Deploy FastAPI Application on Ubuntu with Nginx, Gunicorn and Uvicorn

FastAPI is a promising new Python framework that supports concurrency and type system out of the box. It has many cool features that I like and it's fast. In this post, I will briefly go over the process of deploying a simple FastAPI application on Ubuntu running on an EC2 instance. This post will assume that you know and already configured quite a few things and will only focus on deploying the actual application.

Pre-requisites

1. First of all, you will need a running EC2 instance with Ubuntu 18.04 or later. This will also work for VPS servers like DigitalOcean or Linode.
2. If you are using EC2 instance, make sure you configured the VPC and security groups properly and your port 80 is open for inbound and outbound traffic.
3. You have to have Nginx installed and running.
4. You will need Python 3.7 or above installed and any Python virtual environment tool like virtualenv, pipenv, conda, etc. installed.
5. You need to have your FastAPI application on Github/Bitbucket/Gitlab for easy deployment.

After you have ensured that all of the above requirements are met you can move on to the next step. The actual deployment.

Prepare the Application

Now, SSH into the server, create and navigate to the directory that you want your application to be stored into. Say it's /var/www/myapp.

$ mkdir /var/www/myapp 
$ cd /var/www/myapp

We have to make sure that our user has proper read-write access to this directory. In this post, I will use virtualenv to manage a Python virtual environment. You can use whichever you like or prefer. Inside /var/www/myapp we have to do the following one by one:

$ virtualenv -p python3.8 venv
$ mkdir src
$ . venv/bin/activate
(venv) $ cd src

Once inside the src directory, we have to pull our application source code there. Our application source code should contain a requirements.txt file with all the dependencies unless we are using a PIPFILE.

(venv) $ git init
(venv) $ git remote add origin <your-repo-url>
(venv) $ git pull origin <your-branch-name>
(venv) $ pip install -r requirements.txt
(venv) $ pip install gunicorn uvicorn

Configure Nginx

Now our application is ready to be run and tested. To be able to serve the application over HTTP we have to make an Nginx config for our application.

(venv) $ sudo vim /etc/nginx/sites-available/myapp

Put the followings on that file:

server{
       server_name <your-site-name>;
       location / {
           include proxy_params;
           proxy_pass http://127.0.0.1:8000;
       }
}

Now we save the file and exit. Then we make a symbolic link to this config file in the /etc/nginx/sites-enabled directory.

(venv) $ sudo ln -s /etc/nginx/sites-available/myapp /etc/nginx/sites-enabled/

Then we restart the Nginx service.

(venv) $ sudo systemctl restart nginx.service

Now we can start our uvicorn server to check if our application is working or not.

(venv) $ gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app

In the place of main:app we can use whatever is correct for our application. Now that our application is running and the proxy server is configured properly we should be able to visit the URL and see our application from a browser.

Configure ASGI Server

Now that our application is deployed and configured properly one last thing to do is to create a service for the Gunicorn server so that it is always running and it automatically starts when the server is rebooted. We will user systemd to create the service.

(venv) $ deactivate
$ sudo vim /etc/systemd/system/myapp.service

In this new file we have to put the following:

[Unit]
Description=Gunicorn instance to serve MyApp
After=network.target

[Service]
User=<username>
Group=www-data
WorkingDirectory=/var/www/myapp/src
Environment="PATH=/var/www/myapp/venv/bin"
ExecStart=/var/www/myapp/venv/bin/gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app

[Install]
WantedBy=multi-user.target

Modify the locations as of your own setup and then save and exit the file.

$ sudo systemctl start myapp.service

This will start our new service and our ASGI server will be running in the background.