Finally catching up with this thread (have been in cobbler-land rather than ansible-land for the past few days, doing provisioning stuff…). So, I’m about to do a Folsom deploy of OpenStack internally. I’m looking through this thread, and it wasn’t clear to me whether a concrete plan had formed about what the starting point will be. Are we going to use Jimmy’s jcftang/ansible-openstack repo as the basis for the canonical ansible playbooks? I’m happy to offer mine up as well (lorin/openstack-ansible). I’ve seen lots of references to it, but haven’t gotten much direct feedback on it, so I don’t know how suitable it is outside of the original goal of illustrating an install on Vagrant. I’d like to try to update it for Folsom, although it may become obsolete once the official repo ramps up.
I would really like to see ansible modules like those David has mentioned. It would be great if these eventually used the Python APIs, although shelling out to the command line is a fine start (yes, there are Python APIs for all of the OpenStack command-line tools. They are great, although poorly documented. See http://docs.openstack.org/developer/language-bindings.html)
Side note: for OpenStack best practice stuff, take a look at the new OpenStack Operations Guide: http://docs.openstack.org/ops/
BTW, Curtis, your Cybera blog post was how I discovered Ansible in the first place.
Lorin -- I'm very much leaning on us taking lead from Packstack (and
porting it) at the moment. This way we have a pretty well
evolved/entrenched example.
But I want it to be something everyone can use as a baseline.
I was thinking about making my ansible-openstack repo private, I don't
think it's going to be suitable as a set of examples as I'm beginning
to customise it for my needs (for a year down the road). I think I was
a little too eager to just dive right in to get things going and
sharing it early.
Lorin -- I'm very much leaning on us taking lead from Packstack (and
porting it) at the moment. This way we have a pretty well
evolved/entrenched example.
But I want it to be something everyone can use as a baseline.
I believe packstack is limited to RedHat/CentOS. Would we just be
supporting that OS to start with openstack-ansible? Not that it's a big
deal to me, though most openstack clouds I interact with are based on
Ubuntu 12.04.
We’re an Ubuntu shop, so I can contribute by Ubuntu-izing the CentOS stuff. I think the Fedora folks have their own openstack config tool for doing some of the initial painful setup that isn’t upstream, and they default to Qpid instead of RabbitMQ, but otherwise I think it will be pretty similar.
Oh, I didn't mean something private. I was referring to the
"openstack-config" tool, which I think(?) is bundled with the Fedora
OpenStack packages, but isn't in the Ubuntu ones. Packstack seems to use it
here:
I would likely use a dynamic host group (group_by) to segment by OS
and then have different plays use a mix of common and OS specific
include files to keep it dead simple -- though even just a few
conditionals sprinkled here and there would not be too bad.
I have have updated the script for Grizzly and added Cinder block storage. The new dashboard is really good with Quantum and should help newcomers get their heads around it.