Skip to main content

Single Server Web App Deployment

Deploying small web applications can take up more time than creating the application itself, for this reason I've put together this small guide to document how I go about deploying applications. I've used this method for things like small business sites, demo sites and application prototypes. Although we will go through a Python Flask application the same method would apply to other micro web apps.
The final running application will have a few moving parts, here is how it will handle two incoming requests:
Our final implementation will comprise three Docker containers:

- Our frontend reverse proxy server - nginx
- Our WSGI server Gunicorn and application server Flask
- An example ancillary service - a redis cache

To follow along grab the code from https://github.com/hardbyte/python-app-deploy

In this post we’ll create a small standalone flask application that uses redis for a cache. The same deployment method works for applications that require services like memcached, postgresql, etcd.

Start small

So first up we write and test the application locally. Our flask app is just one file - here on github. The application will just have two endpoints: /bump and /hits. We will skip the implementation details for now (see them on github) and jump straight to building a docker image ready for deployment. We will serve our app using the WSGI HTTP server gunicorn. In the Dockerfile we install the package requirements, copy over the Python files and bind to port 5001:

FROM python:3.5

# If you have a src folder add that instead
ADD . /var/www

COPY requirements.txt /var/www/requirements.txt

WORKDIR /var/www

RUN pip install -r requirements.txt

EXPOSE 5001

# Useful for debugging
#CMD python3 server.py

# Serve using gunicorn
# Note this should additionally be behind nginx
CMD gunicorn server:app -w 4 -b 0.0.0.0:5001


We build the docker image and test it locally. Here we just use a stock standard redis container from dockerhub, and use Docker’s linking mechanism to facilitate connecting the two containers. When we create these container links we can use specific names on the system (myapp-redis) but then internally expose them more generically as e.g. redis. This becomes very useful if we deploy multiple applications on a single server.
docker build -t hardbyte/myapp-server
docker run -d --name myapp-redis redis
docker run -d --name myapp -p 5001:5001 --link myapp-redis:redis hardbyte/myapp-server
This image is available on docker hub as hardbyte/myapp-server.

Adding a proxy server

Next we want to put a proxy server such as nginx in front of our application server, this will help protect our application from DDOS attacks by buffering slow clients. Do read through the deployment documentation for gunicorn yourself though!

So to do this we will add another docker container into the mix. I’ve created a top level folder myapp-nginx which is where we will put our nginx configuration files and the nginx Dockerfile.
The Dockerfile is really simple (github) - we just copy the configuration directory and any static assets to the image. The more interesting part is the nginx configuration file nginx/default.conf which contains the routes:

server {
  listen 80;
  location / {
    # Pass the request to Gunicorn/Flask app
    proxy_pass http://backend:5001;
    proxy_redirect     off;
    proxy_set_header Host $host;
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
  }

  # Serve our static files too
  location /static/ {
    alias /usr/share/nginx/html/;
  }
}
Although here we are serving the static files as /static often this is switched around - we could easily have the backend at e.g., /api, and the website being served on root (/).
Now we’ve added hardbyte/myapp-nginx into the mix we have to link the nginx container to our application backend and finally we might want to serve traffic on port 80. Build or pull the docker image as before and start the container with:

docker run -d --name myapp-nginx -p 80:80 --link myapp:backend hardbyte/myapp-nginx

Deploy with systemd

Once we have the containers all working locally there are a few directions in which we can go for actually deploying the application. We don’t want to have to manually ssh into a server and run the docker commands. We want to make sure if the server restarts that so do all our docker containers. For this post we will assume just a single web server, however a future post will talk about deploying on a cluster.
Most modern linux operating systems now use systemd to handle running services. One such lightweight server operating system is CoreOS, it comes with a solid security model and predictable automated update behaviour. If you don’t have access to a machine with systemd it is very easy to spin up a coreos VM on your favourite cloud provider.
The remaining piece of the puzzle is creating systemd .service files, they comprise the instructions to the server on how to run, stop, restart, and initialize services. Each service is very similar, they each pull their docker image, and run a container with the appropriate command line incantation.
Due to the docker containers having required links, we will create services with the same dependencies. One to run each of redis, the app, and nginx.
The service files are available on github, we will just show one service file in full.
The Requires and After statements are both required. Because we are launching docker containers we rely on the docker.service.

myapp.service - An example systemd service File

[Unit]
Description=My App Server

# Requirements
Requires=docker.service
Requires=myapp-redis.service

# Dependency ordering
After=docker.service
After=myapp-redis.service

[Service]

# Let processes take awhile to start up (for pulling initial Docker containers)
TimeoutStartSec=500

KillMode=none

ExecStartPre=-/usr/bin/docker kill myapp-server
ExecStartPre=-/usr/bin/docker rm myapp-server

ExecStart=/usr/bin/docker run --rm --name myapp-server --link myapp-redis:redis hardbyte/myapp-server

ExecStop=/usr/bin/docker stop myapp-server

[Install]
WantedBy=multi-user.target


Note directives with =- are allowed to fail, this is useful because docker complains when trying to rm containers that don’t yet exist.
These service files have to be both enabled and started before systemd will know to run the docker containers at boot. To do this we have to copy the service files to /etc/systemd/system/.

systemctl enable myapp
systemctl start myapp

Modification if using a private Docker repository

If you use a quay.io account, just add these ExecStartPre commands to login to quay.io. Robot accounts are a good idea here too!

ExecStartPre=-/usr/bin/docker login -e="." -u="quayaccount+myapp" -p="PASS" quay.io
ExecStartPre=-/usr/bin/docker pull quay.io/quayaccount/myapp-server

Accessing logs and containers’ stdout

You can use journalctl to get the logs that systemd collects, or as these are just normal Docker containers running on a single host we can use docker logs. Consider mounting a logging directory from the host machine to collect logs outside of docker.

Brief cloudy diversion

I usually have another couple of layers in front of nginx which I’ll briefly mention:
  • AWS Load Balancer - Even when deploying on a single machine a load balancer is useful as it can monitor server health and alert you when your appliction is down.
  • AWS CloudFront - Backed by the ridiculously reliable S3 service this can serve your static assets from edge locations all around the globe.

How do you deploy?

We have explored a couple of powerful tools for deploying web applications to a single server so you end up with a reliable running application. Automated deployment is a big issue that I’ve not mentioned, a great start is to have a continuous delivery pipeline that tests code and then builds and uploads new docker containers. A sidekick service can be written that will update the server too. As you’d expect with additional servers you can dramatically increase the reliability, but if anyone else has different ideas for single server deployment please share them with me by commenting or mentioning me on twitter @thorneynz.

Popular posts from this blog

Bluetooth with Python 3.3

Since about version 3.3 Python supports Bluetooth sockets natively. To put this to the test I got hold of an iRacer from sparkfun . To send to New Zealand the cost was $60. The toy has an on-board Bluetooth radio that supports the RFCOMM transport protocol. The drive  protocol is dead easy, you send single byte instructions when a direction or speed change is required. The bytes are broken into two nibbles:  0xXY  where X is the direction and Y is the speed. For example the byte 0x16 means forwards at mid-speed. I was surprised to note the car continues carrying out the last given demand! I let pairing get dealt with by the operating system. The code to create a  Car object that is drivable over Bluetooth is very straight forward in pure Python: import socket import time class BluetoothCar : def __init__ ( self , mac_address = "00:12:05:09:98:36" ): self . socket = socket . socket ( socket . AF_BLUETOOTH , socket . SOCK_STREAM , socket .

Matplotlib in Django

The official django tutorial is very good, it stops short of displaying data with matplotlib - which could be very handy for dsp or automated testing. This is an extension to the tutorial. So first you must do the official tutorial! Complete the tutorial (as of writing this up to part 4). Adding an image to a view To start with we will take a static image from the hard drive and display it on the polls index page. Usually if it really is a static image this would be managed by the webserver eg apache. For introduction purposes we will get django to serve the static image. To do this we first need to change the template. Change the template At the moment poll_list.html probably looks something like this: <h1>Django test app - Polls</h1> {% if object_list %} <ul> {% for object in object_list %} <li><a href="/polls/{{object.id}}">{{ object.question }}</a></li> {% endfor %} </ul> {% else %} <p>No polls

Homomorphic encryption using RSA

I recently had cause to briefly look into Homomorphic Encryption , the process of carrying out computations on encrypted data. This technique allows for privacy preserving computation. Fully homomorphic encryption (FHE) allows both addition and multiplication, but is (currently) impractically slow. Partially homomorphic encryption just has to meet one of these criteria and can be much more efficient. An unintended, but well-known, malleability in the common RSA algorithm means that the multiplication of ciphertexts is equal to the multiplication of the original messages. So unpadded RSA is a partially homomorphic encryption system. RSA is beautiful in how simple it is. See wikipedia to see how to generate the public ( e , m ) and private keys ( d , m ). Given a message x it is encrypted with the public keys it to get the ciphertext C ( x ) with: C ( x ) = x e mod m To decrypt a ciphertext C ( x ) one applies the private key: m = C ( x ) d mod m The homomorphic prop