ansible-django-stack

Ansible Playbook designed for environments running a Django app. It can install and configure these applications that are commonly used in production Django deployments:

• Nginx
• Gunicorn
• PostgreSQL
• Supervisor
• Virtualenv
• Memcached
• Celery
• RabbitMQ

Default settings are stored in roles/role_name/defaults/main.yml. Environment-specific settings are in the env_vars directory.

A certbot role is also included for automatically generating and renewing trusted SSL certificates with Let's Encrypt.

Tested with OS: Ubuntu 18.04 LTS (64-bit), Ubuntu 16.04 LTS (64-bit).

Tested with Cloud Providers: Digital Ocean, AWS, Rackspace

Getting Started

A quick way to get started is with Vagrant.

Requirements

• Ansible
• Vagrant
• VirtualBox or Docker

It's recommended to use the version of Ansible specified in requirements.txt, although any version greater than Ansible 2.7 will work with this repository. When choosing an Ansible version, consider:

• Ansible only issues security fixes for the last three major releases.
• The included version of molecule has requirements on the Ansible version (currently, Molecule requires Ansible 2.5 or later and the 2.23 release will require Ansible 2.7 or greater)

Ansible has been configured to use Python 3 inside the remote machine when provisioning it. In Ubuntu 16.04 LTS, compatible Ansible versions are not in the main package repositories, but can be installed from the Ansible PPA by running these commands:

sudo add-apt-repository ppa:ansible/ansible
sudo apt-get update

Configuring your application

The main settings to change are in the env_vars/base.yml file, where you can configure the location of your Git project, the project name, and the application name which will be used throughout the Ansible configuration. I set some default values based on my open-source app, YouTube Audio Downloader.

Note that the default values in the playbooks assume that your project structure looks something like this:

myproject
├── manage.py
├── myapp
│   ├── apps
│   │   └── __init__.py
│   ├── __init__.py
│   ├── settings
│   │   ├── base.py
│   │   ├── __init__.py
│   │   ├── local.py
│   │   └── production.py
│   ├── templates
│   │   ├── 403.html
│   │   ├── 404.html
│   │   ├── 500.html
│   │   └── base.html
│   ├── urls.py
│   └── wsgi.py
├── README.md
└── requirements.txt

The main things to note are the locations of the manage.py and wsgi.py files. If your project's structure is a little different, you may need to change the values in these 2 files:

• roles/web/tasks/setup_django_app.yml
• roles/web/templates/gunicorn_start.j2

Also, if your app needs additional system packages installed, you can add them in roles/web/tasks/install_additional_packages.yml.

Creating the machine

Type this command from the project root directory:

vagrant up

(To use Docker instead of VirtualBox, add the flag --provider=docker to the command above. Note that extra configuration may be required first on your host for Docker to run systemd in a container.)

Wait a few minutes for the magic to happen. Access the app by going to this URL: https://my-cool-app.local

Yup, exactly, you just provisioned a completely new server and deployed an entire Django stack in 5 minutes with two words :).

Additional vagrant commands

SSH to the box

vagrant ssh

Re-provision the box to apply the changes you made to the Ansible configuration

vagrant provision

Reboot the box

vagrant reload

Shutdown the box

vagrant halt

Pulling from a private git repository using SSH agent forwarding

If your code is in a private repository you must use an SSH connection along with a key so ansible can checkout the code. HTTPS connections and SSH connections with a username and password do not work because ansible cannot deal with interactive logins.

Using SSH agent forwarding we can get the authentication request from the repository sent to the machine where the playbook is run. That keeps the private key to the repository as safe as possible. To set this up you need to:

• Add a public key to the server hosting your repo.
• Make sure ssh-agent is running on your local machine
• Add the public key to ssh-agent

Connecting to GitHub with SSH has all the information on key generation, adding keys to the server, setting up ssh-agent and troubleshooting any problems.

Your server SSH configuration should work out-of-the-box. The "Server-Side Configuration Options" section in SSH Essentials: Working with SSH Servers, Clients, and Keys has good advice on locking down who can access the server over SSH connections.

Getting the playbook to use agent forwarding

The first thing you need to do is set the ssh_agent_forwarding flag in env_vars/base.yml to true:

ssh_forward_agent: true

This flag is used when configuring sudoers so that any user you become on the remote server will also use the same socket connection when requesting to unlock keys.

To enable SSH agent forwarding on the Vagrant box, change the following flag in VagrantFile and set it to true:

config.ssh.forward_agent = true

When running a playbook to provision a server, you enable SSH agent forwarding using the --ssh-extra-args option on the command line:

ansible-playbook --ssh-extra-args=-A -i production site.yml

This is a little bit clunky but it does not restrict you from setting other SSH options if you need to.

Security

NOTE: Do not run the Security role without understanding what it does. Improper configuration could lock you out of your machine.

Security role tasks

The security module performs several basic server hardening tasks. Inspired by this blog post:

• Updates apt
• Performs aptitude safe-upgrade
• Adds a user specified by the server_user variable, found in roles/base/defaults/main.yml
• Adds authorized key for the new user
• Installs sudo and adds the new user to sudoers with the password specified by the server_user_password variable found in roles/security/defaults/main.yml
• Installs and configures various security packages:
  • Unattended upgrades
  • Uncomplicated Firewall
  • Fail2ban (NOTE: Fail2ban is disabled by default as it is the most likely to lock you out of your server. Handle with care!)
• Restricts connection to the server to SSH and http(s) ports
• Limits su access to the sudo group
• Disallows password authentication (be careful!)
• Disallows root SSH access (you will only SSH to your machine as your new user and use a password for sudo access)
• Restricts SSH access to the new user specified by the server_user variable
• Deletes the root password

Security role configuration

• Change the server_user from root to something else in roles/base/defaults/main.yml
• Change the sudo password in roles/security/defaults/main.yml
• Change variables in ./roles/security/vars/ per your desired configuration

Running the Security role

• The security role can be run by running security.yml via:

ansible-playbook -i development security.yml

Running the Ansible Playbook to provision servers

NOTE: to enable the Security module you can use the steps above prior to following the steps below.

Create an inventory file for the environment, for example:

# development

[all:vars]
env=dev

[webservers]
webserver1.example.com
webserver2.example.com

[dbservers]
dbserver1.example.com

Next, create a playbook for the server type. See webservers.yml for an example.

Run the playbook:

ansible-playbook -i development webservers.yml [-K]

You can also provision an entire site by combining multiple playbooks. For example, I created a playbook called site.yml that includes both the webservers.yml and dbservers.yml playbook.

A few notes here:

• The dbservers.yml playbook will only provision servers in the [dbservers] section of the inventory file.
• The webservers.yml playbook will only provision servers in the [webservers] section of the inventory file.
• An inventory var called env is also set which applies to all hosts in the inventory. This is used in the playbook to determine which env_var file to use.
• The -K flag is for adding the sudo password you created for a new sudoer in the Security role (if applicable)

You can then provision the entire site with this command:

ansible-playbook -i development site.yml [-K]

If you're testing with vagrant, you can use this command:

ansible-playbook -i vagrant_ansible_inventory_default --private-key=~/.vagrant.d/insecure_private_key vagrant.yml [-K]

Using Ansible for Django Deployments

When doing deployments, you can simply use the --tags option to only run those tasks with these tags.

For example, you can add the tag deploy to certain tasks that you want to execute as part of your deployment process and then run this command:

ansible-playbook -i stage webservers.yml --tags="deploy"

This repo already has deploy tags specified for tasks that are likely needed to run during deployment in most Django environments.

Advanced Options

Changing the Ubuntu release

The Vagrantfile uses the Ubuntu 18.04 LTS Vagrant box for a 64-bit PC that is published by Canonical in HashiCorp Atlas. To use Ubuntu 16.04 LTS instead, change the config.vm.box setting to ubuntu/xenial64.

Changing the Python version used by your application

Python 3 is used by default in the virtualenv. To use Python 2 instead, just override the virtualenv_python_version variable and set it to python.

It is possible to install other versions of Python from an unofficial PPA by Felix Krull (see disclaimer). To use this PPA, override the enable_deadsnakes_ppa variable and set it to yes. Then the virtualenv_python_version variable can be set to the name of a Python package from this PPA, such as python3.6.

Changing the Python version used by Ansible

To use Python 2 as the interpreter for Ansible, override the ansible_python_interpreter variable and set it to /usr/bin/python. This allows a machine without Python 3 to be provisioned.

Creating a swap file

By default, the playbook won't create a swap file. To create/enable swap, simply change the values in roles/base/defaults/main.yml.

You can also override these values in the main playbook, for example:

---

...

  roles:
    - { role: base, create_swap_file: true, swap_file_size_kb: 1024 }
    - db
    - rabbitmq
    - web
    - celery

This will create and mount a 1GB swap. Note that block size is 1024, so the size of the swap file will be 1024 x swap_file_size_kb.

Automatically generating and renewing Let's Encrypt SSL certificates with the certbot client

A certbot role has been added to automatically install the certbot client and generate a Let's Encrypt SSL certificate.

Requirements:

• A DNS "A" or "CNAME" record must exist for the host to issue the certificate to.
• The --standalone option is being used, so port 80 or 443 must not be in use (the playbook will automatically check if Nginx is installed and will stop and start the service automatically).

In roles/nginx/defaults.main.yml, you're going to want to override the nginx_use_letsencrypt variable and set it to yes/true to reference the Let's Encrypt certificate and key in the Nginx template.

In roles/certbot/defaults/main.yml, you may want to override the certbot_admin_email variable.

A cron job to automatically renew the certificate will run daily. Note that if a certificate is due for renewal (expiring in less than 30 days), Nginx will be stopped before the certificate can be renewed and then started again once renewal is finished. Otherwise, nothing will happen so it's safe to leave it running daily.

Useful Links

• Ansible - Getting Started
• Ansible - Best Practices
• Setting up Django with Nginx, Gunicorn, virtualenv, supervisor and PostgreSQL
• How to deploy encrypted copies of your SSL keys and other files with Ansible and OpenSSL
• Using SSH agent forwarding - GitHub developer guide

Contributing

Contributions are welcome! Please make sure any PR passes the TravisCI test suite.

Running the test suite locally:

The test suite uses a Docker container - make sure Docker is installed and configured before running the following commands:

pip install -r requirements-dev.txt
molecule test