My Approach to Self-Hosting Git

I have often tried to self-host my Git repositories, but have always fallen short when I tried to find a suitable web interface to show on the front-end.

After a few years, I have finally found a combination of methods that allow me to easily self-host my projects, view them on the web, and access them from anywhere.

Before I dive into the details, I want to state a high-level summary of my self-hosted Git approach:

This method uses the ssh:// (read & write) and git:// (read-only) protocols for push and pull access.
- For the git:// protocol, I create a git-daemon-export-ok file in any repository that I want to be cloneable by anyone.
- The web interface I am using (cgit) allows simple HTTP cloning by default. I do not disable this setting as I want beginners to be able to clone one of my repositories even if they don't know the proper method.
I am not enabling Smart HTTPS for any repositories. Updates to repositories must be pushed via SSH.
Beyond the actual repository management, I am using cgit for the front-end web interface.
- If you use the scan-path=<path> configuration in the cgitrc configuration file to automatically find repositories, you can't exclude a repository from cgit if it's stored within the path that cgit reads. To host private repositories, you'd need to set up another directory that cgit can't read.

Assumptions

For the purposes of this walkthrough, I am assuming you have a URL (git.example.com) or IP address (207.84.26.991) addressed to the server that you will be using to host your git repositories.

Adding a Git User

In order to use the SSH method associated with git, we will need to add a user named git. If you have used the SSH method for other git hosting sites, you are probably used to the following syntax:

git clone [user@]server:project.git

The syntax above is an scp-like syntax for using SSH on the git user on the server to access your repository.

Let's delete any remnants of an old git user, if any, and create the new user account:

sudo deluser --remove-home git
sudo adduser git

Import Your SSH Keys to the Git User

Once the git user is created, you will need to copy your public SSH key on your local development machine to the git user on the server.

If you don't have an SSH key yet, create one with this command:

ssh-keygen

Once you create the key pair, the public should be saved to ~/.ssh/id_rsa.pub.

If your server still has password-based authentication available, you can copy it over to your user's home directory like this:

ssh-copy-id git@server

Otherwise, copy it over to any user that you can access.

scp ~/.ssh/id_rsa.pub your_user@your_server:

Once on the server, you will need to copy the contents into the git user's authorized_keys file:

cat id_rsa.pub > /home/git/.ssh/authorized_keys

(Optional) Disable Password-Based SSH

If you want to lock down your server and ensure that no one can authenticate in via SSH with a password, you will need to edit your SSH configuration.

sudo nano /etc/ssh/sshd_config

Within this file, find the following settings and set them to the values I am showing below:

PermitRootLogin no
PasswordAuthentication no
AuthenticationMethods publickey

You may have other Authentication Methods required in your personal set-up, so the key here is just to ensure that AuthenticationMethods does not allow passwords.

Setting up the Base Directory

Now that we have set up a git user to handle all transport methods, we need to set up the directory that we will be using as our base of all repositories.

In my case, I am using /git as my source folder. To create this folder and assign it to the user we created, execute the following commands:

sudo mkdir /git
sudo chown -R git:git /git

Creating a Test Repository

On your server, switch over to the git user in order to start managing git files.

su git

Once logged-in as the git user, go to your base directory and create a test repository.

cd /git
mkdir test.git && cd test.git
git init --bare

If you want to make this repo viewable/cloneable to the public via the git:// protocol, you need to create a git-daemon-export-ok file inside the repository.

touch git-daemon-export-ok

Change the Login Shell for `git`

To make sure that the git user is only used for git operations and nothing else, you need to change the user's login shell. To do this, simply use the chsh command:

sudo chsh git

The interactive prompt will ask which shell you want the git user to use. You must use the following value:

/usr/bin/git-shell

Once done, no one will be able to SSH to the git user or execute commands other than the standard git commands.

Opening the Firewall

Don't forget to open up ports on the device firewall and network firewall if you want to access these repositories publicly. If you're using default ports, forward ports 22 (ssh) and 9418 (git) from your router to your server's IP address.

If your server also has a firewall, ensure that the firewall allows the same ports that are forwarded from the router. For example, if you use ufw:

sudo ufw allow 22
sudo ufw allow 9418

Non-Standard SSH Ports

If you use a non-standard port for SSH, such as 9876, you will need to create an SSH configuration file on your local development machine in order to connect to your server's git repositories.

To do this, you'll need to define your custom port on your client machine in your ~/.ssh/config file:

nano ~/.ssh/config

Host git.example.com
  # HostName can be a URL or an IP address
  HostName git.example.com
  Port 9876
  User git

Testing SSH

There are two main syntaxes you can use to manage git over SSH:

git clone [user@]server:project.git
git clone ssh://[user@]server/project.git

I prefer the first, which is an scp-like syntax. To test it, try to clone the test repository you set up on the server:

git clone git@git.example.com:/git/test.git

Enabling Read-Only Access

If you want people to be able to clone any repository where you've placed a git-daemon-export-ok file, you will need to start the git daemon.

To do this on a system with systemd, create a service file:

sudo nano /etc/systemd/system/git-daemon.service

Inside the git-daemon.service file, paste the following:

[Unit]
Description=Start Git Daemon

[Service]
ExecStart=/usr/bin/git daemon --reuseaddr --base-path=/git/ /git/

Restart=always
RestartSec=500ms

StandardOutput=syslog
StandardError=syslog
SyslogIdentifier=git-daemon

User=git
Group=git

[Install]
WantedBy=multi-user.target

Once created, enable and start the service:

sudo systemctl enable git-daemon.service
sudo systemctl start git-daemon.service

To clone read-only via the git:// protocol, you can use the following syntax:

git clone git://git.example.com/test.git

Migrating Repositories

At this point, we have a working git server that works with both SSH and read-only access.

For each of the repositories I had hosted a different provider, I executed the following commands in order to place a copy on my server as my new source of truth:

Server:

su git
mkdir /git/<REPOSITORY_NAME>.git && cd /git/<REPOSITORY_NAME>.git
git init --bare

# If you want to make this repo viewable/cloneable to the public
touch git-daemon-export-ok

Client:

git clone git@<PREVIOUS_HOST>:<REPOSITORY_NAME>
git remote set-url origin git@git.EXAMPLE.COM:/git/<REPOSITORY_NAME>.git
git push

Optional Web View: `cgit`

If you want a web viewer for your repositories, you can use various tools, such as gitweb, cgit, or klaus. I chose cgit due to its simple interface and fairly easy set-up (compared to others). Not to mention that the Linux kernel uses cgit.

Docker Compose

Instead of using my previous method of using a docker run command, I've updated this section to use docker-compose instead for an easier installation and simpler management and configuration.

In order to use Docker Compose, you will set up a docker-compose.yml file to automatically connect resources like the repositories, cgitrc, and various files or folders to the cgit container you're creating:

mkdir ~/cgit && cd ~/cgit
nano docker-compose.yml

# docker-compose.yml
version: '3'

services:
  cgit:
    image: invokr/cgit
    volumes:
      - /git:/git
      - ./cgitrc:/etc/cgitrc
      - ./logo.png:/var/www/htdocs/cgit/logo.png
      - ./favicon.png:/var/www/htdocs/cgit/favicon.png
      - ./filters:/var/www/htdocs/cgit/filters
    ports:
      - "8763:80"
    restart: always

Then, just start the container:

sudo docker-compose up -d

Once it's finished installing, you can access the site at <SERVER_IP>:8763 or use a reverse-proxy service to forward cgit to a URL, such as git.example.com. See the next section for more details on reverse proxying a URL to a local port.

Nginx Reverse Proxy

I am using Nginx as my reverse proxy so that the cgit Docker container can use git.example.com as its URL. To do so, I simply created the following configuration file:

sudo nano /etc/nginx/sites-available/git.example.com

server {
        listen 80;
          server_name git.example.com;

        if ($host = git.example.com) {
                return 301 https://$host$request_uri;
          }

          return 404;
}

server {
        server_name git.example.com;
        listen 443 ssl http2;

        location / {
                # The final `/` is important.
                    proxy_pass http://localhost:8763/;
                add_header X-Frame-Options SAMEORIGIN;
                add_header X-XSS-Protection "1; mode=block";
                proxy_redirect off;
                proxy_buffering off;
                proxy_set_header Host $host;
                proxy_set_header X-Real-IP $remote_addr;
                proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
                proxy_set_header X-Forwarded-Proto $scheme;
                proxy_set_header X-Forwarded-Port $server_port;
        }

        # INCLUDE ANY SSL CERTS HERE
        include /etc/letsencrypt/options-ssl-nginx.conf;
        ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;
}

Once created, symlink it and restart the web server.

sudo ln -s /etc/nginx/sites-available/git.example.com /etc/nginx/sites-enabled/
sudo systemctl restart nginx.service

As we can see below, my site at git.example.com is available and running:

Settings Up Git Details

Once you have cgit running, you can add some small details, such as repository owners and descriptions by editing the following files within each repository.

Alternatively, you can use the cgitrc file to edit these details if you only care to edit them for the purpose of seeing them on your website.

The description file within the repository on your server will display the description online.

cd /git/example.git
nano description

You can add a [gitweb] block to the config file in order to display the owner of the repository.

cd /git/example.git
nano config

[gitweb]
    owner = "YourName"

Note that you can ignore the configuration within each repository and simply set up this information in the cgitrc file, if you want to do it that way.

Editing `cgit`

In order to edit certain items within cgit, you need to edit the cgitrc file.

nano ~/cgit/cgitrc

Below is an example configuration for cgitrc. You can find all the configuration options within the [configuration manual] (https://git.zx2c4.com/cgit/plain/cgitrc.5.txt).

css=/cgit.css
logo=/logo.png
favicon=/favicon.png
robots=noindex, nofollow

enable-index-links=1
enable-commit-graph=1
enable-blame=1
enable-log-filecount=1
enable-log-linecount=1
enable-git-config=1

clone-url=git://git.example.com/$CGIT_REPO_URL ssh://git@git.example.com:/git/$CGIT_REPO_URL

root-title=My Git Website
root-desc=My personal git repositories.

# Allow download of tar.gz, tar.bz2 and zip-files
snapshots=tar.gz tar.bz2 zip

##
## List of common mimetypes
##
mimetype.gif=image/gif
mimetype.html=text/html
mimetype.jpg=image/jpeg
mimetype.jpeg=image/jpeg
mimetype.pdf=application/pdf
mimetype.png=image/png
mimetype.svg=image/svg+xml

# Highlight source code
# source-filter=/var/www/htdocs/cgit/filters/syntax-highlighting.sh
source-filter=/var/www/htdocs/cgit/filters/syntax-highlighting.py

# Format markdown, restructuredtext, manpages, text files, and html files
# through the right converters
about-filter=/var/www/htdocs/cgit/filters/about-formatting.sh

##
## Search for these files in the root of the default branch of repositories
## for coming up with the about page:
##
readme=:README.md
readme=:readme.md
readme=:README.mkd
readme=:readme.mkd
readme=:README.rst
readme=:readme.rst
readme=:README.html
readme=:readme.html
readme=:README.htm
readme=:readme.htm
readme=:README.txt
readme=:readme.txt
readme=:README
readme=:readme

# Repositories

# Uncomment the following line to scan a path instead of adding repositories manually
# scan-path=/git

## Test Section
section=git/test-section

repo.url=test.git
repo.path=/git/test.git
repo.readme=:README.md
repo.owner=John Doe
repo.desc=An example repository!

Final Fixes: Syntax Highlighting & README Rendering

After completing my initial install and playing around with it for a few days, I noticed two issues:

Syntax highlighting did not work when viewing the source code within a file.
The about tab within a repository was not rendered to HTML.

The following process fixes these issues. To start, let's go to the cgit directory where we were editing our configuration file earlier.

cd ~/cgit

In here, create two folders that will hold our syntax files:

mkdir filters && mkdir filters/html-converters && cd filters

Next, download the default filters:

curl https://git.zx2c4.com/cgit/plain/filters/about-formatting.sh > about-formatting.sh
chmod 755 about-formatting.sh
curl https://git.zx2c4.com/cgit/plain/filters/syntax-highlighting.py > syntax-highlighting.py
chmod 755 syntax-highlighting.py

Finally, download the HTML conversion files you need. The example below downloads the Markdown converter:

cd html-converters
curl https://git.zx2c4.com/cgit/plain/filters/html-converters/md2html > md2html
chmod 755 md2html

If you need other filters or html-converters found within the cgit project files, repeat the curl and chmod process above for whichever files you need.

However, formatting will not work quite yet since the Docker cgit container we're using doesn't have the formatting package installed. You can install this easily by install Python 3+ and the pygments package:

# Enter the container's command line
sudo docker exec -it cgit bash

# Install the necessary packages and then exit
yum update -y &&                      \
yum upgrade -y &&                     \
yum install python3 python3-pip -y && \
pip3 install markdown pygments &&     \
exit

You will need to enter the cgit docker container and re-run these yum commands every time you kill and restart the container!

If not done already, we need to add the following variables to our cgitrc file in order for cgit to know where our filtering files are:

# Highlight source code with python pygments-based highlighter
source-filter=/var/www/htdocs/cgit/filters/syntax-highlighting.py

# Format markdown, restructuredtext, manpages, text files, and html files
# through the right converters
about-filter=/var/www/htdocs/cgit/filters/about-formatting.sh

Now you should see that syntax highlighting and README rendering to the about tab is fixed.

Theming

I won't go into much detail in this section, but you can fully theme your installation of cgit since you have access to the cgit.css file in your web root. This is another file you can add as a volume to the docker-compose.yml file if you want to edit this without entering the container's command line.

:warning: Remember to Back Up Your Data!

The last thing to note is that running services on your own equipment means that you're assuming a level of risk that exists regarding data loss, catastrophes, etc. In order to reduce the impact of any such occurrence, I suggest backing up your data regularly.

Backups can be automated via cron, by hooking your base directory up to a cloud provider, or even setting up hooks to push all repository info to git mirrors on other git hosts. Whatever the method, make sure that your data doesn't vanish in the event that your drives or servers fail.

Self-Hosting a Personal Git Server