zrepl VS lxd-snapper

Lastly there is zrepl. This is an automatic snapshot creation, pruning and replication daemon. It lets you automate the creation of ZFS snapshots at specific intervals, apply a retention policy to them, and replicate them out to a remote system with ZFS, like a NAS with TrueNAS on it.

I run it in a FreeBSD jail and take frequent ZFS snapshots using zrepl. I’ve had to restore after failed updates and it worked flawlessly.

As others have noted, I use syncoid but zrepl is an alternative that could be considered.

I’m going to throw out Zrepl again because it’s amazing: https://zrepl.github.io/

Current idea: We offer some form of container each other, where we have that zvol mounted somewhere for the replication use. This would mean that I could create an "inside zfs" where I create a filebased zpool. How is performance on a filebased zpool? (Although it is technically not so important that it performs critically, just trying to find the - or one of - "best" ways). Then we would offer each an ssh endpoint into that container (It doesn't have to be ssh, but it is so far convenient to setup and resilient to be open to the public). And my current plan is to use zrepl (https://zrepl.github.io/) to organise replication from my home server to this "inside" zpool.

Why would you want to use snapper instead of LXD's built-in snapshots or lxd-snapper?

lxd-snapper is a tool that takes snapshots.schedule to the next level - it's a fancy wrapper for lxc snapshot & lxc delete that supports rich, Borg-style retention policies and hooks: