Fuel sets force_config_drive = True which causes issues

[Ed Balduf]

Fuel 9 sets the force_config_drive=True option in /etc/nova/nova.conf which causes all kinds of issues like https://github.com/coreos/bugs/issues/774 and prevents live migration from working (unless you have the config drive on a shared volume).

[Dmitry Klenov (dklenov)]

@Ed Balduf, can you please also specify steps to reproduce?

[Roman Podoliaka (rpodolyaka)]

...at the same time it allows us to have reliable file and metadata injection:

https://review.openstack.org/#/c/322132/

So there is a trade-off here. force_config_drive just seems to be a better default for now.

[Ed Balduf (ebalduf)]

To reproduce, I'm going to upload 3 files, which include 2 heat templates and a user-data script. These produce a working CoreOS cluster (when force_config_drive=False). To reproduce, get CoreOS image and call it CoreOS. Then run go here https://discovery.etcd.io/new?size=3 and get a cluster start token, the drop it into this command line in place of the token shown.

openstack stack create --template multi_coreos_template.yaml --parameter "cluster_init_token=ae75bff9aa1aeffd8947316dda4d56d9" coreos-stack

To confirm issue, consult the file /run/systemd/system/etcd.service.d/20-cloudinit.conf on the CoreOS instances.. it should have the private IP address... it won't it will either be blank or still say $private_ipv4.