Hi,

We went the „long“ way.

- first emptied osd node by node (for each pool), purged all OSDs
- moved the OS from centos 7 to ubuntu 20 (reinstalled every node)
- removed the cache pool and cleaned up some config
- installed all OSDs and moved the data back
- upgraded ceph nautilus to octopus (containered) and now we are moving to pacific/quincy.

This worked (slow - lot’s of data) without any bigger outage or problems. The only thing was, that a pool of NVMe nodes was flooding other nodes while backfilling :) … 
… so start with low backfill values and CHECK before you carrie on with other pools or nodes.

Best . Götz




Am 17.01.2024 um 08:09 schrieb Szabo, Istvan (Agoda) <Istvan.Szabo@agoda.com>:

Hi Goetz,

Which method you finally choose?
We've done a successful migration from Centos 8 to ubuntu 20.04 but we have a centos 7 nautilus cluster which we'd like to move to Ubuntu 20.04 octopus same as you.
 Wonder any of you tried to skip Rocky 8 from the flow?

<…>