hashproduct+rsync at gmail.com
Mon Sep 24 23:50:45 GMT 2007
On 9/24/07, Stephen Zemlicka <stevezemlicka at gmail.com> wrote:
> I think I got it (with the help of a friend). You can, in fact, rsync to a
> mapped drive efficiently. You must include --no-whole-file. My 100MB file
> only transferred a few MB using that method and it opened up with the
> changes just fine. Now it works just like I want. It's not the quickest
> but definitely quicker than uploading hundreds of MB for each database.
Of course --no-whole-file will make rsync report a smaller number of
"bytes sent", but is it actually reducing the amount of I/O to the
mapped drive? If so, I find this really surprising. The one
explanation I can think of is that the network filesystem has cached
the file, so reading the basis file costs nothing and writing only the
changed areas (due to --inplace) is a big saving. Is this the case?
More information about the rsync