Ceph 21.2 Umbrella Release Available as test preview and Ceph 19.2 Squid EOL

[t.lamprecht]

Hi Community!

The recently released Ceph 21.2 Umbrella is now available in the Proxmox Ceph no-subscription and test repositories for installation or upgrade.

Upgrades from Tentacle to Umbrella:
You can find the upgrade how-to here: https://pve.proxmox.com/wiki/Ceph_Tentacle_to_Umbrella

New Installation of Umbrella:
Since pve-manager 9.2.23, the Ceph installation wizard offers Umbrella as a technology preview. Select it together with the No-Subscription or Test repository.

Current State:
We have been running and testing the Umbrella release candidates and this release internally for a few weeks and found no major issues.
The current default Ceph 20.2 Tentacle will stay supported until June 2027 for the time being.

Road to Enterprise Stability:
We plan to lift the preview state and provide Umbrella as a fully supported Ceph release once we have gathered even more test time and feedback from QA. Naturally, we would be happy to hear about observations from our great community! Once we deem the release and its integration into Proxmox VE fully production-ready, we'll also populate the Ceph Umbrella enterprise repository.

Reminder: Ceph 19.2 Squid EOL
Please also remember that Ceph 19.2 Squid is close to the end of its life (EOL), with only one last update currently in progress. You should upgrade any existing Ceph Squid setups sooner rather than later, either to Ceph Tentacle or, once it is fully supported, to Ceph Umbrella. See the upgrade how-to: https://pve.proxmox.com/wiki/Ceph_Squid_to_Tentacle.

We welcome your feedback!

 

[J-Rod]

Ran this on my test cluster. Tiny nit-picks:
- ceph.sources or ceph.list section maybe could be a tiny bit clearer/cleaner for someone just marching from top to bottom. Some hosts could have one file or the other, some hosts may have both with/without some lines commented out.
- after the upgrade, I had two MDSs still on the old version so I ran a rolling restart in the GUI to fix. Looking back, I noticed of course unlike the ceph commands, systemctl has to be run on each host individually. Perhaps that could be made more clear.
- ceph osd pool autoscale-status returned nothing in my test cluster, which is a known result when allowing pool(s) to span multiple device classes. I ultimately decided to constrain all of the pools so the PG autoscaler could run.