[NCLUG] I think

[Zak Smith]

On Fri, Feb 21, 2014 at 09:26:40AM -0700, Stephen Warren wrote:
> I'd suggest simply copying the whole tree with rsync, to a separate
> (dated) backup tree. That way, you can easily do something like:
..
> As an extension to this technique, you can then hard-link your backup
> trees together, so you can keep N backups in the space of 1 backup,
> ignoring effects due to actual changes to the files. Of course, you need
> some code to prune old backups so that the backup sizes don't grow in an
> unbounded fashion if your files keep changing.
>
> rsnapshot appears to be a tool to automate this. IIRC, I derived my

The cool thing about doing this is rsync's "--link-dest" feature (ie,
you give --link-dest=<the most recent snapshot directory>).  This does
not use storage for all the files that have not changed.   rsnapshot
is a wrapper around rsync that is based on this feature.  The downside
of this is if you get one screwed-up snapshot directory, you may lose
the "hard link" continuity, ie, if you had the backups

20140101 - original backup
20140201 - next backup, links non changed files back to 20140101
20140301 - this one is partially screwed up becasue of a power failureor something
20140401 - this one will duplicate the data instead of re-using via
   hardlinkes from 20140201 (20140101) for files that are missing in
   20140301.

Another alternative is to run ZFS with snapshotting on your
destination rsync volume.  Or run it on your primary volumes (but this
is not a "backup" more like version control).




--
# Zak Smith    mobile 970-232-4468