[Noisebridge-discuss] Data integrity, rdiff-backup, Reed-Solomon codes

[Michael C. Toren]

On Sun, Dec 13, 2009 at 11:39:56PM -0800, John Magolske wrote:
> Though I've heard good things about dirvish, I've read that its reliance
> on hard link trees means "...Apart from not being a 1:1 backup (you lose
> hard links!), the filesystem metadata storage explodes for any
> reasonable sized filesystem and any reasonable frequency of backup."
> http://lists.debian.org/debian-user/2009/07/msg02022.html

I'm not familiar with dirvish, but I don't believe usage of hardlinks
necessarily means you lose the ability to perform a 1:1 restore.  For
example, if your restore method involves using rsync to copy data to the
new disk, you should be able use rsync's -H argument to preserve hard
links on the new device.  Does anyone have experiences to the contrary?

(His other argument, that it explodes the meta information in your file
system, is valid, but it needs to be weighed against the cost of exploding
disk utilization of storing multiple copies of the same file.)

-mct