Juju Bootstrap to OpenStack Rocky with LXD hypervisor

Hey Guys!

I have a running instance of OpenStack Rocky using LXD hypervisor. Everything appears to be perfectly OK and functional. All basic checks are passing. Uploading images, creating flavours, the network is nicely set up, instances in different series (trusty, xenial, bionic) are running. OpenStack Dashboard is functioning properly. There is not a single detail that would suggest that something is wrong.

However, Juju bootstrap is failing. From what I can get it appears that initial cloud-init setup is failing.

bootstrap debug output: https://pastebin.com/prN4c9BX
cloud-init logs: https://pastebin.com/Nae4YcYC and https://pastebin.com/Jkv6F8pY

Any ideas how to debug and eventually solve this problem would be highly appreciated. :slight_smile:


Ok, so that’s running Juju against an openstack cloud which is using nova-lxd rather than kvm?

If so, I suspect the Juju guys may have a better idea as to what’s going on here.

@ajkavanagh may also have ideas :slight_smile:

Yes, that’s correct. OpenStack cloud is running well. It’s Bionic deployment. I did not mention that.

I did not get into the details yet, but Juju bootstrap is using cloud-init to preconfigure the node before SSH-ing into it. Adding SSH keys, agent binaries, etc. The container itself is healthy like any other that I would create and start using OpenStack CLI or Dashboard.

Might be interesting to look at the cloud-init logs inside the container, maybe it’s not properly fetching its metadata from openstack.

I provided those two logs. I can’t find anything about adding Juju agent.

Did you have something like this in mind?

2018-12-12 11:49:35,059 - url_helper.py[DEBUG]: Calling '' failed [3/-1s]: bad status code [500]

I was looking for Juju metadata. :frowning:

Looks like it did hit a metadata server but then got a 500 error back, so may need to go look at the relevant openstack logs to see what’s up with that.