... for its connection back to the originating broker when using multicast discovery?
I keep running into problems when the discovered broker attempts to connect back to the originating
broker using a domain-name that the discovered broker can not resolve properly. In some cases
I can set -Djava.rmi.server.hostname on the originating broker to fix this, though I couldn't
get that to function on an Ubuntu 11.04 using OpenJDK, no clue why.
It would almost be better if it simply used the discovered brokers IP address and/or as part
of discovery included all of the addresses/domain-names it was listening for connections on
so that the discovered broker could give others a try if the preferred address failed... er
something.
Or, and I have no idea how/if this would work, there was a way to effectively use duplex connections
with multicast discovery. Then only the originating broker would need to make a connection,
removing issues with figuring out what the right domain-name/address would be for the discovered
broker to connect back to? Is that even possible with discovery? Or would that effectively
create 2 separate channels for messages to flow on between 2 connected brokers?
--jason
|