On Wed, May 21, 2003 at 12:46:55PM -0400, Paul L Lussier wrote:
> In a message dated: Wed, 21 May 2003 18:22:28 +0200
> Olivier Fourdan said:
> >1) Does migrating the repository data from an architecture to another
> >one requires any special procedure ? For example, what if someone hosts
> >his repository on a Win32 (or MacOS X) system and later wants to migrate
> >to a Linux or UNIX server for hosting his data under subversion.
> You should be able to simply 'svnadmin dump' the repo on one system,
> redirect to a file, then 'svnadmin load' that file on the other
Yup. The svnadmin dump/load and the corresponding dumpfile format is
compatible (it's a text file) as long as you don't try to run any newline
conversion on the dumpfile.
> >2) Does upgrading Subversion requires (or will require in a near future)
> >any special procedure for the existing repositories ? Basically, if
> >someone uses Subversion right now, does a migration is to be planned
> >when Subversion 1.0 comes out, for example ?
See my previous email.
> >3) Backuping the server data (Berkely DB) requires a shutdown of the
> >database or the backup can be performed with Subversion being running
> >and accessed ?
> Well, Berkeley DB is not like a typical RDBMS in that there's a
> process to shut down or start up. You can copy the db files anytime
> you want, however, you may wish to guarantee that there is no
> activity operating on those files when the copy is made. So, it is
> recommended that you prevent access to the Subversion server while
> the backups are being performed. This may be as simple as shutting
> down Apache, or, as complex as preventing SSH access to the system as
This is incorrect.
Berkeley DB is fully capable of performing a hot backup. You can back up the
repository even while commits are being performed. There is no need to
prevent operations, stop servers, or anything like that.
As Garrett pointed out, see the hot-backup.py script. The key is simply how
you copy the database files (the order of the copies).
> >4) How does Subversion scale ? Can it support (or will it support) huge trees ?
> Can you define 'huge' trees? There are some large repositories
> listed on the Subversion website on the order of 1+GB. There is at
> least one which is over 7GB. Is that large by your definition?
Right. What is "huge" ? Our current design *should* scale to very large
trees and very large numbers of revisions. Experience so far seems to
indicate that is going well.
Greg Stein, http://www.lyra.org/
To unsubscribe, e-mail: email@example.com
For additional commands, e-mail: firstname.lastname@example.org
Received on Thu May 22 00:18:39 2003