juju 2.8.5 bridge activation failed in bionic machines upgraded from xenial

[Giuseppe Petralia]

We have a cloud where all xenial machines were upgraded to bionic.
Netplan was installed, /etc/netplan folder created but empty and machines are still using /etc/network/interfaces

When trying to create a new lxd on one of those machines, it fails with

2021-09-07 14:55:23 WARNING juju.worker.provisioner provisioner_task.go:1173 failed to start machine 124/lxd/0 (failed to bridge devices: bridge activation error: device - name "bond0" MAC "98:03:9b:6f:47:0a" not found), retrying in 10s (1 more attempts)
2021-09-07 14:55:34 ERROR juju.worker.provisioner provisioner_task.go:998 cannot start instance for machine "124/lxd/0": failed to bridge devices: bridge activation error: device - name "bond0" MAC "98:03:9b:6f:47:0a" not found

If we create an lxd on a freshly deployed bionic machine that is using netplan this works as expected.

[John A Meinel (jameinel)]

So as I understand this, the issue is that if you have a system upgraded from Xenial to Bionic, it isn't actually using netplan for networking configuration, and we are failing to detect that it is using /e/n/i as the correct source of configuration.

[Peter Sabaini (peter-sabaini)]

For future travelers:

This seems to be caused by the netplan package creating an empty /etc/netplan dir. It appears that Juju provisioner probes for an /etc/netplan dir to decide if it should create bridges via netplan. In the case of upgraded machines this is not a good test as even though the netplan pkg is installed and /etc/netplan is present networking is configured in /e/n/i

Workaround: remove the empty /etc/netplan

h/t lathiat for analysis

[Jorge Merlino (jorge-merlino)]

2.9: https://github.com/juju/juju/pull/14510

[Canonical Juju QA Bot (juju-qa-bot)]

This Medium-priority bug has not been updated in 60 days, so we're marking it Low importance. If you believe this is incorrect, please update the importance.