birthday-demo

Simple HTTP API application with Cloud deploy playbook

In Scope

• Simple Hello World style Python Flask application with unit and functional tests.
  • I've opted not to use a Database to reduce the amount of time required to complete this excerise.
    • ie the playbooks, cloudformation, localdev enhancements to support a Database.
• System diagram of production AWS Infrastructure.
• Cloud build and deploy script.

Out of Scope

• Productionisation of application

  • Dockerise localdev and split requirements.txt to reduce docker image size.
  • Front with Nginx and WSGI
  • Use an API framework like Django Rest Framework
  • Use Clustered HA DB
  • Log to CloudWatch Logs
  • add /healthz on an internal only port

• Developer quality of life enhancements

  • repo setup script (distributed separately).
    • verify docker and pyenv installed.
    • clone with ticket number included in the folder and branch names.
    • create pyenv and install required python packages.
    • run pre-commit install.
  • pyinvoke/make for commonly used developer actions.

• CICD: AWS CodePipeline/CodeBuild

  • Enable branch projection rules on main branch.
  • Codebuild for PRs that runs pre-commit run --all-files and tests with coverage.
  • CodePipeline triggered on merge to main branch.
  • Post staging deploy smoke tests followed by auto deploy to prod if in release window.
  • Centralised channel reporting deployments of all applications.

• Monitoring: AWS CloudWatch Dashboard

  • Load Balancer request and error rates plus application response times
  • AutoscalingGroup CPU load and instance count
  • Per instance CPU, memory and disk usage

• Alerting: AWS CloudWatch Alarms

  • Email and or Messaging of owning team for
    • Disk, CPU, memory, error rate or response time out of bounds
    • Instance healthcheck failures
    • CICD stage failures

• Exception Handling: Sentry

  • Configuration of App to send exceptions to Sentry
  • Configuration of Sentry to notify teams

• Secret/Env-variable Management: Envars

  • pre-commit checks for unencrypted secrets
  • I've developed a solution that leans towards developer ease of use over centralisation
    • https://github.com/timeoutdigital/envars
  • Other solutions are available and appropriate.

• Disaster Recovery

  • Copy nightly RDS snapshots to another AWS region

Architectural design choices

There are many ways to skin a cat but generally:

• Avoid vendor lock-in.
  • For instance we could use AWS API Gateway but Django Rest Framework is an excellent open source option.
• Maintainability over raw performance.
  • NodeJs can be more performant than Python but in a Python shop it will induce more developer overhead.

Code implementation choices

• My first choice would be Django Rest Framework but this is hardly a 'simple' solution.
• Considered using python http.server as the simplest possible python implementation but it doesn't support PUT http method.
• Python Flask looks like the next simplest option.

Localdev

• Install required python packages

pip install -r requirements.txt

• Install pre-commit hook

pre-commit install

• Run app

cd birthday_demo
flask run --debug

Testing

• Run tests.

pytest

Deployment

• Run deploy playbook

cd devops/ansible
./birthday_demo_deploy.yml -e env=<staging|prod>