Losing RMI-IIOP connection at server end

Question

Hi all
I have a system running on Solaris 9, JVM 1.4.2_06 with a webapp hosted in Tomcat 5.5 connecting over RMI-IIOP to another java service on the same machine. The connection is established at webapp initialisation via a lookup to the naming service (currently tnameserv).
Unfortunately, after a period of time (varies between a few hours and a couple of days), the RMI connection simply disappears from the server-side point of view, and the webapp grinds to a halt. Taking a thread dump of Tomcat and the service at the point of failure shows that Tomcat still has "JavaIDL Reader" threads listening to the service, but the service has lost all of its "JavaIDL Reader" threads listening to Tomcat.
I've traced the problem as far as line 518 of com.sun.corba.se.internal.iiop.messages.MessageBase:
bytecount = is.read(buf, offset + n, size - n);This read on the InputStream is throwing an IOException. I have configured the system so that IP address 127.0.0.1 is being used in all places.
It may well be that this is not the right forum for this question, but has anyone else ever come across a problem like this? All suggestions most welcome.
Regards
Brian

Answer 1

In trying to find an answer to your question, have actually discovered that the root cause of the issue is actually a little lower in the stack.
SocketInputStream.read(byte[],int,int) is returning -1 (EOF), and this is being interpretted into an IOException by the MessageBase class. So now I need to work out why this unexpected EOF may be occurring on this platform.
I imagine this is no longer very relevant to the RMI forum. I've posted on the Socket Programming forum.
Regards
Brian