Proxmox Datacenter Manager - First Alpha Release

[bjsko]

that's not yet implemented - it requires adding the "remote" part to our replication feature.

Hi @fabian and @t.lamprecht, sorry if mentioning you here is not appropriate.

I'm just trying to figure out if we are talking about the same thing here as it is not clear to me when I interpret the first message as it "should be possible", and the last one as "not implemented yet"

In any case, looking at the bugzilla (https://bugzilla.proxmox.com/show_bug.cgi?id=4928) request I see a comment there that the functionality is on the table, but that it is not a trivial task to do. Has there been any further movement on this topic, or is it far down the list of stuff to look into?

Thanks for all your hard work.

BR
Bjørn

 

[fabian]

I'm just trying to figure out if we are talking about the same thing here as it is not clear to me when I interpret the first message as it "should be possible", and the last one as "not implemented yet"

both comments are referring to different things.

initially, export/import support for the RBD storage plugin was missing, so no remote migration was possible at all with disks on such a storage. that part got implemented, so it is now possible to migrate a VM from cluster A with disks on Ceph to a different cluster B with disks on Ceph (or from/to some other storage) with a storage migration, i.e., a full live copy of the disk.

what is not yet implemented is setting up a replication between cluster A and B (and their corresponding ceph clusters) so that you can re-use the already replicated volume as a base for the live migration (this is supported within a cluster with the ZFS-based replication feature).

what is also not implemented is a way to tell both PVE clusters "this Ceph disk/storage over here, and this Ceph disk/storage over there, are actually the same" - because that is fundamentally opposed to the notion that each cluster fully owns their (view of the) storage, and we'd need to implement a lot of safeguards first to actually track and hand over ownership of individual volumes from one cluster to the other.