I would be interested how you made the two servers (Linux and Win2k) notice each other.
I have a similar setup but whatever I do, the JBoss instance on Linux doesn't seem to "see" the one on Win2k and vice versa. They simply don't cluster.
However, if I start an instance on my "old" Win98 box, it happily joins the default cluster running on my Win2k box.
Any light in your setup would be more than welcome.
P.S. Yes, I have entries in my routing table on both machines (18.104.22.168, mask 240.0.0.0) and my router also supports UDP, since the two Win boxes are also hooked up to the same LAN segment.
Well, the fact is I did not do anything special :)
My linux install is from Linux Mandrake 8.2 with no special routing, same on the W2K boxes. They are on the same LAN 192.168.xxx.xxx, the linux box gets a dynamic IP and the W2K boxes have both a static IP. They all use Jboss 3.0.0 + Tomcat 4.0.3 (and sun jdk 1.3.1_03).
This is probably a Win2K issue with JavaGroups. All other WinXXX systems work okay (I didn't test WinME though), but Win2K fails when attempting to create a socket and setting options.
Sacha and I are looking into the issue and hope to have a fix soon.
You can cheat by using different JavaGroups properties, such as TCP.
I am seeing the same error on Solaris with all JBoss instances running on the same machine. I am configuring two seperate clustering partitions, one for EJB clustering and one for JNDI clustering. Some servers participate in both partitions and some servers only participate in the JNDI partition. I am using jboss_3.0.1RC1_tomcat-4.0.4. The two partition configurations are attached.
At present the error does not seem to cause a problem in behavior, but it is a troubling message to see in the log files.