16

I just bought a disk to backup all my documents and to store it apart and I don't know what is the filesystem under Linux that gave me better reliability. Maybe I'll backup one time each month and it will not be connected all the time, but I want to be sure that my data will remain.

Maybe ext4?

1
  • No specific file system will make your data disappear while the disk is disconnected, so that's a non-requirement. (The media itself may still fail, but then the file system won't help you much.) This question might be answerable if you state your requirements, such as the backup disk size, maximum file size you need to backup, number of files and any specific file system features needed (hardlinks, ACLs, extended attributes, ...). Also as pointed out it may be very well worth your while to consider a file system other than that on the main disk.
    – user
    Aug 8 '13 at 13:13
7

The question is subjective.

EXT4 is considered stable and supports journaling (as does EXT3 but not EXT2) which helps with corruption.

BTRFS is still experimental but offers a lot more features.

Reiser4, TUX3, XFS, various ZFS implementations, etc. there is a lot of options.

Personally, I use EXT4 on all my drives and haven't had issues with the FS.

I'm not aware of any stability benchmarks (there may be some 'white papers' though) but Phoronix has performance benchmarks

Lastly, this question may be a duplicate of Rock-stable filesystem for large files (backups) for linux

3
  • 1
    to note modern (post 2012) ext4 does checksums on most but not all metadata , and sadly not the actual on file contents data (which BTRFS, or ZFS does). Also some parts the stable (as in does not change anymore and code base is well tested) on disk layout of ext4 is redundant as I displayed unix.stackexchange.com/a/499668/24394, but most data is not. If ext4 is used as backup, consider increasing its "recoverability" via e2image tool! Feb 10 '19 at 9:14
  • Since the answers suggestion ext4 does not neither provide checksum nor increased protective redundancy measures for the files it stores, it would be advisable to invest to compute - at least very important files in the backup fs - par2/parchive files, which would provide verificability and repariability (to some extent) and protect against unreadable sectors occuring, by adding a variable (by default 5%) additional Reed-Solomon redundancy data to those files. This would protect against "silent corruption" and "noticable" bit rot. Feb 10 '19 at 9:22
  • @humanityANDpeace: COuld you provide a link to further information about this?
    – einpoklum
    Mar 26 '21 at 16:11
7

I always take a different FS for the backup drive than on the main drive just because if I hit a strange FS bug I won't hit it on my backup medium as well.

9
  • 1
    Could you please explain what kind of strange FS bug you were thinking of? (Ideally with an historic example which could have affected a backup drive, too) Oct 15 '15 at 10:50
  • Well, normally you don't copy the file system, but the files. This means you just transmit the data, and the files are actually created again in the target file system. This way "FS bugs" don't play any role. Only if you backup the file system as a whole (e.g. dump it with dd), you will actually copy your "FS bugs". But then, you normally don't bother about the target file system and dump the copy to the hard drive partition directly.
    – R Kiselev
    Jan 30 '18 at 13:12
  • @RKiselev yes and no. You still have some advantages/disadvantages with different files systems, like HFS+ getting corrupted out of no where and other file systems that support checksumming it's files (other just the metadata), so it's not a bad idea to have different fs
    – hashier
    Jan 31 '18 at 19:09
  • @RKiselev Or APFS just losing files youtube.com/watch?v=k60NvrJnNOY when coping. Which is exactly your named case. There was another bug just couple of days/weeks ago that the file system executed files if they had a specific name. Another FS would have not done that, so FS bugs definitely can bite you, even thought I admit it's very unlikely but since the cost is almost zero two have two drives with different FS I don't see any reason not to do it. (Further different FS have different features, very little FS support for example check summing of files)
    – hashier
    Feb 19 '18 at 8:44
  • 1
    Actually it was helpful; it helped me gauge the situation — I find that helpful on Stackexchange. The video is relevant, it has me a bit more worried about that my networked Time machine backups are in sparsebundles, even if they're HFS+.
    – Andreas
    Nov 18 '18 at 17:32

Not the answer you're looking for? Browse other questions tagged or ask your own question.