The other issue you are going to have is with galera spread across 2 DCs you aren't getting the high availability you think you are. In order to meet quorum requirements you need an odd number of nodes otherwise your cluster will split brain on loss of a datacentre. The only way around that would be to have one datacentre have more mysql nodes than the other, but that effectively means if you lost that DC your cluster would go down anyways. ________________________________ From: Nguyễn Hữu Khôi <nguyenhuukhoinw@gmail.com> Sent: 13 November 2023 07:20 To: Dmitriy Rabotyagov <noonedeadpunk@gmail.com> Cc: openstack-discuss <openstack-discuss@lists.openstack.org> Subject: Re: [kolla-ansible] Galera DR across multiple datacenters CAUTION: This email originates from outside THG ________________________________ Hello. I think kolla ansible cannot deploy galera cluster per DC. If u want to do, so u need custom kolla ansible, Nguyen Huu Khoi On Sun, Nov 12, 2023 at 5:16 PM Dmitriy Rabotyagov <noonedeadpunk@gmail.com<mailto:noonedeadpunk@gmail.com>> wrote: Hey, Sorry, I can't help with kolla answer, as I have close to no idea, but was wondering what is wrong with ProxySQL and why not worth using it? Except cumbersome user management:) As I was also considering it lately for adoption. PS: I would wait for a more native solution for kolla, but fwiw ... OpenStack-Ansible has quite a decent Ansible role for MariaDB+Galera cluster deployment, that we use outside of OSA a lot as well, as it's quite agnostic: https://opendev.org/openstack/openstack-ansible-galera_server<https://opendev.org/openstack/openstack-ansible-galera_server> On Fri, Nov 10, 2023, 21:57 Jahns, Jay <jahns.jay@heb.com<mailto:jahns.jay@heb.com>> wrote: We currently run OpenStack on top of two datacenters, and we need to ensure we have datacenter redundancy for the database. With this, we would have 2-3 nodes in each DC running galera. Its been recommended by multiple people that we use an externally managed database server (or cluster) that has its own redundancies built around the two DC layout, and not have kolla manage that. Is there a way we can use kolla to deploy a galera cluster per DC, and use that to provide redundancy in the event that we lose one of the DCs? We were currently looking at proxysql but folks have recommended against it. Thanks.