Hacker News Reader

Comment by ciniglio | Hacker News Reader

Do you do anything to check against corruption (e.g. bit flipping)? I've been experimenting with md5deep recently to accomplish this, but it's a little time-consuming. I'd be very interested in hearing about more efficient ways of dealing with this problem.

View on HN

No, except daily snapshots. If I would find main version corrupted I can always recover old version a few month back. I

I found bit flipping problematic only when transferring larger amount of data over network.

If you are not checking it, how do you know your drives are not flipping bits?

Just use ZFS. It works.

But, as with the old adage "RAID is not backup", don't forget that ZFS is not a backup either.

ZFS has bugs like any other codebase, and metadata corruption will make your whole pool unrecoverable. RAID-Z and block checksums won't save you from that.

I speak from personal experience with ZFS. Backups are important, no matter what filesystem you use.

ZFS is a good answer to the question "Do you do anything to check against corruption?"

If you're doing differential backups on external USB disks (generally of dubious reliability), you definately want some integrity checking. Bitrot happens!

I'm using ZFS on my home server. The backups of my main system are created with the ubuntu desktop utility and written to the server via NFS. This means I have incremental backups and the backups themselves are protected from a hardware failure using raidz. A problem with this approach is the possibility to rm -rf /mnt/backup - that's why I only start and mount the server when I need it. It's not running 24/7.

Using Windows? Try ExactFile

Reply on news.ycombinator.com