After pulling my hair out for nearly a week, I found a working
solution to hosting a two-node setup that handles multiple overlapping
client subnets, while keeping clients uniquely identifiable in the
main application.
It's solved by combining the regular two-node lvs approach with VLANs
(or multiple interfaces), and recent netfilter/ipvs/network features
that allow the source client networks to be remapped. The main
application would then see clients coming in on unique subnets, so
there would have to be some reverse translation if you want to do log
analysis or accounting.
The full howto is posted here:
http://kb.linuxvirtualserver.org/wiki/Two-node_setup_with_overlapping_client_subnets
sven
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|