On Wed, Aug 18, 2010 at 2:21 PM, Christopher McCrory <chris...@gmail.com> wrote: > Hello... > > I wrote up a short perl program to backup a subversion repository via > amanda ( http://www.amanda.org ). Basically it uses the amanda > application API to call 'svnadmin dump'. It still needs a few > enhancements, but it does work. I developed it on Fedora 13 where it
This actually came up..... 4 hours ago with a professional colleague, and recently on the rsnapshot mailing lists as a something that deserves documentation. "svnadmin dump" discards all your file ownership (necessary to administering post-commit and configurations as an admin user and the database with direct access, critical for direct "ssh" based access). So does "svnadmin hotcopy", which is usually a lot more effective than an "svnadmin dump". Both are profoundly less efficient to avoid backup churn than "svnsync". > should "just work". Other platforms might have to adjust the perl path > and/or use lib dir. There are more detailed notes in the source code. > > If you manage both amamda and a svn repo, please give it a shot and > report any problems, suggestions, or successes. The best place for > reports is the amanda-hackers@ list, otherwise to me directly. The > current source is here: http://216.152.245.114/svn/amanda/trunk/ . Amanda (and its commercial version, Zmanda) are pretty cool. But I urge you to review the use of svnsync and a parallel tool to duplicate configurations and file ownership: That gives you a working repository easily switched to active use, rather than something that needs to be reconstructed from an svnadmin dump. It also reduces the backup churn: incremental backups do not need to backup the full contents of "svnadmin dump", which matters a lot for a set of large repositories.
