Re: [BackupPC-users] Keep backup data when source host/directory changes

[Arnold K. <ar...@ar...>]

On Wed, 6 Mar 2013 10:10:17 -0600 Les Mikesell <les...@gm...>
wrote:
> On Wed, Mar 6, 2013 at 9:56 AM, Maarten <mt...@mi...> wrote:
> >>>
> >> Yes - the pool/cpool directory is full of extra hardlinks where the
> >> hash of the content is the name to quickly match up new items
> >> (plus a little extra work to deal with collisions).  Rsync just
> >> follows the existing pc directory tree though, so you do have to
> >> copy the data from each new location once, but storage is all
> >> pooled.  If the target system is local, that usually isn't a
> >> problem.  If it is over a low-bandwidth connection it might be
> >> worth some extra work to fake the initial tree with a copy of the
> >> old link tree.
> >
> > OK, thanks for clarifying this. I feel silly that I have used
> > BackupPC for years but never realized this.
> >
> > The bandwidth is not an issue, so I'm about to start a new full
> > backup right away.
> 
> I'm not sure how much temp space will be required for the initial run.
>  I think backuppc does something clever to handle matching small files
> in memory to avoid any extra disk writes, but large ones may be saved
> then subsequently replaced with links which might be a problem in an
> extreme case.

Which might be a good opportunity to split one big server with one big
backup-share into several "servers" with maybe even smaller shares.
Then you can distribute full backups of that one machine onto several
days. And the first full-backup doesn't need to transfer all files to
the backup machine before deduplicating (which is the "link"-step in
the status afaik).

Have fun,

Arnold