[Dirvish] Local vs remote

Noel Kelly nkelly at citrusnetworks.net
Wed Feb 7 02:22:38 PST 2007


I am preparing a remote server to be dirvish-ed each night and am 
currently synchronising the files to a removable USB disk.  The idea is 
then to bring the USB disk to the backup server and start dirvish off 
with the initial image from the disk.

Some of the files that are going to be backed up are these ugly Outlook 
PST files.  One of these is around 450Mb.

I use this command to rsync the files to the USB disk:

rsync -vrltH --delete -pgo -D --numeric-ids -x --delete-excluded 
--exclude-from=/root/exclude.list /home /mnt/USBdisk/roar/

When this runs it labours over the big PST file, taking ages to complete 
(~30mins for this one file).  It seems to recopy the entire file each 
night rather than just the differences.

However when I run the dirvish test via SSH (using a 50Mb PST file) with:

command="rsync  --server --sender -vlHogDtprxz --numeric-ids . 

it clearly only transfers the differences.

There is a big difference between the two PST file sizes but this is 
tempered by the remote link being a painful 250K as opposed to a USB 1.1 
bus which even if running at 1.5Mb/s should still transfer a few Mb of 
changes quite quickly).

I did not think (and nor can I find) any reference that rsync and/or 
dirvish change adapt their algorithms for local file systems and remote 
file systems.

What I am worried about is moving the initial image to the backup server 
and then finding dirvish is trying to transfer these huge PST files each 
night rather than just the changes!  I have tested moving large PSTs 
from other remote systems and it has always worked as expected with the 
differences only being copied.

Are my fears unfounded?  Is there a difference between local and remote 
copies?  I have the latest version of rsync on both sides.

Thanks for any ideas,

More information about the Dirvish mailing list