Currently, if the Keel-server loses it's connection to the JNDI or
JMS servers, it fails and exits. We need to be able to configure a
"set" of JDNI/JMS servers, so that if the connection throws an
exception, the server will try the "next" JNDI/JMS server, if any,
or sleep and retry the only one it has a set number of times before
giving up. Ideally, a failure to reconnect should be logged to a
special logger set up to email the cluster sysadmin, rather than
just going silently into the night...
The client is a bit better - it will try a rotating set of servers, but a
request made will throw an exception back to the user, then the
*next* request will try the next server. It should try all available
servers before throwing the exception, hiding the failure from the
user if at all possible.