FWIW, I'm seeing this on 1.25.x. I didn't try with with 2.0.
Another difference is that we first do 'juju add-machine' for each physical machine and LXC that we deploy to. So we never let 'juju deploy' or 'juju add-unit' create the machine, we always use --to. Not sure whether that has an impact.
FWIW, I'm seeing this on 1.25.x. I didn't try with with 2.0.
Another difference is that we first do 'juju add-machine' for each physical machine and LXC that we deploy to. So we never let 'juju deploy' or 'juju add-unit' create the machine, we always use --to. Not sure whether that has an impact.