Recoverable filesystems?

[crispy]

I am going to upgrade my fileserver, and now Im wondering what kind of setup is should use. I don't have any preferences one way or the other so whatever solution fits my requirements is great. These are the situations I would like to avoid:

1) Broken disk -> data loss
2) rm -rf * in wrong folder -> data loss
3) mk2fs on wrong partition -> data loss
4) delete wrong file (or regret some days after) -> data loss

Some kind of raid and btrfs can probably save me from 1, 2 and 4.

[Kano]

Well, rm can be 100% destructive now as soon as TRIM is used on SSDs, ext4 needs discard mount option or some distros trigger fstrim from time to time. As soon TRIM is executed the controller can only show 0 for the deleted areas. Basically uncompressed/unencrypted filesystems can be analyzed with low level access via file signatures with tools like photorec (Debian has it inside the testdisk package). I see no way to restore files with compressed filesystems (like btrfs supports), but maybe there are some tricks i don't know. If you have data that you don't want to lose, store/mirror it on different systems or at least different hds, preferred not active the same time while not mirroring.

PS: Edit you post! It is 100% unsecure to write commands in there that can be copy/pasted by accident and delete data!!!

[nanonyme]

I am going to upgrade my fileserver, and now Im wondering what kind of setup is should use. I don't have any preferences one way or the other so whatever solution fits my requirements is great. These are the situations I would like to avoid...

Some kind of raid and btrfs can probably save me from 1, 2 and 4.

Yeah, 1 would be solved with raid, 2, 3 and 4 with backups. Btrfs supports pushing incremental changes through snapshotting so it's nice on both ends of snapshots.