In a clustered environment, under some circumstances (for example one server in a cluster crashes or is killed), EJB client can receive an exception instead of a response.

EJB client library can respond automatically by retrying invocation on some other node in the cluster. However, this happens only for NoSuchEJBException.

Please document, for which cases the EJB client library retries the invocation automatically, and for which cases it does not. In these cases, please describe best practices - how should ejb client code respond to such a case.

Please ask pferraro, rachmato@redhat.com, rhn-engineering-rhusar, dlloyd@redhat.com for the content.