Django SSH Deployer

This package provides a Django management command to deploy your site to various instances (develop, stage, production) over SSH via Paramiko.

Pre-Requisites

With great power comes great responsibility! Target servers (DEPLOYER_INSTANCES['instance']['servers']) must each have git and Python 3.3+ installed, and support Linux-style OS commands. Target servers must have a user (DEPLOYER_INSTANCES['instance']['server_user']) with keys set up from the control machine where you run the Django command from. This typically means installing the control machine account's public key into the target server's user account's AUTHORIZED_KEYS.

Installation and Required Django Settings

Install via pip into your development environment:

pip install django-ssh-deployer

Then add django_ssh_deployer to your INSTALLED_APPS. Next, we need to configure your instances in Django's settings; these can live in your development or local settings, as they won't be required by production.

DEPLOYER_INSTANCES = {
    "dev": {
        "name": "your-project",
        "repository": "git@github.com:youruser/your-project.git",
        "branch": "dev",
        "settings": "config.settings.dev",
        "pip_command": "-e .[dev]",
        "code_path": "/var/django/sites",
        "venv_python_path": "/usr/bin/python3",
        "upgrade_pip": True,
        "servers": ["devserver.example.com"],
        "server_user": "deploy_user",
        "save_deploys": 3,
        "selinux": False,
        "collectstatic": False,
        "migrate": False,
    },
    "prod": {
        "name": "your-project",
        "repository": "git@github.com:youruser/your-project.git",
        "branch": "prod",
        "settings": "config.settings.prod",
        "requirements": "requirements/prod.txt",
        "code_path": "/var/django/sites",
        "venv_python_path": "/usr/bin/python3",
        "upgrade_pip": True,
        "servers": ["prodserver-1.example.com", "prodserver-2.example.com"],
        "server_user": "deploy_user",
        "save_deploys": 3,
        "selinux": True,
        "additional_commands": [
            "chmod -R a+rX /var/django/sites/your-project-prod",
            "curl -kLs -o /dev/null --max-time 5 --resolve 'your-domain.com:443:127.0.0.1' https://your-domain.com/",
        ],
    },
}

• name: A name for your project.
• repository: The repository for your Django project, which will be cloned on each target server.
• branch: The branch to check out for the instance.
• settings: A full path to the Django settings for the instnace.
• requirements or pip_command: Either a relative path to a requirements file, or a pip path to be pip install'd for the instance.
• code_path: The root path for your code repository to be checked out to on the target servers.
• venv_python_path: The full path to the version of Python for the venv to use on the target servers.
• servers: A list of servers to deploy the Django project to.
• server_user: The user on the target servers which has been set up with keys from the control machine.
• (optional) upgrade_pip: If set to False, will not upgrade pip to the latest version.
• (optional) collectstatic: If set to False, will not collect static files.
• (optional) migrate: If set to False, will not run migrations.
• (optional) save_deploys: If a positive integer, will only keep the most recent number of deployments. By default, will keep all.
• (optional) selinux: If set to True, the deployer will run chcon command to set the necessary security context on files for SELinux. It will set all files in the codepath to httpd_sys_content_t, and any *.so files in the venv to httpd_sys_script_exec_t.
• (optional) additional_commands: A list of commands to run after the deployment is complete.

Optionally, you can customize the directory created by the git clone in your Django settings:

DEPLOYER_CLONE_DIR_FORMAT = "{name}-{instance}"

The following keywords will be replaced in the checkout directory format: instance, name, branch, and server_user. The default is "{name}-{instance}", which in the example above, would be your-project-develop and your-project-production.

Running the Command

python manage.py deploy --instance=develop

• --instance: Required. The name of the instance to deploy in DEPLOYER_INSTANCES. In the example above, either develop or production.
• --quiet: Less verbose output. Does not display the output of the commands being run to the terminal.
• --no-confirm: Publishes without a confirmation step. Be careful!
• --stamp: By default, Django SSH Deployer will append a datetime stamp to the git clone. This overrides the datetime default.

What It Does

The deploy command will SSH to each server in servers as the server_user, and perform the following functions in two passes.

First, it will connect to each server and prepare the new deployment:

• clone the repository from git with a stamp
• create a venv with a stamp
• run the collectstatic command

After the deployment has been prepared on all servers without error, it will proceed to the final deployment steps:

• run the migrate command on the first server only
• create or update the symlink to point to the completed deploy on each server

Known Limitations and Issues

• Windows servers are not supported, however, you can use Windows as your control machine.
• Your repository's host must be in your target server's known hosts list, as git checkouts over SSH require an initial fingerprint.
• This is not meant to be a replacement for a fully featured continous integration product, like Jenkins.

Release Notes

Release notes are available on GitHub.

Contributors

• Timothy Allen (https://github.com/FlipperPA)