Hi.
I was buzzing through the clustering pdf and discovered the "farm" section. This sounds like it might solve another issue if I can get it to work.
After searching the clustering topic and all forums, I discovered that the documentation isn't quite current. I found a thread though that got me up using TCP rather than multicast. The nodes recognize each other when they come up.
However, dropping a war into the ./farm directory deploys it on the local machine but does not seem to transfer the war to the other node. I thought this is what it's supposed to do?
Our need is not to duplicate beans, but rather to just get the war distributed between two JBoss servers. I don't care about state of beans, just having redundant servers up at this point with a guarantee of the same version of war at both hosts.
One thing that I noticed is that when I added the tag to the web.xml suddenly the RAM usage skyrocketed, as did CPU usage. Both systems are Redhat 8, one has 512mb RAM and the other is a POS test box, 64mb with a P3-450. The test box, though, is fairly trim and is only doing JBoss.
Am I doing something wrong here? Does farming cause that much extra memory usage? What's reasonable? Does farming actually distribute a war someplace? What sort of system do I need for a minimum testing system, as a proof of concept? What sort of networking bandwidth, etc?
Thanks.
