Problems with JBoss client on Linux, and server on Solaris

[tysmeister]

Hi

I am responsible for a site that is experiencing problems when an RMI based
client (Tomcat JSP/Servlet 4.0.4 engine running on Linux) attempts to
connect to a JBoss 3.0.3 instance running on Solaris 8. More often than not the connect
request block waits until timing out. However when this behaviour occurs the
JBoss instance is subsequently unable to accept connections from any other
client on the network, despite apparently still listening on the JNDI port.
It is almost as if the JNDI listener thread in the JBoss instance has
crashed. The only option is to restart the JBoss instance.

The VM on Solaris is 1.3.1-b24, and the VM on Linux is 1.3.1_08.

Anyone have any thoughts?

Thanks and regards,
Andrew

[adrian.brock]

I can think of a number of reasons.

1) JNDI has deadlocked
try kill -3 {jboss.pid} to get a threaddump
and look for a deadlock

2) There is some problem with the network.
Host configuration, dns, is your server multi-homed?

3) You are running out of network connections
what does netstat show?

The timeout actually comes from a UDP broadcast
when the initial connection attempt fails, it tries
to find any jboss server on the network.

One obvious question, have you tried a later
version of jboss. I remember some improvements
to JNDI around 3.0.4/3.0.5

Regards,
Adrian