The only thing that jives with the introduction of a load balancer is that the session is not getting populated correctly on the jb2 instance that was not being accessed in the first testcase. When the jb1 instance comes back up it is not receiving a completely valid session from jb2.
-- xxxxxxxxxxxxxxxxxxxxxxxx Scott Stark Chief Technology Officer JBoss Group, LLC xxxxxxxxxxxxxxxxxxxxxxxx
Joao Clemente wrote:
Well, I can't explain why, but I've setup my cluster with the a loadbalancer in front, instead of interacting directly with one of the jboss servers, and this way I don't have an exception... If someone has followed the problem I described here, I have an hipoteses for what was happening, but I would like it validated:
By using a load balancer, when JB1 goes down, the requests are directed to JB2 wich is able to provide updated references to the user stub (by piggybacking). When JB1 comes up again, those references are again updated with a pointer to a NEW AND DIFFERENT ejb stub in JB1... and that is why ot works!
What could be happening was that JB1 was goind down, but the user stub kept the reference for it, one that would become invalid. When JB1 came up again, it would create a new ejb with the correct session values but whose reference was different. User then tried to interact, the user stub would see JB1 up and try that old reference (invalid). I don't know why at that time it would not failover to JB2 but...
I'll try to validate this using the java application that I got working, but using a longer time separating 2 interactions, so long that it would enable for JB1 to go down and up before a connection got redirected to JB2..
------------------------------------------------------- This SF.Net email sponsored by: Parasoft Error proof Web apps, automate testing & more. Download & eval WebKing and get a free book. www.parasoft.com/bulletproofapps1 _______________________________________________ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
