Ask Your Question

Revision history [back]

Three options in my opinion:

Go Kilo-Liberty-Mitaka-Newton-Ocata-Pike-Rocky. Some distros allow jumping over a release, Helion OpenStack for example, which can go from Newton to Pike (with downtime). SUSE OpenStack cloud as well, I think.

Build an equivalent Rocky cloud in parallel to your Kilo cloud, copy all resources (instances, images, networks, volumes, ...) to it, then switch clouds. Requires more servers and storage, but no or little downtime.

Stop the cloud, save all resources, build the Rocky cloud, restore the resources. Downtime, but no need for additional servers.

Or option 2b, run the Kilo cloud on half of the servers, install Rocky on the other half, copy resources, switch, add the former Kilo servers to the Rocky cloud. No downtime, or partial downtime if you freeze the less important instances.

Three options in my opinion:

Go Kilo-Liberty-Mitaka-Newton-Ocata-Pike-Rocky. Some distros allow jumping over a release, Helion OpenStack for example, which can go from Newton to Pike (with downtime). SUSE OpenStack cloud as well, I think. The main problem is that each release has a different database schema, and schema conversion tools only exist for going from release N to N+1.

Build an equivalent Rocky cloud in parallel to your Kilo cloud, copy all resources (instances, images, networks, volumes, ...) to it, then switch clouds. Requires more servers and storage, but no or little downtime.

Stop the cloud, save all resources, build the Rocky cloud, restore the resources. Downtime, but no need for additional servers.

Or option 2b, run the Kilo cloud on half of the servers, install Rocky on the other half, copy resources, switch, add the former Kilo servers to the Rocky cloud. No downtime, or partial downtime if you freeze the less important instances.