On Thu, 05 Feb 2015 10:05:22 +0100
Miroslav Suchý <msuchy(a)redhat.com> wrote:
On 02/05/2015 01:13 AM, Kevin Fenzi wrote:
> Could we instead call it 'openstack.cloud.fedoraproject.org' or
> 'controller.cloud.fedoraproject.org' or something? Not sure if that
> needs us to rename/reinstall the node, or can just be done in the
> cert...
It can be just cname + name in cert. Reinstall is quite fast with
ansible, that is no problem. I automated all but one workaround
(there is still usually need one reboot).
Sure, true.
> Along those same lines, how about we move the existing host
> playbooks to a group/openstack-controller.yml (currently just
> fed-cloud09, but I'd like to see if we can allocate one machine
> moving forward to be our test for the 'next' openstack) and
> group/openstack-compute.yml (fed-cloud10/11, but some more will be
> installed next week) to make them more generic and ready for more
> nodes?
Compute node is already in roles/cloud_compute/tasks/main.yml so
migration to groups should be easy (not my priority thou).
Sure. Just makes more sense to me.
I see no
benefits in migrating controller playbook to group or roles. It is
only one. I +1 to controller-next instance, because upgrade of
OpenStack is not supported. However those playbook will be quite
different and it does not have sense to have them in one playbook
with "when" directives.
Good point. So how about:
hosts/fed-cloud09.cloud.fedoraproject.org.yml -> hosts->
openstack-icehouse-controller.yml
hosts/fed-cloud* -> groups/openstack-icehouse-compute.yml
Of course this is all just somewhat cosmetic. I just wanted to do it
before we added more compute nodes.
kevin