Arun >If one of the nodes freezes(like powering down the >node)the TAF clients are taking a very long time to >failover to the working node almost i can say around >10 to 15 mins I don't know if it's relevant for your problem, anyway, with LOAD_BALANCE=ON I was able to reproduce at will the following problem by one of my customers a couple of years ago (was 9.2.0.4, I think...). Let's say we have two nodes. Node A is much more loaded than node B (I know, it should not happen since LOAD_BALANCE=ON, but we work with real systems...). Now, when node B crashes (e.g. powered off...), the new connections should go to node A. But, the listener on node A is still thinking that node B is there AND is less loaded. Therefore they are forwarded to node B... Since node B is not up and running the client gets a nice error! We find only two methods to solve this problem: - don't use LOAD_BALANCE=ON - execute "lsnrctl reload" on node A Therefore, IMO, you should do some tests without LOAD_BALANCE=ON. HTH Chris -- //www.freelists.org/webpage/oracle-l