Not sure where to post this so I figured I’d ask here. I have a low priority ticket with Canonical, but I figured I’d see if the community has any suggestions or experiences that I can learn from.
The short of it is that I’ve got a small cluster of hosts that I’m trying to use to replicate an Openstack production deployment. Production has 17 compute nodes and 3 infra nodes, while testbed only has four compute nodes and three infra nodes.
In MAAS, I’ve defined my OAM, internal, and public subnets and they all appear when I run juju spaces and juju subnets. The problem is that the units I deployed through the initial walkthrough was designed for a flat network whereas now I’m trying to model production with separate networks for OAM/internal/public.
When I try to change the bindings to match prod, I get an error as the existing units/LXC containers don’t have network interfaces into anything but OAM so the binding never applies.
Would removing the units (but not the application), changing the binding, and using juju to re-add units create the networking interfaces or is there a way to trigger Juju to create the interfaces in LXC so I can modify the necessary bindings?
Thank you in advance!