can't co-locate nova-lxd with controller services deployed in containers

[Jason Hobbs]

21:10 < jhobbs> rockstar: do we need to avoid putting nova-lxd compute nodes on on the same system as juju managed lxd containers?
21:11 < jhobbs> zul: ^^
21:11 < rockstar> jhobbs: there is a bug, currently, where if you put the control plane in lxd containers, and then try to but compute nodes on the same system, the configuring of nova-lxd breaks some things.
21:12 < jhobbs> ok
21:12 < rockstar> This is a well-known bug, and one that we probably need to find some time to address.

[Ryan Beisner (1chb1n)]

I believe this is a conflict between Juju 2.x and nova-lxd, both needing to configure lxd's storage on the host.

[Anastasia (anastasia-macmood)]

Please clarify:
1. What version of Juju are using?
2. What version of LXD?
3. What "well known bug" are you referring to?
4. What steps did you follow to reproduce?
5. Do you have logs for your deployment?

Thank you for filing the bug and the conversation snippet :D
Details will help to triage/investigate and address the issue.

[Anastasia (anastasia-macmood)]

@Ryan, @Jason,

From Juju's perspective this is a duplicate of bug # 1573681 which also targets Charms Collection.

I'll mark it as such. Feel free to add to the original bug.