Publish-Subscribe PoC

Overview

Some notable tools utilized in this PoC:

• k3s - Very lightweight Kubernetes distribution running in Docker
• .NET 6/7 Minimal APIs
• RabbitMQ - Message queue
• MassTransit - Abstraction for messaging with RabbitMQ
• KEDA - Kubernetes autoscaler
• k6 - Load testing tooling used to stress test the system

Installation

1. Create the cluster:

   k3d registry create registry.local --port 0.0.0.0:5000
   k3d cluster create -c local-setup/k3d.config.yaml --registry-use k3d-registry.local:5000
   

2. Build the pubi and subi services:

   docker-compose build
   

3. Import the images from local Docker to the k3d cluster:

   k3d image import pubisubi-subi pubisubi-pubi -c local-k3s
   

4. Install rabbitmq operator (run 2x as CRDs are pre-requisite for install):

   kubectl apply -k ./local-setup/rabbitmq/
   

5. Deploy pubi and subi services:

   kubectl apply -f ./local-setup/subi/ -f ./local-setup/pubi/
   

6. Deploy KEDA autoscaler (run 2x as CRDs are pre-requisite for install):

   kubectl apply -k ./local-setup/keda/
   

API Load Testing

docker run -it -v ${PWD}/k6/scripts:/scripts grafana/k6:latest run -u 100 -d 5s /scripts/work-script.js

TODO

1. Run the setup to verify that it still works
2. Draw a diagram of the setup
3. Create a CI for Pubi and Subi services
   • Build the containers
   • Scan the containers for vulnerabilities using Trivy (Optional)
4. Use GitOps for deployment
   • Install FluxCD
   • Add Flux Kustomizations and regular Kustomizations
   • Build OCI artifact from manifests
5. Improve k6 load tests
   • Verify the request responses
   • Run as a Kubernetes job