When testing AM6.3 patch12, session failover feature was reported not working. There is a NPE when checking whether a server instance is up or not. The amsessiondb process never received any READ requests. After some debugging, it appeared that the server ID of the Load Balancer (03, suppose 2 AM servers 01 and 02 in the cluster) have to be added to make it working.
However, in Deployment Planing Guide of AM6.3, it clearly states "Do not include the server ID of load balancer." in "Session Cluster Server List".
The root cause is that the person who set up the environment modified the value of "com.iplanet.am.localserver.host" in configure file AMConfig.properties when changing other properties "com.iplanet.am.server.host", "com.iplanet.am.console.host", "com.iplanet.am.profile.host" and "com.iplanet.am.naming.url". Properties "com.iplanet.am.localserver.xxx" should never change unless you have to modify the host name of the box.
Actually with localserver.host changed to the host name of LB, adding server ID of LB to the "Session Cluster Server List" could help session failover to work, but it may cause some performance issue, some unnecessary internal session routings.