Question

I want to use GIT as a Backup solution.
I have a windows shared network in my Company and a Centos Server.
Each user have a unique shared folder, where is located all files of user. We have 2 developers, they can have a git repo(s).

How Can I make the entire folder of user a GIT repo, a perform usual git operations, commit, push, etc, without import if subfolders have git repos?

Note: Submodules is not a solution, I don’t know what repositories is under the main folder of user.

Answer 1

There are a number of reasons why Git does not make a good backup solution:

• you cannot have nested Git repositories — there is no way to coerce Git to treat a nested .git directory like a normal directory
• it doesn’t save complete permissions and ownership
• its index format does not scale to many files
• it does not handle large files well (it runs out of memory because it needs to put the entire file in memory to compress — or at least that was true about a year ago, maybe they’ve fixed it)
• while the compression is good for source code, it’s not very good for virtual machine images, etc.
• backing up to a remote system involves backing up locally first
• no defense against backup file corruption

That being said, there are some reasons why Git would be awesome as a backup solution:

• every backup is a complete snapshot — you don’t need to restore an old full backup followed by some number of incremental backups in order to get the latest files
• ridiculously good deduplication performance
• good space efficiency in general
• great support for efficiently transferring backups to remote systems

My recommendation: take a look at bup. It’s not very mature, it’s not seeing much active development, and it lacks some important features, but it has a lot of promise. It’s an open source backup program based on Git. They’ve kept the things Git does well, but replaced/augmented some of the things Git does poorly (for backups).