Hello.. I have a PVE 6.1 cluster with ceph. Presently the OS / Boot disk is deployed via LVM / XFS. We want to move towards a ZFS RAID-1 ROOT disk configuration.
Is there a procedure to migrate (could be offline) a host from XFS / LVM to ZFS RAID 1 without redeploying PVE?
I imagine redeployment would generally mean removing+rejoining node to cluster, as well as rebuilding OSDs.
If the nodes (one-by one) are reinstalled using ZFS RAID 1 ROOT, rejoined to cluster, and ceph monitor+manager+metadata redeployed, is there a method to "reattach" OSDs to the cluster -- e.g. Ceph-volume activate?
Would like to avoid cluster from rebuilding OSDs whose data is intact.
I realize the above is a bit of surgery. #1 is a Proxmox specific question, #2 is more ceph centric. Curious what others may have done.
A procedure as in written form that you can just follow? I don't think so, but it should be doable without any problems.
I'd approach like this:
- create a vm or use other hardware to test the migration
- bootup the new machine with a LiveSystem that is ZFS capable and create the pool (e.g. like this)
- copy files over the files from the production server
- try-and-fix loop until all potential problems are gone
- you should have a procedure and go on to the real nodes
If you want to improve your procedure, try on a real node (move all VMs away and exclude the node in HA) over and over again after creating a 1:1 backup of your-disk, so that you can play with the node, restore and still have a working node.
