Send-and-forget work-flow?

[gcb]

Hi. I'm trying to migrate my scp workflow to borg.

Basically i'd scp or rsync everything form the client (sometimes N clients, but let's keep it simple), to a server. And then scp the files i needed back to the client.

with borg, i'm planning to:

Start a new repository on the server:
borg init --encryption repokey --info --progress --show-version --show-rc --umask 0077 --rsh ssh ssh://server/path
(mostly defaults)

then every week or so i run:
borg create --info --progress --stats ssh://server/path::client1_$(date +%F) --exclude '*cow2*' ~/Documents ~/Picutres
(date +%F is just to get the current "YYYY-MM-DD")

now, on the client, I DO NOT have the entire files from the backup. For example, every week I import the pictures from my phone, and send them to backup. And from time to time I fetch one month from the backup as I need.

for example, on snapshot ::client1_2026-02-01 i might have:

Pictures
  `- 2026
      |-01
      |  `- some 100 files
      `-02
        `- some 10 files

Then on ::client1_2026-03-01 i will have

Pictures
  `- 2026
      `-02
        `- some 100 files (10 same as before)

Q1. Will creating snapshots like this cause problem? will borg ever drop the old files not present on the client?

Q2. Also, files in ~/Documents are most of the time always present on the client. Should I create two borg repos or borg handle that mix (one path always adding files, another path client always missing files) well?

[ThomasWaldmann]

A1: A borg archive contains exactly what you put into it, it does not have a dependency on other archives.

As long as you do not delete or prune archives with stuff that you still need, there won't be an issue.

borg prune does not fit this use case, because you can't just keep some of these archives by a time-based pruning policy. So you would be very careful, never use prune and only use borg delete manually after making sure you still have all the data you need in the remaining archives.

A2: Usually, one only needs to split repos if they would get too big. Or if one wants to run multiple backups at the same time (from one or from multiple backup clients).