Why is retention configued per plan, and not per repo?

[zolakt]

Hi, first time using Backrest (or Restic), so I'm a still a little confused on the best approach to set everything up.

I need to backup 2 sets of data, from the same machine: photos and document.
Initially, I've created 2 repos, and 2 plans (each for it's own repo), and that works fine.

However, multiple people have suggested that I don't need 2 repos, and that I could just have both plans push snapshots to the same repo. It would be more efficient, since it would prune/verify once, and easier to replicate to the cloud with "restic copy".

So now retention policies come into play. What if I want to have different retention for those 2 sets of data? Is a single repo off the table in that case?

From what understood, actual retention (as in --keep-daily) is done during prune, not during shanpshot creation. If that is correct, I'm really struggling to understand why did you put that option on the plan, and not on the repo? Is there some reason or use case that I'm missing? Otherwise, it just seems misplaced.

Also, what will it actually do in that case, lets say:

• plan A keep last 7 days
• plan B keep last 30 days

Different data, with no overlaps, both pushing to the same repo.

[garethgeorge]

Hi -- this is a good question and I can see the confusion coming from the restic model of the world where forget operations would typically apply to every snapshot in the repo.

Backrest is using tags on your snapshots to allow it to identify which plan created each snapshot, this means that when forget is run to enforce retention, it is running only on snapshots created by the current instance and tagged by the plan. You can look at snapshot metadata to see the tag:<plan ID> tag is added to each snapshot.

This means that there isn't any problem with mixing different retentions across plans, while storing their data in the same repository.

Finally you ask about prune-- restic deduplicates chunks by their hash, the way prune works is essentially a GC algorithm. A block is removed only if it is not reachable from any snapshot (e.g. because forget removed all snapshots referencing that chunk).

[zolakt]

Thanks for the reply. I think I get it.

To summarize, I could have both plans pushing to the same repo. That way I could spare one verification cycle and do it only once. And I could replicate the whole repo to the cloud with a single "restic copy". However, pruning I still have to do twice, since each is prunning just it's own tag. Right?