We have a pair of Ruckus (Brocade) ICX 7750s that we use as our core routers. They currently handle a BGP connection with our ISP and then use OSPF for internal routes. Our ISP is finally going to turn on our IPv6 transit and we discovered in testing that the ICX7750 does not support running MCT and IPv6 on the same virtual interface (vlan). (This is not documented in the Layer 3 Routing Configuration manual and the engineer assigned to my ticket didn't even know, but it turns out it is mentioned in the Layer 2 Switching Configuration manual.) We will eventually get new routers, but in the mean time I need to develop a workaround (so we can finish getting IPv6 set up before our ISP loses interest).
Our current configuration:
We have 2 10g fiber uplinks to our ISP, running through transparent firewalls and terminating on our 7750s. The 7750s have a BGP session with the upstream routers for our IPv4 transit. Our BGP links on configured with an active/passive configuration - preferring the link to
dc-cr00. The 7750s use
VRRPe to direct all traffic to
dc-cr00 and failover to
A lot of the reasons for the design of our network come from legacy design decisions. It is also a production network handling thousands of users, so I can't break things or have extended downtime for reconfiguration.
My goal is to make the fewest changes to our existing infrastructure and also be able to easily upgrade to new routers as soon as we get them. To that end, I would like to put the BGP for IPv6 on the 7750s and avoid any MCT links.
With this solution, I would need to add a (non-MCT) link between the two 7750s. All the trunking would basically simulate the full-mesh OSPF links that we have with our other switches and the new VyOS VMs would serve as the IPv6 routers on each VLAN.
What are the downsides to this approach (assuming I have enough link bandwidth)? Any hidden pitfalls? Better ideas (not counting get new routers with reasonable support for IPv6)?