Re: [Duplicity-talk] How to use duplicity for local backup

[Nate Eldredge]

On Fri, 24 Apr 2020, MRob via Duplicity-talk wrote:

> Hello, I want to ask opinion about backup tactic please.
>
> I found a backup system that using only rsync to a local server. A few 
> scripts run on the server that create tree of historical directrories 
> that are filled with hard links to the main rsync target. I never saw 
> this but my understand of the benefit:
> * saving space for historical backups because they only use inodes
> * simple scheme to use rsync and ln
> * But am I correct, to understand if a file *change* then the historical 
> data is lost so its a shortcoming of the scheme.

Well, it could work either way.  Maybe it only keeps the latest version of 
files that change, maybe it keeps all past versions.  I can't tell from 
your description alone.  It'd help if you would provide a link to the 
system you are talking about.

Duplicity uses the rdiff algorithm to only track the changes in a file's 
contents.  So if you have a 1 GB file and you add 1 KB, duplicity will 
only need an additional 1 KB (or so) to make an incremental backup of that 
file, and you only use a total of 1 GB + 1 KB of backup storage, while 
maintaining the ability to recover either the old or the new version. 
With the system you describe, it sounds like either you lose the old 
version, or you use 2 GB of backup storage.

Note this does not apply if files are moved: duplicity currently doesn't 
detect moved files, and treats them as if the file was deleted and an 
unrelated file created under a new name.  If you have a 1 GB file and then 
rename it, even if you don't change the contents, duplicity will use 2 GB 
of backup storage.  It's not clear how the other system you describe would 
handle that.

> Is this scheme popular?

I've never heard of it before.

> Is it useful in comparison to local-rsync

They seem very different.

--
Nate Eldredge
address@hidden