Hi Ashu, thank you very much for this nice overview. As you have described a lot of details I'm quickly going to understand that if I do have legacy middleware (based on CORBA/EJB using RMI-over-IIOP) which uses TCPIP and I would like to put its server side on Sun Cluster, then several things needs to be done (I'm talking here only about fail-over services): 1) modify middleware server side to communicate its state from master to slave(s) 2) modify middleware client side to "survive" getting "conn reset by peer" and try to reconnect transparently for defined amount of "conn reset by peer" in the defined amount of time (i.e. to correctly detect "real" conn reset from just master -> slave fail over) 3) modify server side to provide some "agent" hooks for SC. Am I right that in this case not only failure of hardware (NIC/CPU/machine) would be detected, but also general failure of the software (crash of the server)?
Is that everything or is there any detail I've overlooked so far? Thanks! Karel -- This message posted from opensolaris.org
