I have three IPv6 Tunnels, one with SixXS, one with HE and the generic 6to4 tunnel. That might seem pointless but what I am really trying to do is simulate how a deployment with multiple native IPv6 ISPs would work. My goals are to load balance sessions across the different connections and provide fault tolerance. The clients are end users that do not host services and do not care what their public address is. With IPv4 and NAT this was very easy; the router had the power to route outbound traffic to any interface it chose. How would I do the same thing with IPv6? The problem I face is that each host on my local network has 3 IPv6 addresses, one for each tunnel prefix. I setup a routing map to send traffic through the appropriate tunnel based on the source address. However, I seem to be at the mercy of whatever source address the host selects. Even if I shutdown a tunnel the hosts still tries to use its address (and fails). I am using a Cisco router, but this is more of a theoretical question than a question of how to configure the router.