YH Yang wrote:
>
>
> 2008/1/2, Ryan Schmidt <subversion-2007b_at_ryandesign.com
> <mailto:subversion-2007b_at_ryandesign.com>>:
>
> On Jan 1, 2008, at 22:54, YH Yang wrote:
>
> > 2008/1/1, George Georgalis:
> >
> >> Can I get some clarification on the --clean-logs
> >> switch for svnadmin? What are "redundant Berkeley
> >> DB log files" and when would they arise?
> >>
> >> I'm using hotcopy weekly and doing tar bzip2 to the
> >> result. If there is an md5 difference between
> >> my hotcopy archive and the real backup I move the
> >> new archive to the backup.
> >>
> >> My problem is every time I tar bzip2 a hotcopy, I
> >> get a few bytes difference, even when there is no
> >> change to the repo. I'm looking to get an exact
> >> archive from each hotcopy taken when the repo hasn't
> >> changed.
> >>
> >> Maybe the access times are updated or not duplicated
> >> with hotcopy? Is there any way to change this
> >> behavior?
> >
> > I saved the "svnlook youngest" result to a plain text file to
> > backup what exactly revision in the last hotcopy. And then
> > before hotcopy, I check this plain text file first to determine
> > skipping or going.
>
> That's fine for new revisions, but will not result in accurate
> backups if you allow changes to revision properties through the use
> of a pre-revprop-change hook.
>
>
> You're correct. In my case, the revprop modification at both client
> side (through the use of a pre-revprop-change hook) and server
> side (through svnadmin setlog --bypass-hooks ...) can not be detected.
Perhaps you should consider a "post-revprop-change" hook that updates
that file or otherwise flags the revprop changes for the backup script
to know it should be backed up again?
Deven
Received on 2008-03-24 17:59:52 CET