Which do you prefer and why for HDD or SDD hard disks ? And how is Ext4 with power failures nowadays ? iirc a few years ago there were issues with Ext3 (Yes, Ext3) and I remember a no barrier option was needed ? Is Ext4 much better ? And what FS do you prefer on usb storage ?
ok. making snapshots with btrfs is also no backup-strategy (alone). But sending the snapshot to an external drive or backup-store is a really good thing for making backups. Much faster than doing this with rsync.
Don’t use rsync. I do them with Nextcloud or Syncthing and sometimes manually.
For other cases, I use a manager for an Hypervisor for VMs to make programmed backups, also including snapshots.
For me, an snapshot is a backup method.
BTRFS-snapshots only is no backup-method. Only if you send them to an external Store (external HD, SAN, NAS…) In case of deduplication and Copy-On-Write, a snapshot is physically the same store-location as the original. Only the few changed parts of the whole filesystem are stored on another physical place on the same HD.
So if you use BTRFS-snapshots you MUST send/receive the snapshots to another medium, if you want to use snapshots as backup. Data-loss is so a big danger.
I think you understand backup not as the real abstract concept but as a sub-class called “a secure way to make a backup”.
If you copy a file to other folder to the same hard drive it is still a backup.
The same happens with the btrfs-snapshots if the idea are for the files or changes you made in an Operating System installation (OpenIndiana with ZFS snapshots made for every system upgrade).
In BTRFS if you copy a File to another folder, only the reference is created. The physically information, what IS the file, is still the same. If you use [code]cp --reflink=always[/code] for sure.
So if you make a Snapshot with BTRFS, it’s almost like a hardlink. The file disappears, when the last reference gets deleted. But al reference-files break, if the physically location on the HD breaks.
So… making a snapshot in btrfs is not a backup for me. It is a good way for a fast rollback. It is a good way for deduplication. It is a good way for daily work. But a real backup for me is a physically separated store for the information.
Make snapshots on a computer for fast rollback, or after system upgrade - even here for fast rollback. Or if I’m working on my photos, i do automatically every 10 minutes a snapshot. So a wrong command does only destroy the work of maximum 10 minutes… never the whole work.
But backup… i make differential snapshots and send them with btrfs send/receive to another drive, so the snapshot is pyhsically duplicated on a separate store. One HD can break… i have a backup.
Do you understand, what i mean?
Maybe the example I gave was not enough.
The point of snapshots is serving as some kind of incremental backups always dependent of a main backup or the main instance.
Always the changes are copied, it is valid, even if a delta of a file was the only thing saved.
This applies to systems with deduplication enabled and COW.