Charm deploys LXDS to incorrect subnet

[Boris Lukashev]

When adding ceph-fs to an existing cluster with `juju deploy ceph-fs -n 3 --to lxd:0,lxd:1,lxd:2 --config ceph-public-network=10.10.122.0/24`, the containers never resolve as they are brought up on the ceph-cluster-network without the vNIC for the ceph-public network which both the mon and radosgw containers on the same host are given... Locks up charm provisioning since the containers exist but not on the right network. Bit of a show-stopper. Quincy/stable on 22.04 (both host and LXD)

[Alex Kavanagh (ajkavanagh)]

So the ceph-public-network config option sets "public network = <...>" and "public addr =" in ceph.conf:

{% if options.networks %}
public network = {{ options.networks|join(',') }}
{%- endif %}
{%- if options.public_addr %}
public addr = {{ options.public_addr }}
{%- endif %}

It doesn't do anything more from a networking perspective, other that telling configuring ceph-fs.

So how have you set up the networking for the containers (from juju) in terms of networking and the underlying infra? Is there a bundle you can share? Is it on MAAS?

Thanks.