Cloud backup from NFS server

I’m working on the backup & recovery process using NFS and rdiff-backup.

A while ago, an article about Duplicity was published here.

Comments from a user about backup policy that are hard to grasp are;

  1. Backups must be done in two stages, sync and increment.
  2. Your laptop should not have ssh keys for admin access to the backup server.
  3. Many online backup services have this separation built-in.

IMHO,

  1. rdiff-backup folders show mirror (of the source data) and increments at the same time. How is it possible to separate these steps in two different devices using rdiff-backup or Deja Dup/Duplicity?
  2. My main laptop has ssh access to my server for headless setup. So if I run rdiff-backup from NFS server to backup server (cloud), is it okay?
  3. Which software has built-in separate of two processes?

Which devices (workstation 1 or server) in the diagram below should be source directory?
What would you recommend as best practice for data backup?

:thinking: I’ll just give my thoughts without any experiences.

For performance on each workstations, that diagram looks good.

But if performance are not a priority, then backup source from each workstations should be safer and use lokal server act like a local backup/sharing for fast access.

1 Like

Sure, the local server should act as a backup, and sharing, but it does not require fast transfer. The just in case scenario is when a laptop becomes unusable (temporarily due to malware or long-term).

One thing I overlooked is duplicate hardware (server) for performing a complete restore with the recovery image. My NFS server runs on R PI, so it is inexpensive to have a spare. I’ll carry out the recovery process.

[Updates to my initial questions]
Separate initial backup (or snapshot) and increments are practiced/tested in rdiff-backup, Deja Dup/Duplicity.

In addition, ReaR can create a disaster recovery image of a live server - system backup. As soon as I can get a duplicate Pi (all sold out here in UK), I’ll test recovery process from system backups which is much faster than rebuilding systems.

I better understood the intention of the original article by practicing it.