Recently I have a brand new stand-alone server for my repositories and other minor thingies. And the challenging thing: moving the repositories from my home desktop computer to this server.
My setup was this: I have my home desktop computer that runs Windows XP, and the repositories were stored and managed using VisualSVN, a great tool to manage SVN repositories, if you’re running Windows. The applications were deployed on my Windows machine (for developing purposes) and on the hosting server (live, production deployment) that runs Linux. And here’s what I’ve done…
On the source, Windows machine
I’ve stopped the SVN service, so no svn updates/commits would be initiated. I’ve opened up a console and ran the following command:
svnadmin dump d:\Repositories\repo_a > repo_a.dump
Then I’ve packed the repo_a.dump file because it’s almost plaintext and copied the archive to the destination server.
On the destination, Linux machine
I’ve unpacked the dump file and then:
svnadmin create repo_a
svnadmin load repo_a < repo_a.dump
After that, I needed to chown -R www-data:www-data ./* to make the files belong to apache, and then add the authentication back.
This is just about it for the relocation of the repositories. But we need to do one more thing: alter all the checkouts to point to the new repositories. This is done in a way on a Windows machine, and a bit different on a linux machine.
Checked out repositories fixing
To make all the deployed checkouts point to the new server, you need to edit every file called "entries" that is located in every ".svn" folder of a checked out repository.
To do that, on Windows you can use grepWin, a lovely tool that does all the work:
There you have it. On linux, you'll need perl:
find ./ -name "entries"|xargs perl -w -i -p -e "s/svn\.example\.com/new\.server\.com/g"
This command replaces all the occurences of "svn.example.com" with "new.server.com" in all the files named "entries" in the current folder and below. Replace for your own needs.