Re: hdup 2.0.1 argggg

Boris Goldowsky <[email protected]> Thu, 02 Dec 2004 13:19:52 -0500
Newsgroups gmane.comp.sysutils.backup.hdup.general
Message-ID <[email protected]>
[On 17 Nov, @ 21:38, Boris wrote in "Re: [hdup-user] hdup 2.0.1 arg ..."]
> > Ie, if I move /opt to /foo, then /foo/fileX ought to be backed up on the
> > next run even though it may not have been modified for a year.  I don't
> > know of a way to catch this without keeping a list of what files were
> > backed up "last time".  If pax doesn't have that ability, you may want
> > to implement it within hdup.

On Wed, 2004-12-01 at 15:46 +0100, Miek Gieben wrote:
> If you move /opt to /foo the ctime changes - at least under Linux.
> (if you put something in a dir the mtime changes). So when you
> recursively walk a tree and encounter such a directory - you must
> backup the entire subtree, no matter what. This is a bit like the
> opposite of the .nobackup keyword.

That would work, but it may be overkill - there are other reasons the
directory ctime could change (e.g., deleting a file in it).  Backing up
the entire subtree rooted at that directory would be unnecessary in that
case, and could result in much-bigger-than-necessary backups.

I think the only way to know what really needs to be done is to have a
list of what files were backed up the last time -- hence the concept of
the listed-incremental used by tar.  Anything not on the list is backed
up unconditionally; anything that is on the list has its modification
time checked and is only backed up if recently changed.  If a file is on
the list but no longer exist, that fact is noted so that it can be
removed during a restore of the containing directory.

Bng