Azure App Service is a Paas solution to run Enterprise-grade Web Applications. Azure has recently announced App Services on Linux. It supports running web apps natively on Linux.

App Service Linux provides pre-defined application stacks(Nodejs, PHP and others ). App Service on Linux uses Docker containers to host these pre-built application stacks. You can also use a custom Docker image to deploy your web app to an application stack that is not already defined in Azure. More Info : https://docs.microsoft.com/en-us/azure/app-service-web/app-service-linux-using-custom-docker-image

In this blog, I would like to provide steps for running a Python app on App Service Linux. To make it little more fun, I would be using Magenta to play Piano. Final project is available on Github @ https://github.com/prashanthmadi/ai-duet-tensor

Below is the architecture of App Service Linux Python App on App Services Linux

Request would hit one of the App Service front-end and navigates to worker-role. Inside worker-role, There would be two different containers

  • Kudu Container(Used for Tooling Purpose : https://your_app_name.scm.azurewebsites.net)
  • App Container (Your App runs here)

Consider them as two different OS having their own file-system while share only /home directory(Remote persistent Storage). i.e, /var on Kudu is not same as /var on App container

Note : Only /home content is persistent, Anything written to other folders will be gone after App restart.

####Running Sample Python app on Local environment As I have mentioned earlier, we would be using Magenta to play Piano that responds.

  • Navigate to github page of Magenta and download ai-duet demo.
  • Follow steps in README to run this sample app.

Using waitress

  • Create run_waitress_server.py file in app and copy below content into it. 
    import os
from waitress import serve
from server import app
serve(app, host="127.0.0.1", port=8080)
  • Add waitress in requirements.txt file

####Add Docker Related Files

  • Create a new folder and copy above step sample app into app folder in project root.
  • Copy DockerFile and others at root folder from here
  • DockerFile would install
    • Python
    • nginx (Check nginx.conf for configuration)
    • openssh-server(Used to make web-ssh connection between Kudu and App Container. Check sshd_config file)
    • supervisor (To start nginx and python app simultaneous. Check supervisord.conf file)
    • Install Python App dependencies in requirements.txt file
  • init_container.sh will run on first request and starts ssh, supervisord.

Publish to DockerHub

  • Publish app to github
    • ex: https://github.com/prashanthmadi/ai-duet-tensor
  • Setup Automated build from git repository in Dockerhub.
    • ex: https://hub.docker.com/r/prashanthmadi/ai-duet-tensor/

Note: I have used Dockerhub but you can use other options like Azure Container Services.

Create App Services Linux

  • Follow steps @ https://docs.microsoft.com/en-us/azure/app-service-web/app-service-linux-how-to-create-web-app
  • Select Docker Hub in Configure container and set Image,Tag.
  • Ex : prashanthmadi/ai-duet-tensor:latest

    Final output

    url : http://ai-duet-sample.azurewebsites.net/

AI Duet on Azure App Service Linux