[ale] Backup strategies

[Pat Regan]

Chris Kleeschulte wrote:
> Excellent point. One question though:
> 
> 
> would it be imprudent of me to full backup once and then rsync from  
> then on or am I being paranoid about that? I mean what if something  
> corrupted the full back that you did once, then the incrementals  
> would also be bunk, correct?

If you are using straight plain old rsync the checksums are computed
during each transfer.  That means that if your backup has a corrupt file
it would just show up as a change.  The good file would then be copied
over the wire.

With most of the rsync-style backup methods your most current backup is
stored as the full backup and the history is stored as differentials.
So if your current full backup has a corrupt file, that corruption will
be moved to a differential and the current full will be repaired.  I
hope I'm explaining that clearly :).

The common rsync+hardlink backup gives you the least space efficient
differentials.  Entire copies of changed files are stored on disk.

rdiff-backup computes and stores, ummm, reverse differential binary
patches, I suppose.  With rdiff-backup you can retrieve the most recent
backup with simple filesystem operations.  Previous revisions require
the software to rebuild the file from the differential patches.  It is
amazingly space efficient.

Pat

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: OpenPGP digital signature
Url : http://mail.ale.org/pipermail/ale/attachments/20080912/6d4a9ead/attachment.bin