Practicality of backing up huge hard drives (was Re: [mythtv-users] Dieing Hard Drive)

[Tony Godshall]

According to Johannes Becker,
> My experiance with HD backup is not so good. Usually you have only one
> drive that holds one backup (maybe two). But what can happen is that a
> disk goes down slowly and has failures that you copy to your backup
> before you notice that. The HD backup is not helping then, because you
> only have the broken copy there.

I find cp -al and rsync to make a good combo.

You can cp -al to make a shadow tree (backup.yyyymmdd) and
then do your rsync (over ssh) with --backup.

The two new tree should transfer with minimal bandwidth used
and be identical to the current state, and the old one
should have everthing the way it was at the last backup, and
you can compare and diff and du just fine, but unchanged
files take up no extra space.  Of course the old backup does
have timestamps updated, so be sure you can tolerate that,
but you have the best of differential and full backup.

So when you detect corruption you can just go back to the
backup that was good as your starting point and collect
stuff from later ones selectively.