I want to make a virtual clone of my production server so I can do some testing without the chance of messing up the actual server. I mainly want to try out stuff like a dist-upgrade, or updates of installed packages etc. I don't have physical access to the server (well, I do, but it's a 3 hour drive) so I need to be able to make an image and download it over my internet connection.

The server hosts a few directories with a huge amount of images of which I already have a local copy so I don't want to include those in the disk-image.

The clone is going to run on a virtual PC.

So what would be the best way to make such a disk-image (keep in mind that I want to be able to exclude certain directories from the image)?

Build clean guest, install / updates packages until it has the same status as the production host then rsync data?

It is common to clone a system without some folders. As to the method, that would be your choice.

In most cases, linux doesn't care about being cloned as long as there is support. Cloning of systems running databases and other apps need special consideration.

It is also a common task to create a golden or reference image to run as unSpawn suggests.

That's not as easy as it may seem since the production server is already running quite a few custom compiled packages.

I was hoping I could do something as simple as a "tar". Can I?

Yes, it's running databases (MySQL).

Anyone? Please?

Tar may or may not correctly copy a running database. I'd think you'd need to stop the service and then tar it or rsync it. There are many web pages on how to do that usually.

Tar in general is a very old and trusted tool that compacts a single file or any number of groups of files. One generally further tries to compress it for storage via some pipe.

I'd say check the MySQL documentation as it shows you how to make backups. And I'd still argue rsync trumps tar when it comes to efficiently mirroring (whole, sparse, partial) files taking into account permissions, ownership and extended attributes and making use of compression (default) and bandwidth limiting and offering log files and statistics. But it's your choice, well-informed or not.