Hi, I'm not familiar with kolla-ansible but since nobody else replied to this thread I just wanted to bump it. A year ago Gorka commented on my thread [3] about active/active setup, but we don't use any of the deployment tools, we have our own deployment method. I also don't really have answers for you, I don't think it makes a big difference which DLM tool you use, maybe the one you're more familiar with. I tested cinder-volume a/a with zookeeper and seemed to work well in my lab setup, I don't have an actual deployment to speak from experience. Regards, Eugen [3] https://lists.openstack.org/pipermail/openstack-discuss/2022-September/03055... Zitat von 韩光宇 <hanguangyu2@gmail.com>:
Hello,
I saw that kolla-ansible had allow cinder coordination backend to be configured. And it can be set to "redis" or "etcd" in /etc/kolla/gloabls.yml [1]. Based on my current understanding, this option enables cinder-volume active/active.
1. I would like to ask, between Redis and Etcd, which one is more recommended or what are their pros and cons? 2. And I have saw the spec of cinder-volume a/a[2]. Is it sufficient to configure [coordination]backend_url and [DEFAULT]cluster to use Cinder volume active/active, or are there any other considerations? Could any reference for Cinder volume active/active configuration and usage documentation be recommended ?
I would appreciate for any help. Best wishes.
Han Guangyu
[1] https://github.com/openstack/kolla-ansible/blob/master/etc/kolla/globals.yml... [2] https://specs.openstack.org/openstack/cinder-specs/specs/ocata/ha-aa-job-dis...