PSS: I would suggest spawning up some all-in-one setup for deployment tools you pick.

* For OpenStack-Ansible you can find a doc here:

https://docs.openstack.org/openstack-ansible/latest/user/aio/quickstart.html

But eventually it can be even more trivial, just:

git clone https://opendev.org/openstack/openstack-ansible; cd openstack-ansible; ./scripts/gate-check-commit.sh aio_metal_horizon

You can also add more "scenarios"/projects separated with underscore if needed. "aio" includes core only - keystone, placement, nova, cinder, glance and neutron.

* For kolla-ansible https://docs.openstack.org/kolla-ansible/latest/user/quickstart.html

Also I'd try to get 16gb of RAM for aio for better experience, as 8gb is really bare minimum.

On Tue, Mar 12, 2024, 09:45 Dmitriy Rabotyagov <noonedeadpunk@gmail.com> wrote:

As PTL of one of deployment toolings I'm not in position to tell about
benefits/disadvantages of any, as I am obviously quite opinionated.

About OpenStack-Ansible - yes, it can deploy just on bare metal
without containers and does support both CentOS and Rocky Linux (while
I'd suggest Rocky Linux according to my reflection on our CI stability
for it).

About configuration examples, for bare metal it's enough to set in
openstack_user_config.yml:

global_overrides:
  ...
  no_containers: true

One thing that you completely skip during configuration when
no_containers is set - provider_networks definition, which is probably
most non-trivial thing in a setup, and use `neutron_provider_networks`
instead to define mappings for neutron [1]

You can also do a mix of containers/no containers depending on how
inventory will be generated. There's some doc here: [2]

Regarding usage statistics of deployment tools, I'd suggest referring
to the user survey [3] on the tab "deployments" (or deployment
decisions, though data is slightly different on these 2 tabs somehow).

[1] https://opendev.org/openstack/openstack-ansible-os_neutron/src/branch/master/defaults/main.yml#L417-L428
[2] https://docs.openstack.org/openstack-ansible/latest/reference/inventory/configure-inventory.html#deploying-directly-on-hosts
[3] https://www.openstack.org/analytics/

вт, 12 мар. 2024 г. в 09:22, Ruslanas Gžibovskis <ruslanas@lpic.lt>:
>
> Hi team,
>
> Could you, please, share your opinion and your experience with deployment methods. Also what are the tools you use?
>
> I have some deployments running TripleO tool.
> What I like is that we can deploy on empty servers and customizable OpenStack Installation.
> Sadly TripleO stopped working on new releases last year.
>
> Question:
> Which deployment method you would recommend which would support deploying on empty servers?
> My plan is to use CentOS Stream as base OS.
> My setup: 3x controller and 10+ computes. All with generic things except cinder, I do not have volumes and telemetry.
> But this year I need to add SR-IOV and some computes with RT kernels.
>
> Suggestions?
>
> My research:
> OSP-ansible - looks like it could deploy without containerization (I find it as a benefit).  (could someone share sample config you use?)
> Kolla-ansible - looks like old time here and most used? (could someone share sample config you use?)
> But all above are deploying on existing OS.
>
> Tho I have found the Kayobe, quite new to me, have you tried? How multi-controller and multicompute deployment config looks like?
>
> Thank you for your opinion or sharing your experience with the tools you use.
>
> --
> Ruslanas Gžibovskis
> +370 6030 7030