Python wrapper for git that allows it to read a GET-only HTTP mirror of repositories that use Large File Storage (LFS). It effectively extends git's so-called dumb git protocol with the capability to serve LFS files.
This means you can use S3 to host mirrors of LFS repositories, without additional infrastructure. However, no authentication mechanism is currently supported. From the point of view of the calling code, it must be a publically readable bucket.
It works by temporarily firing up an LFS server during the lifetime of the git command.
pip install git-lfs-http-mirror
Configuration is by command line arguments
python -m git_lfs_http_mirror
--upstream-root 'https://my-bucket.s3.eu-west-2.amazonaws.com/a-folder'
--bind '127.0.0.1:8080'
-- git clone http://127.0.0.1:8080/my-repo
You can make this slightly less verbose my making a wrapper bash script. If the below is in an executable file glfsm
in your PATH
#!/usr/bin/env bash
exec python -m git_lfs_http_mirror \
--upstream-root 'https://my-bucket.s3.eu-west-2.amazonaws.com/a-folder' \
--bind '127.0.0.1:8080' \
$@
then you can call
glfsm -- git clone http://127.0.0.1:8080/my-repo
The server configued as upstream_root
should be a static server, serving copies of git repositories, for example S3. Each copy can be created using:
git clone --bare https://server.test/my-repo
cd my-repo.git
git lfs fetch
git update-server-info
and then uploaded to the server in its own folder. If the server is S3, this can be done using the Upload folder feature in the AWS S3 Console.