I will see if i can set a snapshotable filesystem too, i know friends are doing this with #zfs, #btrfs should achieve smae purpose.

And yes it should work else it could not be called snapshot. What bother me is memory footprint of such solution i doubt it is free lunch ...

for #selfhosting, it will be enough,i'm confident that my service can endure a powercut, so snapshot will be fine.

i haven't figured out how i will manage off-site backup, maybe export a tar from borg ? or rsync the repo even if it's not advised.

@mistersheep even network traffic should be lower with #borgBackup than plain #btrfs send. If haven't worked with btrfs send yet, but I guess it's faster than borg if you leave network speed out of the equation.

Now I'm thinking about a new strategy:

- stop service
- make #btrfs snapshot (seconds at max)
- restart service
- run #borgBackup from snapshot, but via #bubbleWrap so it sees it as the original path and inodes for consistency&performance!
- run as many borg backups as desired to any remote, even in parallel, as the service is running again

Thoughs?

#nixos