Each WLS Server has it's own copy of the JNDI tree. Only Cluster Aware Stubs (Such as Queues and EJBs that are marked as such) are replicated. And when this happens, each server still has their own copy of these objects. Just that the implementation of these objects knows how to get a hold of the other instances of these objects in the other servers.
Originally posted by Senthil Kumar SS: In our application we bind an HashMap to the InitialContext. My doubt is will this be replicated to the available managed servers when the current managed server goes down.