Tonight I migrated my VPS from vpslink which I no longer recommend to GeekStorage. They both run Debian on OpenVZ/Virtuozzo, so I figured it wouldn’t be hard to migrate over.
I didn’t really find much about migration on the user side. Most of the migration pages were about how to migrate containers from the host server’s perspective. I did manage to cobble together enough clues to figure out what to do.
Basically to summarize the essential points, you can just copy everything from the old server to the new, clobbering whatever is there, EXCEPT for the following:
/aquota.group
/aquota.user
/etc/fstab
/etc/hostname
/etc/hosts
/etc/resolv.conf
/etc/network/interfaces
/etc/network/interfaces.template
/dev
/proc
/sys
In more detail:
On vpslink I have Debian on OpenVZ. So when I signed up the GeekStorage, I requested Debian on Virtuozzo. OpenVZ is the open-source variety of Virtuozzo, so they are close enough.
When GeekStorage said the VPS was ready, I logged in and used rsync to copy everything from the old server to /backup/old. That took a few hours. Then on the old server, I shutdown all the services (sendmail, apache, mysql, tinymuck, etc.) and then ran rsync again. That updated everything that changed while the first rsync was running, and was fairly quick.
Once that was done, I put the server in repair mode. This creates a temporary container with the real container’s contents mounted under /repair. I logged into the repair mode serve. First I moved everything in /repair except aquota.*, backup, dev, proc and sys to /repair/backup/orig. This moves the new VPS’ original OS files out of the way.
Then I moved everything in /repair/backup/old except aquota.*, dev, proc and sys to /repair. That moves the old VPS’ OS files to the new VPS’ root directory. Then I copied the etc files (see above) from /repair/backup/orig/etc to /repair/etc.
Caveats: Remember that there are probably hidden dotfiles in the old and new root dirs you need to move too. Also remember to always work in /repair when in repair mode. It’s easy to forget and start twiddling in /etc instead of /repair/etc. That disappears as soon as repair mode is finished.
Then I took the server out of repair mode, which boots into the VPS container again. And miraculously, it booted up just fine with everything looking just like it was on the old VPS. I had to twiddle a couple of other files with hardcoded IPs, but the OS came up just fine. I also had to update a bunch of things in DNS and wait for DNS caching to time out.