. Nothing. Nada. They
just disappear.
I'm running on Windows 2000 and I had the same problem using JDKs
1.3.1_07 and 1.3.1_08. Sometimes I also notice a Windows error dialog
say that the "Application log is full" but I'm not sure it is related
(because when the problem happens and the dialog is NOT there we can't
see any entries in the application log).
The application is a JBoss server process which is part of a cluster.
In the process we have a RMI application running. The nodes that crash
are the nodes where this RMI client/server is deployed. When the app
server process starts we bind the RMI server stub in to the JNDI tree
and this server object communicates with a number of different
applications.
I suspect that this could be related to RMI (distributed gc, maybe?)
or the fact that this service running within the JBoss process is
multi-threaded (relies heavily on waits and notifies). But I don't
really know. The process simply disappears.
We have deployed the RMI service in different machines to rule out the
possibility of hardware failure but the crash happened anyway.
We have a memory monitor that checks the memory usage every 30
seconds. The logs show that the memory usage should not be a problem.
It is possible, though, that we have a memory spike that takes less
than 30 seconds to show up and crash the system in such a dramatic
way.
Has anyone here ever experienced or seen anything like this?
Thanks in advance for any help.
JS