Lets go to the original design and the original intent of how this is supposed to work.
It does *not* matter whether this IP is a VIP or a real IP assigned to a machine that is on the same subnet, or on a routed subnet. The premise is that the IP needs to be routable for the machines.
As such, the design is that machines need to be told two things:
1. The IP of maas_url from the rack they boot from, because that's the only place where we know where the region is (regardless of whether this is a VIP or not).
2. The *other* region IP address that face the machines.
What's missing here is maas_url which should *always* be present regardless of it being a single region/rack, a multi-region in the same subnet, a multi-region with VIP's or a split region/rack with a routed l3.
Mike,
Lets go to the original design and the original intent of how this is supposed to work.
It does *not* matter whether this IP is a VIP or a real IP assigned to a machine that is on the same subnet, or on a routed subnet. The premise is that the IP needs to be routable for the machines.
As such, the design is that machines need to be told two things:
1. The IP of maas_url from the rack they boot from, because that's the only place where we know where the region is (regardless of whether this is a VIP or not).
2. The *other* region IP address that face the machines.
What's missing here is maas_url which should *always* be present regardless of it being a single region/rack, a multi-region in the same subnet, a multi-region with VIP's or a split region/rack with a routed l3.