Hi Dmitri, I was thinking of the reverse engineering route but it’s a production cluster so I don’t want to miss something and change configuration so I think I’ll go down the manual route. Thanks for the advice, very much appreciated as always. Will let you know how I get on. Regards, Derek
On 4 Dec 2024, at 20:32, Dmitriy Rabotyagov <noonedeadpunk@gmail.com> wrote:
Hey,
Well. You basically have 2 choices here. First would be to reverse-engineer config and re-create the state of the deploy host before it was wiped. Tricky part would be to get proper inventory back. As with the new config, container names and their IPs will be different from actual values. So you will also need to either manually patch the inventory afterwards, or destroy old containers manually and create new ones. But that is really tricky with 1 control plane and no HA, especially when it comes to rabbit and mariadb.
Second would indeed be to manage the cluster manually from now on. Adding a compute is in fact quite a trivial process. Given, that you're running Ubuntu, you pretty much can: * Add UCA repository [1] for the release you're on * Install required packages for Nova [2] and Neutron [3] depending on plugins and functionality you need * Sync /etc/nova and /etc/neutron directory from other compute * Issue/sync SSH keypairs for Nova (they're in /var/lib/nova/.ssh) * Sync /etc/hosts for name resoluiton * smth else?
[1] https://wiki.ubuntu.com/OpenStack/CloudArchive [2] https://opendev.org/openstack/openstack-ansible-os_nova/src/commit/61be9e722... [3] https://opendev.org/openstack/openstack-ansible-os_neutron/src/commit/65419f...
ср, 4 дек. 2024 г. в 15:42, Derek O keeffe <derekokeeffe85@yahoo.ie>:
Hi all,
I have a controller and 2 compute setup that was deployed some time ago with OSA. It's been running fine since it was deployed but we now have the need for more compute and I was looking at the deploy host to add a compute host to the cluster. Unfortunately it seems the deploy host was repurposed and I have lost all the config.
So two questions, 1) Is it possible to manually add a compute to the cluster (I fairly confident this is a yes) and 2) if so is there any documentation or how to that someone could point me to/share? any help would be greatly appreciated.
Regards, Derek