I have the following Situation:
I'm calling EJBs on a Websphere Application Server 6.1 from a Servlet running on an external WAS 6.1 - that works fine.
For security issues, I have a second application server, which is a clone of the first one where the EJBs are arranged.
If the first application server occurs errors so that the EJBs are not longer reachable, I have to switch to the second application server (backup).
This switch happens automatically in the source code of the application running on the external server, which calls the EJBs.

Now I have two failure situations:

Note: Both situations are exactly at this time, when the external server calls an EJB and the ejb-processing is still working, e.g. the EJB makes a longer database call!

1. Stop the first EJB-Server with kill -9:
In this situation I have the possibility to respond to the errors (timeout) immediately and make a switch to the backup server.

2. remove network cable from the machine, the second application server is running:
In this situation I do NOT have the possibility to respond to the errors immediately.
I only get a timeout from the TCP-Layer after 180 seconds, on which I can react and switch to the backup server - this is too long!!!

My question to the second situation:
Is there a possibility that I can switch to the backup server before I get the timeout-response of the "killed" EJB-server-machine?
Perhaps there is a possibility to configure any one of the involved application servers to do the correct handling?