Transferring Large Files w/ Rsync - Initial Xfer

[Wayne Davison]

On Mon, Jan 03, 2005 at 04:40:03PM -0800, d c wrote:
> The source files have slightly chnaged and I now am trying to sync
> them.  However,  it seems that rsync has to copy the entire file
> itself the first time inorder to build the index (hash) needed in
> order to have more efficient transfers in the future.

The hash is determined at runtime.  Rsync needs to read the whole file
on the receiving side (to generate the hash), the sender then uses that
hash data to send any unmatched data (by reading its version of the
file), and then the receiving side creates a duplicate file based on its
original one and the data from the sender.  It then moves this new file
over the old one.

So, it's hard to know what you're referring to.  You may just be seeing
the new temp file being created (which will always happen unless you use
the --inplace option, which is new for 2.6.3, but is only efficient for
things like log files that get appended data).  You might be seeing a
new copy because you accidentally put the file into the wrong spot on
the recieving side (try sending a small file to the same directory as
the big file to test this).  Or, the file may be compressed, in which
case a small change to the uncompressed data results in a massive change
in the compressed data (there is a patch available for gzip that makes
its compressed files more rsync friendly, if you need it).

..wayne..