DevOpsBootcampUPES

Deploy applications in AWS and Kubernetes

Note: Can be done in 2-3 groups!

Never commit (and push) AWS credentials! No credentials file is needed to be placed in the repo directory.

Background

Your goal is to provision the object detection service as a scalable, micro-services architecture in AWS and Kubernetes.

Guidelines

  1. Deploy the frontend, yolo5 and polybot services as Deployments. Don’t forget to configure resources allocation (don’t give more than 1000mi CPU for request and limit) and probes (you may write some python code to handle probes).
  2. The yolo5 Deployment should have a horizontal pod autoscaler (HPA), min replicas is 1, max is 5. The average CPU utilization is 500mi.
  3. Store the telegram token as a Secret, and define a volumeMount and volume in the Deployment manifest to mount this secret as a .telegramToken file in the container’s file system.
  4. Deploy a stateful mongoDB app using a vanilla StatefulSet (utilize the Grafana manifest we’ve seen in class) or Helm chart.
  5. Expose the frontend add using an Ingress.

Test the app

  1. Make sure you are able to communicate with the service using both the Web UI or the Telegram bot.

  2. Test your app under scale, make sure the yolo5 service is scaling up and down properly.

pip install locust
cd projects/app_development_II
locust

Open http://localhost:8089. Provide the host name of your server and try it out!

  1. Test your service when performing a rolling update for each one of the services, is it updated seamlessly with zero-downtime?

Good Luck

This site is open source. Improve this page.