Hello,
I have a rather simple LVS-NAT setup, that involves balancing UDP traffic
between a number of realservers. The requirement is, that each UDP datagram
should be processed independently, regardless of its source address / port.
To satisfy this requirement, I use one packet scheduling mode (--ops
argument to ipvsadm). This brings in a (maybe) strange side effect:
director stops rewriting the source address / port of the UDP datagrams,
that the realservers send to clients, which, in turn, greatly confuses the
clients.
Is this expected behaviour? If so, how one should work around it if it is
important to "hide" realserver addresses behind the VIP?
Thanks in advance,
Mark
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
I have a rather simple LVS-NAT setup, that involves balancing UDP traffic
between a number of realservers. The requirement is, that each UDP datagram
should be processed independently, regardless of its source address / port.
To satisfy this requirement, I use one packet scheduling mode (--ops
argument to ipvsadm). This brings in a (maybe) strange side effect:
director stops rewriting the source address / port of the UDP datagrams,
that the realservers send to clients, which, in turn, greatly confuses the
clients.
Is this expected behaviour? If so, how one should work around it if it is
important to "hide" realserver addresses behind the VIP?
Thanks in advance,
Mark
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://lists.graemef.net/mailman/listinfo/lvs-users