Sanity check for remote ZFS backup strategy

[proxale]

I'm drawing up plans for remote backups of a few ProxMox servers, non-cluster single machines in different locations.

After reading through the forum, AI tools and other sources, it seems like using ZFS send/receive to a remote server also running ZFS ( rsync.net or my own server) would be the optimal method especially for large datasets over Internet.

However, in case I'm badly mistaken, I would appreciate a sanity check on what I think I understood.

Planned procedure

1. Existing ProxMox zfs pools are unencrypted
2. An additional local HDD will be added with encrypted zpool, independent of the ProxMox VM storage
3. Daily zfs snapshot and sync to the encrypted zpool ( local backup )
4. Daily zfs send from encrypted pool to remote zpool

Expected recovery options possible

1. Server disk failure, restore from local disk quickly
2. Catastrophic local disks failure (i.e. server burnt), complete restore from remote zpool
3. Deletions recovery
   1. create local temporary recovery zpool
   2. sync/receive zpool snapshot
   3. mount received recovery zpool
   4. locate/copy deleted files

Is this a viable plan or are there better ways to do this that can provide both disaster recovery as well as file level recovery?

I understand there is PBS but as it is recommended to install on bare-metal, there isn't a budget for this. Similarly, the existing ProxMox servers don't really have much free resources ( specifically RAM ) to run another VM. A new HDD per server is about the limit of the budget I have to work with. Crucially, it seems that ZFS send/recv over Internet would be much faster and less costly using something like rsync.net than a complete VM to run PBS remotely.