PyPI Server

This is a simple PyPI server that can be used to host internal packages and versions of packages that are suitable for use with proprietary products.

Usage

If you're not dealing with the Git repo but running the container directly from the Docker index, you can use commands such as the following:

sudo mkdir -p /srv/pypi             # local directory where packages reside
sudo touch /srv/pypi/.htpasswd      # credentials file for adding packages
docker run -t -i --rm \             # remove container when stopped
    -h pypi.local \                 # hostname
    -v /srv/pypi:/srv/pypi:rw \     # host packages from local directory
    -p 8080:80 \                    # expose port 80 as port 8080
    --name pypi \                   # container name
    codekoala/pypi                  # docker repository

Once running, you should be able to visit http://localhost:8080 to see the landing page for your very own PyPI server.

You can add Python packages to the server simply by including the tarballs, zips, wheels, eggs, etc in your /srv/pypi directory.

Configuration

There are some environment variables that may be set to override the default behavior:

PYPI_ROOT: path within the container where packages will be stored. Defaults to /srv/pypi.
PYPI_PORT: port to bind to receive requests. Defaults to 80.
PYPI_PASSWD_FILE: path to authentication file. Defaults to /srv/pypi/.htpasswd.
PYPI_OVERWRITE: allow existing packages to be overwritten. Defaults to false.

Building Your Own

You can build a new, up-to-date version of the container by cloning the Git repository and using the following command:

make build

This will create a new container that just has the latest version of pypiserver installed and ready to serve packages out of /srv/pypi. To use this container, run:

make run

This will spin up the pypi-server command within container, and it will be exposed on port 8080 on your host system. To test that the container is working, visit http://localhost:8080 in your browser.

Adding Internal Packages

Internal packages may be uploaded to this PyPI server quite easily. The first step is to create a user account:

htpasswd -s /srv/pypi/.htpasswd yourusername

You will probably need to re-run make run each time you update the htaccess file, as it will copy the password file to the correct location before launching the server.

Alternatively, you might be able to just copy the htpasswd file to /srv/pypi/.htpasswd after each change without restarting your PyPI container.

This command (included with Apache on most distributions) will prompt you for a password for yourusername. You should use a more appropriate username, and enter a password that you want to use to "secure" your PyPI uploads. Then edit your ~/.pypirc (create it if necessary), replacing both yourusername and yourpassword with the values used with the htpasswd command:

[distutils]
index-servers =
    pypi
    internal

[pypi]
username:pypiusername
password:pypipassword

[internal]
repository: http://localhost:8080
username:yourusername
password:yourpassword

Next, you should be able to go into any Python project with a valid setup.py file and run:

python setup.py sdist upload -r internal

Assuming the container is online and your credentials are correct, this should add a package with the project contents to the internal PyPI server.

Adding Third Party Packages

Third party packages can be mirrored on the PyPI server by using a command such as:

pip install -d /srv/pypi pkgname

If you have a requirements file for a project's dependencies, you can easily mirror all dependencies by running:

pip install -d /srv/pypi -r requirements.txt

Be careful to use the correct version of pip--sometimes you might want to run pip2 and other times pip3.

Updating Mirrored Packages

You can update the packages mirrored on the internal PyPI server by running:

pypi-server -U /srv/pypi

Each package in the repo will be checked for updates, and instructions for updating the repo with the latest packages will be displayed.

mark86092 / docker-pypi