Git & external backup. Subdirectory is empty, I'd rather it not be.

Question

I would like to use Git for creating an external backup system.

This was inspired in part by the answer at What is the best way to back up an entire git repository on external disk?, where the following is suggested:

You can also get incremental backups by just starting a new repo on the external disk (git clone) and then pull the changes every time (git pull). There is not much difference between a backup and a working copy in a distributed system like git.

(I use an unbare repository on the external for convenience.)

My dilemma is: On the partition I'm backing up to the external, I have subdirectories which are themselves git repos, each with upstreams at remote repos.

In other words, via ascii art:

          upstream
external ----------> partition
                         |                    upstream
                         +----- subdirectory ----------> remote

The subdirectories that are themselves under version control, show up empty on my external.

I would like git to not skip these directories, but rather, to pull the contents of versioned subdirectories as well.

I have wondered if submodules is relevant here, but the documentation was difficult. (I never found a definition of submodule - or an example - that suggested it might be appropriate for my case.)

At Git is ignoring .git directories in subdirectories, the question includes the line

Git will not allow me to add the subdirectories .git folders.

The responses are discouraging, they suggest to me that git is not designed to include this functionality.

But I'm hopeful.

Answer 1

Submodules are repos recorded in a parent repo as gitlinks (special entries in the index, mode 160000), that is as the exact SHA1 represented by that submodule repo.

The goal is not just to include a subrepo, but to include a specific version of that repo (hence the gitlink, to record the right SHA1 to which that subrepo should be checked out).

In your backup-script scenario, you should go in each submodule, and backup in their own bare repo on your external drive.

You will find the path of those submodules in your .gitmodules file in the main repo.