Kubernetes Status API

This project is intended as an exercise.

Description

This repo contains an HTTP API service that expose the status of a Kubernetes cluster.

Considerations

• I will be using Go. Go is a good choice when it comes to build backend services thanks to its type-safety and rich ecosystem of modules that will help me in this project. Also, the Kubernetes client is very DevX friendly to use in Go.

• To implement the HTTP web server and also to keep the project structure well organized easily extensible in the future, I will be using Gin.

• For Kubernetes I will be using Okteto Cloud. The service will run inside the Kubernetes cluster itself. If the code run elsewhere, some changes are required to retrieve the K8S's config.

• I will be using OpenAPI and {json:api} to define the API specs and data structure

Development

Using a CDE (Gitpod.io)

Launching the project in Gitpod will automatically connect to your Okteto cluster. Add the following env vars in your Gitpod vars:

• OKTETO_KUBECONFIG: get this from the Okteto UI or CLI, and encode it in base64
• OKTETO_TOKEN: get this from the Okteto UI

Make sure to set a strict scope to avoid leaking the variables.

Requirements

As requirement to run this, you need a Kubernetes cluster. In my case I am using Okteto Cloud. Check out their Go sample project to know more about Okteto.

To launch the project in Okteto, follow these steps:

1. Get the Okteto CLI
2. okteto context use https://<YOUR_NAMESPACE>.okteto.net
3. okteto deploy --build
4. okteto up

Testing

Run make test to execute the unit tests

Publish to Docker

This step is only required in case the development image for Okteto needs some changes.

make docker-publish TAG=1.0.0

Future improvements

This is just an exercise so it doesn't come with lot of production-ready stuff :)

These are some of the things I would consider do, to make it closer to be production-ready.

• Use the ./openapi.yml specs for request/response validation and potentially for code generation
• Add error rate and latency alerts or SLOs
• Setup a CI/CD pipeline
• Consider introducing API versioning (e.g /v1)
• Add semantic versioning of the service
• Consider how to protect the API (ie. auth)

Considerations

This project was quite fun to build and it was the first time for me integrating with the Kubernetes API so I had to learn how to do it.

In terms of testing, I've added some essential unit tests to: - validate the sorting logic - validate the integration with the kubernetes's API using the provided mock client

Useful links

I've encountered a couple of useful blog posts that I've used to accomplish this exercise:

• https://betterprogramming.pub/build-a-scalable-api-in-go-with-gin-131af7f780c0
• https://medium.com/the-phi/mocking-the-kubernetes-client-in-go-for-unit-testing-ddae65c4302