A series of articles explaining Kubernetes foundations by developing an application that is deployed to a K8s cluster.
This service will be a customer churn prediction ML pipeline. The prediction model will be exposed as an API from an HTTP server.
This API can be used in two ways.
- A call center employee can enter the data via a simple web form and get a likelihood of churn. If it’s too high, they can offer special discounts.
- An automated process will scan the customer database for people likely to churn and email them a special discount coupon.
- This job will add names to a queue, and another job will process the queue and email, and another job will update the database
The ML will be the customer churn regression from Chapter 3 of Machine Learning Bookcamp.
Custom web app
HTTP API for Model
Multiple pods with a Service between them
External Service (database?)
ConfigMap and/or Secrets
Database installation
Persistent Volumes
HA database?
Monitoring
Prometheseus?
PLONK stack?
Deploy to cloud service