Safely move snapshots from one namespace to another?

[MaeveFirstborn]

Hey guys, I have three clusters geographically separated, connected via OSPF and Wireguard. They operate mostly independent of each other, aside from Active Directory, but we use each one as an offsite for the other two. Unfortunately, Cluster A has 4 TB capacity than Clusters B and C. As a way to prevent it from filling up too fast, what we're doing is using a separate namespace for a particularly large VM (864GB) on Cluster B's backup server, then syncing that over to a namespace that has a specific prune job that only keeps a very small number of snapshots on Cluster A. That way Cluster B and C have the full historical context of that VM, but Cluster A has less, just enough for an emergency. Much less storage cost - 630GB less, and that's with the initial pruning and testing here.
My problem is such: when I moved this to a new namespace, I did so by moving the contents of ns/{original_namespace}/vm/{vmid} to ns/{new_namespace}/vm/{vmid} on both backup servers, which is working fine. The problem is that I accidentally left the backup job on Cluster B pointed to its original namespace, so there are 3 snapshots in there instead of in the new namespace.
With the way deduplication works, it's my understanding that I can simply repeat what I already did, moving the contents of the VM folder in the old namespace to the new one on both servers, fix the backup job to point to the right namespace, and then continue on. I just wanted to clarify that this is the correct procedure before I do this as I don't want to have to copy the data over from Cluster C if for some reason this understanding is wrong, as the WANs connecting these sites all have really low upload speeds.

Is my methodology here sound?

Thanks, Maeve

 

[tcabernoch]

I ran into a similar issue when I made some mistakes building a similar setup.
Compare notes here: https://forum.proxmox.com/threads/pbs-namespace-sync.153411/

So here's the deal. You can delete all the namespaces. That will delete all the backups inside them.
But it doesn't. The data is all still there, and it will be for another three days if your PBS is on default settings.
So nuke the b0rked namespaces and their contents. Go ahead.

Then setup your sync job so it works like you think it should.
Run it. Its gonna scan all the remote backups again, pretend to push the data, and then transfer 0 bytes.
And if you screw up again, just nuke it again, sync again. The data is still there. It will only take minutes.

BTW ... Bravo for the complex namespace sync stuff. It's good to see someone else doing this.