Ondrej Blazek

@NeilW has a point guys.. I have been investigating this for a while and the problem is with the shared codebase of icmpv6. All bpf programs (including to/from-netdev - attached...

> Turns out not to be the case. It's happening in all cases of NA replies, not just over the tunnel in tunnel mode. Yeah this I can confirm (also...

> We are still seeing the same behavior in cilium v1.12.1. yeah you should be seeing that as IIRC nothing has changed in that codepath

similarly to this feature request this is possible in OCCM https://github.com/kubernetes/cloud-provider-openstack/blob/fdba36babb2c4b46e759c99cca50ac7eba2ee06f/pkg/openstack/openstack.go#L103 via configuration: ``` [LoadBalancer] lb-provider=labrador subnet-id=5b4e2c9b-1ab9-4510-87a4-6d2301e16e77 internal-lb=true provider-requires-serial-api-calls=true ```

Basically the VMs only need 1 interface as they are routable across the whole DC. The VIP network is only a subnet within the routable network. > To be clear,...

the VIP network (in our case) is not meant to be used as VM interface, but is announced by our L4 balancer (standalone Cilium L4LB) and is BGP routable just...

@mdbooth any chance you would accept support for this? If we agree on an implementation I can start right away.

great thanks, will start with the implementation then

btw regarding: > Additional VIP subnets must all belong to the same network as the primary VIP. This is not yet possible as `additional_vips` are not yet part of gophercloud...

> I'd still insist that API should have the subnet field as an array even if documented that secondary values are ignored yeah, totally agree