Awesome
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 group_vars
directory.
A certbot
role is also included for automatically generating and renewing
trusted SSL certificates with Let's Encrypt.
Tested with OS: Ubuntu 22.04 LTS (64-bit), Ubuntu 20.04 LTS (64-bit).
Tested with Cloud Providers: Digital Ocean, AWS, Rackspace
Getting Started
A quick way to get started is with Vagrant.
Requirements
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 group_vars/[environment_name]/vars.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.
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 inroles/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 inroles/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
fromroot
to something else inroles/base/defaults/main.yml
- Change the sudo password in
group_vars/[environment_name]/vars.yml
- Change variables in
./roles/security/vars/
per your desired configuration by overriding them ingroup_vars/[environment_name]/vars.yml
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
[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. - 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 22.04 LTS Vagrant box for a
64-bit PC that is published by Canonical in HashiCorp Atlas. To use Ubuntu
20.04 LTS instead, change the config.vm.box
setting to ubuntu/focal64
.
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.
Maintenance mode
The playbook contains a maintenance page option.
roles/web/templates/maintenance_off.html
is the provided maintenance template. To activate the maintenance mode, you can rename
the template to
maintenance_on.html
, in order for nginx
to serve it. This can be done manually. Alternately,
you can include in the playbook a step activating the maintenance page (using the renaming
process) while the site requires downtime. Then switch back to running mode in
the playbook when operations requiring downtime are completed.
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
- Jeff Geerling has some incredible resources for learning Ansible and his scripts and blog post have been instrumental in building this repo:
- His YouTube Channel has a video series on learning Ansible.
- The book Ansible for DevOps is an excellent resource for learning Ansible best practices.
Contributing
Contributions are welcome! Please make sure any PR passes the 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