backup via an intermediate drive

Matt McCutchen matt at
Mon Feb 11 23:51:53 GMT 2008

On Tue, 2008-02-12 at 00:14 +0100, Matthias Schniedermeyer wrote:
> On 12.02.2008 00:03, Matthias Schniedermeyer wrote:
> > I guess the capacity of the transport drive is way lower than the whole 
> > storage needed?
> > 
> > Then the option is:
> > --write-batch
> > 
> > But as far as i understand it, you need a second (in this case local) 
> Strike the second, reading the next line of the man-page reveals the 
> "--only-write-batch"-option. :-)

Using --only-write-batch instead of --write-batch does not remove the
need for a local copy of the backup server's contents that rsync can use
to determine what data to include in the batch file.

Cavan, if you don't like having the extra copy, you might be better
served by GNU tar's incremental archives (--listed-incremental, see
`info tar` for more information).  Tar maintains a state file on the
source machine to remember what changes have been put into incremental
archives.  Each time you ask tar to make an incremental archive, it will
look at the state file to see what changes to include and then update
the state file.  If you transport the incremental archives to the backup
server and extract them there in order, you will maintain a copy of the
source directory.

(I'm recommending GNU tar's state-based incremental mode because it is
ready-to-use now.  It would be possible to add a similar state-based
incremental mode to rsync (or perhaps Unison), but AFAIK no one has done
it.  If you like, you could script your own incremental mode around
rsync using --files-from.)


More information about the rsync mailing list