Re: [rdiff-backup-users] Resuming initial backups

[Maarten Bezemer]

On Fri, 22 Aug 2008, Josh Nisly wrote:

> Ah, that would be the difference. Typically, my end users download and 
> run rdiff-backup themselves.

Ah, nice. So in fact we need a user-friendly tool that is less picky about
who its friends are :-)

> My thinking is 
> that if there is no current_mirror marker, and one or fewer error_log 
> and mirror_metadata files, there cannot be any historical increments, so 
> the operation cannot lose data.

Checking the increments/ directory to see if it exists, and if so,
checking if it's empty, could be done explicitly. That is, you should not
"suppose" anything based on some information, if you can have the definite
answer.

So, checking for a non-existant or empty rdiff-backup-data/increments
directory, AND the absence of a current_mirror, AND the presence of at
most one mirror_metadata and error_log file, could be reasonably safe.
You could also try decompressing the mirror_metadata file to see if it was
complete. That kind of stuff. If all that you remove are empty or broken
files, that would be acceptable... I think...

In your case, with people using rdiff-backup themselves, this might
actually be a better option than forcing them to use --force in their
command lines. It prevents the "it didn't work without --force the first
time, so I left it in" excuse.

-- 
Maarten