Hi Joe,
Joseph Fialli schrieb:
> Please review and provide feedback for the Glassfish v3.1 one pager for
> runtime clustering support (using Shoal GMS) which is available here:
> http://wiki.glassfish.java.net/attach/V3FunctionalSpecs/gms_gfv3_1_onepager.txt?version=6
can you please make sure that the GF V3.1 cluster implementation will
take into account the root cause of/the proposed fix for the scenario as
discussed in the following ias mail alias thread by Jothi and Nacho:
http://archives.central.sun.com/thread/5694445
In short: "When server1 gets up again and comes into contact with
server2 (which at that time is still up and running), it should not just
receive a copy of server2's own sessions, but also a copy of those
active sessions at server2 that initially came from server1 and have not
been fully taken over by server2 only due to the fact that there was no
request from that session hitting server2 in the meantime.
In other words, a server should inform its peer(s) not only about the
sessions it actively owns, but also about the sessions it would be able
to take over in case the next request for such a session were to hit
this server (and not the one which actively owned them)."
Many thanks & best regards,
Andreas
--
Andreas Loew
Senior Java Architect
Sun Microsystems (Germany)