Re: [Duplicity-talk] Seeding a remote backup

[edgar . soldin]

On 13.06.2012 19:03, T.J. Crowder wrote:
> Hi all,
> 
> Duplicity looks like it could well be the answer to a backup issue I'm 
> having. Very cool indeed.
> 
> I have a lot of data (~3TB) that I want to back up to an HDD and then send 
> off-site, keeping the backup up-to-date on a daily basis (overnight). I have 
> a mix of small files and very large ones (virtual machine disks). There's 
> definitely duplication, although that wouldn't be my primary interest in 
> Duplicity.
> 
> Two questions related to this:
> 
> 1. Is it possible for me to use Duplicity to do a full backup to an external, 
> locally-connected HDD, then post that HDD to a remote site (where they've 
> agreed to hook it up), and then have Duplicity maintain that backup vis 
> ssh/scp?

absolutely

> 2. If so, with the VM disk images, for (say) a 200GB disk image, how much 
> transfer would Duplicity require (very roughly) to identify which parts of 
> that file had changed so it could send the deltas?

generally duplicity uses librsync with the included rolling checksum algorithm. 
i seem to remember a 9MByte frame, but am not sure about that. try to find out 
what librsync uses by default or how it works in general.

additionally you might simply use duplicity for a while only locally and 
observe the size of your incrementals.

be aware that duplicity works perfectly for smaller backup sizes, but was never 
developed to backup terrabytes, so you might stumble along the way.
also. it is suggested to do fulls regularly because a bitflip/error in one 
volume essentially makes the backup chain defective from that point in time on.

..ede/duply.net