pve 8 and pre Quincy hyperconverged ceph versions possible

[rainer042]

At the moment I am running two pve clusters both with pve7.4. One of the two is using storage from a external ceph cluster running ceph Nautilus (14.2.22). This is working for me without any problems.
Now in the online docs "Upgrade from 7 to 8" there under Prerequisites i read that for hyperconverged clusters one should upgrade an existing ceph Pacific or Octopus version to Quincy (v17.2) with am emphasis on the hint to this before upgrading pve to 8.0 .

Is there a special technical reason for this? What would happen if I stay with Octopus or Pacific and start the upgrade to 8.0 anyway since I want to stay with the existing version say Pacific?

Any ideas?

Thanks
Rainer

 

[sterzy]

Is there a special technical reason for this?

The reason we mention that for hyperconverged clusters is, that PVE 8 does not support any other Ceph versions currently. Hence, updating to Quincy after upgrading results in an untested and unsupported setup. So we can't give any guarantees about how well that will work. Additionally, our upgrade guides assume you are running PVE 7.x, so they might not apply cleanly to PVE 8.

Also, Ceph Quincy is the only still officially supported version [1], so we would strongly recommend upgrading to that regardless.

What would happen if I stay with Octopus or Pacific and start the upgrade to 8.0 anyway since I want to stay with the existing version say Pacific?

Since you are using an external cluster, this may work fine. I am currently unaware of any major incompatibilities between the Ceph client we use in that scenario and Pacific or Octopus. Regardless, you should think about upgrading ASAP, as Ceph Nautilus, Octopus, and Pacific have all reached their end-of-life stage and will no longer receive updates.

[1]: https://docs.ceph.com/en/latest/releases/#ceph-releases-index