
1-click deployment of OpenVPN with DNS ad blocking sinkhole. Deploys to your favorite VPS machine. Created with Vue.js, Semantic UI and Django. And with love, of course.
Feedback and pull requests are welcome.
OpenVPN is a registered trademark of OpenVPN Inc. © 2002-2019 OpenVPN Inc.
This project is not endorsed by, sponsored or affiliated with OpenVPN Inc.
Managing OpenVPN with PKI authentication is hard. Managing anything beyond hello-world using easy-rsa
package
is a major issue - I could never maintain a config for more than a day. Other solutions are too "enterprise"
for a personal installation or were designed for a tin-foil hat, crypto maniacs hiding from NSA/GCHQ.
This app provides easy management console to keep OpenVPN configuration files in one place, provided in self-contained, easily deployable, clickable package. It's not designed for security - it's meant just to be good enough.
And that works for me better than "no VPN at all".
Features:
- 1-click deployment of OpenVPN server to your favorite VPS provider
- DNS cache and ad blocking for VPN connected clients
- OpenVPN clients management
- generation of self-contained ovpn profiles for servers and clients
- profiles can be sent by e-mail to owner or downloaded as files
- tested on Ubuntu 18.04 and OpenBSD 6.4 (Vultr VPS)
That's all folks.
Note
This is a work-in-progress app, hacked together during x-mas break to solve a specific need of mine. Feel free to submit PRs with improvements.
Why?
To quickly deploy VPN server when I need it. I can spin VPS and deploy my own VPN any time, tear it down when not used and not paying a monthly fee for all my devices.
I travel a lot and I need to have on-demand VPN when browsing stuff in hotels, airports, etc.
Does it hide my ass? Can I haz torrentz?
No. Do not use it to do any stupid things.
Is the app secure?
Since the app manages OpenVPN server deployment, it must have root access to the VPN machine. There is no separate deployment agent (yet), as it would over-complicate things. It is not wise to keep it facing the open internet, I guess, so please don't do it.
So how to host it?
Preferably on your internal network. Keep the server bound to localhost and connect to it via SSH tunnel. This way you don't need to configure SSL certificate and a lot of security headaches go away.
I use it installed on my private laptop, the same way I use CUPS (printer stuff, aka localhost-colon-six-three-one).
How to change server address after it is created?
Use Django Admin panel to modify host field and re-deploy. All client configs must be re-deployed too. You can try playing with DynDNS to work around it.
Why Ansible? It's slow and weights 30MB.
- It does the job like a champ lifting tons of system complexity
- Zero-effort deployment (no master nodes, etc)
- Very easy to extend
- I'd like to have more complex setup in the future and bash won't cut it
Why not language X?
I believe Python is the optimal solution considering platform maturity, libraries quality and skills proliferation. There is not much choice for the frontend.
The project is split into backend, frontend. and ansible scripts.
The backend is written in Django and Django REST Framework. The frontend is a Vue.js SPA application served by Django. That division makes the build slightly more complicated, but provided Makefiles make it a breeze. It should just work.
Ansible is a set of scripts to deploy OpenVPN automatically either on localhost or remote machine.
Scripts located in bin are created either to automate and facilitate various tasks or provide a glue. All scripts have internal documentation (or should have).
- Working Node.js installation (tested with 9.3.0)
- Python 3 with virtualenv
- GNU Make (or compatible)
- Ansible (tested with 2.5.0, but no fancy functionality is used)
- OpenVPN in ${PATH}
- OpenSSL in ${PATH}
- OpenSSH in ${PATH}
- Internet connection (no off-line build possible)
After cloning the repository, you can easily deploy the app for development:
$ git clone https://github.com/ezaquarii/vpn-at-home $ cd vpnathome $ make devel ... backend is bootstrapped ... ... frontend is bootstrapped ... $ make runserver
Open http://localhost:8001/
and you should be able to log-in.
Works out of the box, no prerequisites besides docker needed
- Clone the repo and go to the vpnathome directory.
- Run
docker-compose up
. Docker will install and start a development server for you. - Now you can go to
http://localhost:8001/
and you will be able to login. - Make some changes. The container will automatically pick them up via a volume.
- After you saved the changes, you can refresh
http://localhost:8001/
and will see them immediately.
Docker container can be created with make
:
$ make docker
Created image will be tagged with name vpnathome
.
You can launch a container with a helper script:
$ ./bin/docker_run.sh bootstrap $ ./bin/docker_run.sh run
...or roll out your own fancy scripts for this. Data will be stored in a volume data
.
Package deployment is supported on Ubuntu. Debian should be supported, but I didn't test it there.
$ make deb $ make install_deb
Open http://localhost:8000
and follow on-boarding tutorial.
Note
Building deb
package calls make distclean
, which will zap your development
configuration. Build outside devel environment if you want to preserve your config.
The package needs virtually zero configuration:
deb
is self-deployable, as it contains entire virtualenv- installs into
/srv/vpnathome
(referred to as${ROOT}
) systemd
service scriptvpnathome.service
is installed and starts by defaultdaphne
runs onhttp://127.0.0.1:8000
- bound to localhost only- Contains bootstrapping script to automate app configuration (
${ROOT}/bin/bootstrap.sh
)
Building a package will call sudo
and ask you for a password. Root privileges are required
during Python virtual environment installation step, as we must sudo mount -o bind ...
and
sudo umount ...
virtualenv destination directory. Why? Unfortunately, Python 3 virtualenv
relocation is not reliable (and discouraged), so I decided to hack a bit during the build process
and bootstrap directly into destination directory before packaging. Refer to Makefile
install
target for details.
Note
If make deb
fails for whatever reason, make sure /srv/vpnathome
is left unmounted.
Once the app is up and running, you can log in as admin (using credentials set during bootstrapping phase) and create your server.
After a server is configured, you can deploy it using provided Ansible scripts. Beware that Ansible will modify the target system!
- required packages will be installed
- firewall rules will be altered
- IPv4 forwarding will be enabled
It is advised to deploy the server on a remote machine, but you can do it on localhost too. I personally test it on Vultr VPS.
$ ./bin/deploy_vpn.sh --help Usage: ./bin/deploy_vpn.sh [--help|--local|--host HOST] --help - usage --local - deploy OpenVPN server on the current machine (localhost) --host - deploy selected OpenVPN server only
If you deploy to a HOST
, it must be one of the defined VPN servers.
If make devel
was run, the app is up and running in development mode with default development
configuration:
- Admin login is admin@locahost
- Admin password is admin1234
- Database is located in
${PROJECT_ROOT}/data/db/db.sqlite3
- Settings have
development
flag set to truetrue
To set new superuser, use ./bin/manage set_admin <email> <pass>
command.
Configuration is loaded from settings.json
located in deployment directory. The settings file is generated
during a bootstrap stage, so there is no need to generate it manually. However, should you need to generate the
script during development, you can do it with a supplied Django management command:
$ ${ROOT}/bin/manage configure [--accept] [--devel] [--help] [--force]
Once the file is generated (ie. after bootstrap), you must review and accept it by flipping the configured
flag inside.
Note
settings.json
is excluded from Git repository, so you can safely put your real e-mail credentials there
during development.
You can also access Django Admin app, which is left enabled.
OpenVPN configuration is generated from templates in vpnathome.apps.openvpn.templates
. If the default
configuration doesn't suit your needs, you can alter templates directly there.
There is no frontend config editor, although I was thinking about it.
VPN config files can be send to e-mail account of a user that created a config or downloaded. Once downloaded, the config file (OVPN) can be used directly with OpenVPN client.
If server was deployed with DNS cache enabled, DNS is forwarded to connecting client. Depending on your network this might be slower or faster than popular DNS servers or DNS of your ISP.
To verify if your queries are forwarded to VPN DNS:
ping gateway.vpnathome PING gateway.vpnathome (172.30.0.1) 56(84) bytes of data. 64 bytes from _gateway (172.30.0.1): icmp_seq=1 ttl=255 time=46.5 ms 64 bytes from _gateway (172.30.0.1): icmp_seq=2 ttl=255 time=48.7 ms
where 172.30.0.1
will be your choosen VPN gateway IP. Check systemd-resolve --status
if DNS servers are
properly pushed.
Want to jump in? Fantastic.
I made it as easy to start development as possible. Top-level project directory contains 2 subprojects:
backend
and frontend
.
Top-level Makefile
delegates targets to sub-projects and is provided for convenience. Once make devel
is
done, you can work inside individual subproject with your favourite IDE.
I personally use JetBrains WebStorm and PyCharm, but you can use whatever you want. IDE files are not even in the repo.
This is the Django app. Mostly REST API + single frontend serving view.
App modules have brief documentation inside __init__.py
. Docs are kept up-to-date, as I strongly
believe in code documentation.
Provided Makefile
's default target displays help:
$ make Welcome to VPN@Home make system Available targets: * devel - boostrap project for development (your first choice) * virtualenv - install virtual environment and all dependencies * runserver - start development server * test - run full test suite
In development mode, frontend files are stored outside of this project, in frontend
subproject. Django app
will pick static and templates from frontend build directory.
When development mode is off, frontend resources are taken from vpnathome.apps.frontend
app.
Django Debug Toolbar is provided by default, should you need to check which templates are picked up.
Frontend sub-project contains Vue.js SPA served by Django. By default Django app will serve stable, production version of the frontend app directly.
Provided Makefile
's default target displays help:
$ cd frontend $ make Welcome to VPN@Home make system - frontend sub-project You need running node.js and npm. Available targets: * build-prod - build production build; backend project is NOT updated * build-devel - watch and make development build on change; output is written to './dist' * install - install packages from package.json * distclean - clean project, delete all data (start from 'git clone' state)
To start development of frontend code, you must first switch backend into development mode, by modifying data/settings.json
:
{ ... "configured": true, "development": true, "debug_toolbar_enabled": true, ...
Don't forget to restart the app. Once development mode is enabled, Django will load frontend from frontend/dist
instead of vpnathome.apps.frontend
. You can verify this by inspecting site title - it should say
VPN@Home <version> - development. You can also use Django Debug Toolbar to troubleshoot the configuration.
Django injects some initial state via <script>...</script>
tag. See index.html
and vpnathome.apps.frontent.views
for
details.
GNU GPL v3.
I left this as the last point, hoping not to scare anybody.
- frontend has 0% test coverage :o)
- security is not a major concern for this app, I'm not running a CA company
- no real user management - I rely on Django Admin panel for it
- not tested on Windows, as I don't touch it even with a 10-foot stick, in rubber gloves - patches are welcome, however
- no cert revocation (yet)