Re: [rdiff-backup-users] atomic increment files?

[Maarten Bezemer]

Hi,

Just to fall into a nice discussion...

On Tue, 10 Mar 2009, Andrew Ferguson wrote:

> Interesting idea, except that rdiff-backup doesn't make a snapshot every 10 
> backups, it makes a snapshot every 10 times that a file has changed.

I wasn't aware of that. Given the fact that I do have quite a number of 
quite big files that change almost daily, this means I'm adding those 
sizes to the backup size every (about) 11 days? (Contents are generally 
not easily compressable).
Although I understand the reason for not having to go back over tens or 
hundreds of increments, this idea kind of defeats the purpose. These files 
are just appended to regularly, but the first (huge) part doesn't change.
Also, I can't seem to find any evidence when browsing the 
rdiff-backup-data/increments tree. Maybe this is a 'feature' for the 
development tree?

To go back to the initial discussion, I'd like to add that being able to 
store old increments to a read-only location such as a DVD would be really 
nice. So that you could in fact still go back to previous states but not 
need to have all previous states in the rdiff-backup-data tree.

What I mean is doing something like --extract-older-than doing almost 
exactly what --remove-older-than does, but move the increment files to an 
external location (maybe even creating a tar of them) instead of deleting 
them.
I think having some fun with an external program and regexps could do this 
trick as well, but cleaning up an rdiff-backup tree using rdiff-backup 
sounds like a more clean way to do things.

After extraction of the increment(s), one could tar/zip/... the files to 
one big file and store it somewhere. For backup runs that have already 
been completed, the issue of interrupted backups needing to be rolled back 
is not an issue anymore.


Just my two cents..

Maarten