github.com/Petrosz007/go-k8s-search-autocomplete

Go k8s Search Autocomplete

HTTP service which provides search autocompletion for Kubernetes resources running in the cluster.

Required tools

Assuming Docker and Go is installed on your machine, you'll need these project specific packages:

brew install minikube just bruno golangci-lint

# Optional: to use bruno on the cli install bruno-cli with npm, the intergation test script uses this
npm install -g @usebruno/cli

Running it locally

[!WARNING]
Minikube will edit your Kubeconfig and add its configuration there. It should add it next to your current config, but it's never a bad idea to back it up, just in case.

To run the app in the local minikube cluster:

just run-in-cluster

This will:

• Start minikube
• Build the app
• Build a Docker image inside minikube's docker env
• (Re-)deploy the necessary k8s resources to minikube
• Start port forwarding on 8080 and start showing the logs

You can now query the app, for example with:

curl localhost:8080/search/autocomplete/pods

Integration testing

Integration tests are set up with Bruno and bruno-cli.

To run them automatically:

just integration-test

Notes

I've added // TODO comments for parts which would be nice to write, but would require more effort than it's worth for this POC.

I've added // ? comments for parts, where I'm not too familiar with the pragmatic Go way of writing things, and I'm not sure what's the best choice.

Future improvement ideas

• Write more tests for the Go code
  • Mock out the k8s API to test the my k8s module
  • Mock out my k8s module to test the API layer
• Config loading
  • Right now only the port could be a config
  • k8s client could be configured with other options, timeouts could be set, ...
• Better integration testing with Bruno
  • The tests should better test the values returned by the API. These current tests are enough to check if the API is alive and responds with some correctly formatted data.
• Prometheus metrics
• Tracing with OpenTelemetry
  • It's very cool, but probably far down the priority list
• Swagger UI
• Automated CI/CD pipeline
  • Building and packaging is easy
  • Minikube setup might be a bit tricky
  • The bin/integration-test-in-cluster.sh script runs the tests automatically already, we just need to call it in the pipeline