This project is done is accordance to the udacity specifications to build a complete CI/CD for a flask based application that provide a inference endpoint for house price predictions based on the features of the house. The application is deployed as a web app on Azure using (Azure App Services).
- Trello board for the project: https://trello.com/b/5i2rmPGT/housepredictions-of-project-management-template
- Full project plan spreadsheet: https://docs.google.com/spreadsheets/d/1fb1SpfHJbU5I71yWouwonHabAM7OO0xeGj4jpk9s6sU/edit?usp=sharing
After cloning the project, you need to perform the following steps to deploy it to azure:
- Install azure CLI.
- Login to your azure account
azure login - Deploy the app to Azure App Services:
az webapp up -g houseprediction-rg -p house-prediction-service-plan -n house-prediction-app -l westeurope --sku B1 --runtime PYTHON:3.9 - For automatic deployment, configure azure DevOps, create a new pipeline and make sure to add a Service Connection to the deployed Azure App Services app.
- Make sure to select the pipeline configuration file from the code repository
azure-pipelines.yml - To publish a new release, you need to give your commit a tag for the azure pipeline to trigger a new deployment
Here is an example
git add .git commit -m "Release 4 implementation"git tag release.4git pushgitpush origin release.4 - Once the code is pushed, Github action will trigger the necessary tests and Azure pipeline will perform a new deployment.
- Architectural Diagram
- Project running on Azure App Service
- Project cloned into Azure Cloud Shell
- Output of a test run
- Successful deploy of the project in Azure Pipelines.
- Running Azure App Service from Azure Pipelines automatic deployment
- Successful prediction from deployed flask app in Azure Cloud Shell.
- Output of streamed log files from deployed application
- Load testing results
From a devops engineer prespective, I would definitly consider using infrastructure as code such as Terraform to deploy the web app instead of the CLI. Developers should avoid working on the main branch and use git flow for better repo organziation. I would also suggest using three different environment for development, staging and production.