In this post we are going to see how we can efficiently deploy a web application powered by Flask (a Python framework) to production. We won't code a complex application, actually we will just stick to the Flask Hello World example.
Basic Flask application
If you’ve never heard of Flask before, I recommend you to visit its website and read about it.
To create a Hello World application in Flask, just write the following code in a file called application.py.
1
2
3
4
5
6
7
from flask import Flask
app = Flask(__name__)
@app.route('/')
def index():
return "Hello, World!"
And then run the application with this code in a main.py file.
from application import app as application
if __name__ == '__main__':
application.run()
Now run python main.py and your app should be visible in your browser at localhost:5000.
WSGI server
So far we have a working application, but it won’t be enough to serve a production environment. Python applications can’t be directly deployed in a webserver, because python is not a web compatible language like Javascript or PHP for instance. Hence, we need an additional layer: a WSGI server.
There are various WSGI servers on the market, Gunicorn is a popular choice but my favorite is uWSGI.
Getting started
You can install uWSGI with pip: pip install uwsgi. Once you have it installed, you can run our previous python application within a WSGI webserver by simply doing uwsgi main.py. You can pass many arguments to the uwsgi command, but it is often prefered to have them all in a single configuration file. uWSGI can just do that, let’s create a uwsgi.ini configuration file that looks like this:
[uwsgi]
wsgi-file = main.py
master = true
processes = 2
http = 0.0.0.0:9000
vacuum = true
die-on-term = true
Now we can run the app with uwsgi wsgi.ini. With such configuration, our Flask application will be hosted on 2 processes. This is an arbiratry number we set, but you can increase it either manually or automatically if you need to handle more trafic. The server will respond on 0.0.0.0:9000, which means that every client on the local network will be able to access the app on port 9000 using the HTTP protocol. Check that you can still see your app at localhost:9000. You should also be able to see it from another device connected to the same network at <your-ip>:9000.
Putting uWSGI behing NGINX reverse proxy
I like to run uWSGI behind a reverse proxy such as NGINX. This is fairly easy with this configuration file:
upstream backend {
least_conn;
server localhost:9000;
}
server {
listen 80;
location {
uwsgi_pass backend;
include uwsgi_params;
}
}
With such a configuration, NGINX listens on port 80 and redirects the requests to localhost:9000 which is our uwsgi server. Be careful though, since we use uwsgi_pass command, the uwsgi application must run in a socket and not on http. Just change your uwsgi.ini configuration file accordingly:
[uwsgi]
wsgi-file = main.py
master = true
processes = 2
socket = 0.0.0.0:9000
vacuum = true
die-on-term = true
Now, just run the application with uwsgi wsgi.ini and it will launch the
application with the configuration you just wrote. You can then write a startup
script with upstart or systemd to make your application run in the
background and easily start/stop it.
In another post, I’ll cover how to run all this with Docker, as it is what I tend to use now to deploy apps in production.