This ticket is tracking the QE verification effort for the solution to the problem described below.

What's the feature?

OVN(-controller) currently advertises OVN routes (dynamic-routing-redistribute=connected/connected-as-host/nat/lb/static) as "blackhole" routes in the VRF table associated to the logical router that's configured with dynamic-routing=true.

There are use cases when the same VRF would be shared with the host for actual traffic forwarding.  In that case, the goal is to automatically use OVN advertised routes on the host too.

In order for that to work they need to be installed with an explicit next-hop IP.  To avoid having to guess that value of that IP, the CMS (e.g., ovn-kubernetes/neutron) will have to explicitly provide it to OVN via a NB router configuration.

The new config options should be along the lines of:

LR.options:dynamic-routing-advertise-v4-prefix-next-hop=<IP>
LR.options:dynamic-routing-advertise-v6-prefix-next-hop=<IP>

The reason why we need two configurations (one for IPv4 prefixes and one for IPv6 prefixes) is that OVN (and the kernel) supports mixing address families in routes, i.e., the prefix can be a v4 prefix and the next-hop an IPv6 address (and the other way around).

Why is it needed?

To enable the CMS (ovn-kubernetes) to share the VRF OVN monitors and updates when dynamic routing is configured.

Who will benefit? 

ovn-kubernetes transitioning to use OVN's native BGP support.  Potentially neutron-ovn too.