On Fri, Feb 12, 2010 at 3:56 PM, Justin Connell <[email protected]> wrote:
> I have a repository that has been in use for well over a year and over this > period the size on disk has grown to over 150 GB, I found that when running > svnadmin dump, that the resulting dump file was at 46 GB on disk and then > when loading the dump file into a new repository that the size on disk in > the repository folder was 8 GB in total. > > What's disturbing is the drop in disk usage from 150 --> 46 --> 8 Gig. > > Does anyone have an explanation for this? > > Or rather is there a better way of freeing up disk space back to the OS? (we > are using FS and not Berkley DB storage) Generally speaking a dumpfile ought to be bigger than the repository. Are you sure you did not have some other files that had nothing to with SVN stored in the same filesystem location? Also, generally speaking, a dump load should produce an identical repository. A notable exception is when your version changes. For example, SVN 1.6 included a space saving feature called rep sharing that is allows duplicate items to only be stored one time. One of the earlier releases, perhaps 1.4?, included some zlib compression of contents that also saved space. So if you were coming from earlier versions and dumped and loaded into newer it would be expected to save space. The 150 GB repos to 46 GB dump does not sound right though. Was the dump produced with the --deltas option? That could explain it. A reduction from 150GB to 8 GB is not expected. Most people only see 10-20% savings from these changes. However, both rep-sharing and compression would be heavily influenced by the actual repository content. To answer what I think is your main question, it is not expected that you need to do ANY repository maintenance and it is not expected that a routine dump/load should cause disk space to be recovered. -- Thanks Mark Phippard http://markphip.blogspot.com/
