This docker compose setup runs the Mullvad VPN app inside a docker container and exposes an HTTP proxy using privoxy (port 8118) and SOCKS5 proxy using nginx (port 1080). The Mullvad CLI can still be used to control the application.
Mullvad VPN app works well and has easy to use CLI functionality for switching between their worldwide servers. However, when it is enabled, it will route all of your internet traffic through it and sometimes that is not ideal. If you want have VPN access for individual applications (e.g. Firefox but not Chrome) without routing all your internet traffic through it, then this setup will work with any applications that can use a HTTP or a SOCKS5 proxy server.
HTTP and SOCKS5 proxy ports can now be accessed over IP address on your network.
Previously, the http and socks5 proxy only worked when applications using them were also on the host machine. This was great if you were just using it on your local system (which I was in my case). However, if you wanted to host it on server and have your network access it, then it wouldn't work due to possibly the Mullvad app security filtering out IP addresses that was outside of the docker network IP range on your host machine.
To get around this problem, we serve the ports over the network via an nginx reverse proxy just like how web applications are served. A new container has been added using the docker host network which proxies the exposed ports on the mullvad container. The mullvad container ports are now exposed as ports 61000-61001 to try to avoid collisions with other applications and nginx now will proxy these ports on SOCKS5 (port 1080) and HTTP (port 8118) making them accessible to the network.
This setup has been tested on Ubuntu Linux and should work with all Debian-based Linux OSes. You will need the following to use this repository:
Ubuntu >=21.10
If using ubuntu impish 21.10 or greater, cgroup v2 is enabled by default. This causes the Mullvad VPN daemon to be unable to start up inside a docker container without the docker run --cgroupns=host switch. Since the cgroupns switch has not been implemented in the docker compose spec yet, a temporary fix is implemented in version/cgroupv2 branch. Use this branch if you are using Ubuntu 21.10 or greater for now. Also make sure to upgrade Docker Engine to >=20.10.
Once you have an active account, go to the Mullvad Manage ports and WireGuard keys page and make sure you have at least 1 wireguard key available to allow the Mullvad VPN container to connect. Each account number has a total of 5 wireguard keys.
Make sure ports 8118 and 1080 are available. If you need to bind the proxy servers to different ports, you can create a .env in the same path as the docker-compose.yml file with the following to change the ports:
HTTP_PORT=8118
SOCKS5_PORT=1080If you have ufw or other firewall application enabled, you will need to allow access between containers, something like this for ufw:
sudo ufw insert 1 allow from 172.0.0.0/8This inserts into first position a rule that allows all traffic from 172.0.0.0/8 ip addresses. This example is probably insecure so you should use rules that fits your network.
Once the above is setup, clone the repository:
git clone https://github.com/bernardko/mullvad-proxy.git
cd mullvad-proxyI recommend using a .env file to save you some typing. Copy the .env.example file and make your changes there. The Mullvad account number can also be set in the .env file like the following.
MULLVAD_VERSION=2023.6
ACCOUNT_NUMBER=<Mullvad Account Number>
HTTP_PORT=8118
SOCKS5_PORT=1080With this set, you can just run ./setup.sh to get the containers up and running.
./setup.shThe setup.sh script runs docker-compose up -d to setup Mullvad VPN and proxy docker containers with container networking and then runs several Mullvad VPN CLI commands to setup the app with the account number and connects to the VPN. The containers are configured to auto restart and will be available whenever your machine starts as long as docker is running.
To teardown the containers, use the following:
./down.shTo teardown the containers and delete all images created use the following:
./destroy.shThis script first runs mullvad account get and mullvad tunnel get to output the account and Wireguard key that is used in the Docker container. Finally it will run docker-compose down to teardown the container processes. This allows you to conveniently delete the unused key from the Manage ports and WireGuard keys page (without having to guess which key it is) so that it is not locked up.
All Mullvad CLI commands can be used to control the Mullvad app.
For example, to change VPN location, use the following Mullvad CLI command through docker exec:
docker exec -it mvpn mullvad relay set location us