Re: [rdiff-backup-users] instrument data backup

[Adrian Klaver]

On 07/29/2012 04:58 PM, Z F wrote:
> Hello Again
>
>
>
>
> > >   I need to backup data from a scientific instrument. The problem is that the
> > instrument has limited space, so
> > >   data has to be moved from the instrument onto other drive. From time to
> > time, the old data needs to be copied
> > >   back to the instrument. Rdiff-backup is handy because regular
> > "cp" command will restore the latest version of the datafile.
> > >   Thus users can be given read-access to the backup and they can restore data
> > they need by themselves.
> >
> > The above confuses me. You say the old data needs to copied back, but
> > then you say rdiff-backup is handy because it allows the latest version
> > to be copied back. Not sure I am following the logic:)
>
> Sorry for not being clear about what I would like to do.
>
> Basically, I need an easy access to files which are deleted from the system.
> The files which are present on the system at the backup time and were deleted
> before a subsequent backup can be easily found (using find) and restored with a 
> cp command.
> Files which were deleted from the source directory before the last backup are
> also deleted from the "mirror" but are present in the rdiff-backup history.
>
> I do not see an easy way to browse or search the deleted files by creation date 
> or file name.
> I see
>
> rdiff-backup --list-at-time ??D
> out-dir/subdir
>
> which will work if I know the date when the file was created. Is this my only 
> option?


http://www.nongnu.org/rdiff-backup/rdiff-backup.1.html

See section Time formats.

Using   --list-changed-since you can go back an interval or a 
backup(where 0B is the current backup and 1B would be the next oldest)


> > Still not sure I am following correctly. Still, why not move the data
> > from the instrument to a directory on the hard drive and then
> > rdiff-backup from that directory to another directory?
>
> Ideally, old data do not change only new data are created. But uses sometime do 
> stuff
> to old data and I have to keep track in backup of what they do. It is possible 
> that they
> changed old data by mistake or on purpose. thus, moving data to a different 
> drive does not
> always work. If old data gets modified, I need to create a "revision". It is 
> rare, but might happen.
> I thught rdiff-backup will keep the revision history for me.


It does.
> Should I consider CVS revision control system as a backup tool? it works well 
> on text files, not sure
> if it good for binary data. in this scenario, new data will be simply added to 
> the repository. The old data
> gets revisions (if it happens) and I backup the CVS repository . CVS repository 
> can be searched for filenames
> so the old deleted files can be found...
>
> The CVS route does not sound correct to me, though.

Yea, not sure how this helps. To get revisions you need to do commits. 
You still need to know that a file was changed to schedule a commit.

> Do you have any thoughts or suggestions? Am I talking nonsense again?

I still think the below will work:

instrument --> copy_directory --> rdiff_mirror_directory

Per another post you could use rsync from the instrument to 
copy_directory. If you run rysnc without the --del switch it will keep 
the deleted files in copy_directory. Changed files will naturally be 
changed but the revisions will be kept in the rdiff_mirror_directory.

> Thanks
>
> ZF


--
Adrian Klaver
address@hidden