You should be able to use both scenarios BUT.
The director must have an active interface on the subnet that the RIP is
on (because it needs to inform the switch that the dest MAC address has
changed.
The RIP must have a default route to the client i.e. via firewall or
router (not via the director.)
Just Marc wrote:
Hello,
I'm trying to setup IPVS-DR (because of its scalability promise)
rather than NAT, when setting up the VIP to be on the same logical
subnet as the RIP everything works perfectly.
Now I'm trying exactly the same setup and machines but this time my
VIP is an external, Internet routable address and the RIP remains the
same, belonging to an internal subnet. Unfortunately this doesn't
seem to work, the client connecting to the VIP from the outside world
gets a TCP reset -- this reset is being sent from the real server.
The ldirector machine counts each connection as an InActConn.
Interestingly, if I use netcat to listen on the real server for the
VIP address, I get a full TCP connection.
In both of these setups I'm currently only using one real server to
simplify things, on the real server the loopback interface is set with
the VIP and ARP workarounds are turned on.
I also tried making the RIP be on the same Internet routable subnet
that the VIP is on, that also produced the same TCP reset.
As for versions, I'm using kernel 2.6.16 and the latest heartbeat
debian unstable package, 2.0.5.
Any help or advice or hints are greatly appreciated!
Thanks!
_______________________________________________
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://www.in-addr.de/mailman/listinfo/lvs-users
|