I'm migrating file backups from Restic to Proxmox Backup Server. Infrastructure:
From reading the technical docs and this recent forum thread, I understand that client-side deduplication only kicks in when there is a previous snapshot in the same datastore + namespace + backup group. If the seeding snapshot is in place, the WAN client should be able to download the chunk list from it and only upload genuinely new or changed chunks.
Is this correct and are there any additional gotchas I should be aware of? For example:
- Backup source – remote host in a second datacenter (WAN link)
- Restic server + PBS – co-located in the same primary datacenter
- Install proxmox-backup-client on the Restic server.
- Mount the existing Restic backup volume locally.
- Run the PBS client on the Restic server to push a seeding backup to PBS, over LAN, targeting the exact same datastore + namespace + backup group that the real backup source will use later.
- The real backup source then runs its first backup over WAN against PBS.
From reading the technical docs and this recent forum thread, I understand that client-side deduplication only kicks in when there is a previous snapshot in the same datastore + namespace + backup group. If the seeding snapshot is in place, the WAN client should be able to download the chunk list from it and only upload genuinely new or changed chunks.
Is this correct and are there any additional gotchas I should be aware of? For example:
- Does the change detection mode (legacy vs metadata) used during seeding need to match what the real client will use?
- Does file path structure matter? My Restic mount may have a different path layout than the real backup source. In legacy mode I suspect a path difference would break chunk alignment across the entire pxar stream. Is metadata mode more resilient to this since payload and metadata are separate?
- Does backup group ownership need to be updated after seeding so the real client can read the previous snapshot's chunk list?