kafka-demo

name	kafka-demo
description	Manage the kafka-sensor-city demo. Run with no arguments to start the full demo. Run with "teardown" to remove all resources. Run with "teardown full" to also delete the minikube cluster.
allowed-tools	Bash(minikube *), Bash(kubectl *), Bash(docker *)
context	fork

Check $ARGUMENTS first and branch accordingly:

• If $ARGUMENTS starts with teardown → jump to the TEARDOWN section below.
• Otherwise → run the SETUP section.

---

TEARDOWN

Determine the teardown scope from $ARGUMENTS:

• teardown alone → remove all demo resources but leave minikube running
• teardown full → remove all demo resources and delete the minikube cluster

Step T1 — Delete all demo deployments

kubectl delete deployment sensor-simulator aggregator-service filter-service dashboard-service -n kafka --ignore-not-found

Confirm they are gone:

kubectl get deployments -n kafka

Step T2 — Delete the kafka namespace

This removes Kafka, all services, PVCs, and the ConfigMap in one shot:

kubectl delete namespace kafka --ignore-not-found

Wait until the namespace is fully terminated:

kubectl get namespace kafka

It should return Error from server (NotFound) — retry up to 60 seconds.

Step T3 — minikube (conditional)

If $ARGUMENTS is teardown full:

minikube delete

Inform the user the entire cluster and all cached images have been removed.

If $ARGUMENTS is just teardown: Inform the user that minikube is still running. They can run /kafka-demo to redeploy, or minikube stop / minikube delete manually when done.

---

SETUP

You are setting up the kafka-sensor-city demo end to end. Work through the steps below in order. After each shell command, check the output before continuing — stop and report clearly if anything fails.

Step 1 — minikube

Run:

minikube status

If it is not running, start it:

minikube start

Wait until kubectl get nodes shows one node with STATUS = Ready.

Step 2 — Deploy Kafka

Apply the manifest:

kubectl apply -f k8s/kafka/kafka.yaml

Wait for kafka-0 to reach 1/1 Running. Poll with:

kubectl get pods -n kafka

Retry every 5 seconds, up to 2 minutes. If it is still not Running after 2 minutes, run kubectl describe pod kafka-0 -n kafka and report what you find.

Step 3 — Create Kafka topics

Do NOT use an interactive exec. Run each command directly against the pod:

kubectl exec kafka-0 -n kafka -- /opt/kafka/bin/kafka-topics.sh \
  --bootstrap-server kafka:9092 \
  --create --topic raw-sensor-events \
  --partitions 6 --replication-factor 1

kubectl exec kafka-0 -n kafka -- /opt/kafka/bin/kafka-topics.sh \
  --bootstrap-server kafka:9092 \
  --create --topic intersection-events \
  --partitions 4 --replication-factor 1

kubectl exec kafka-0 -n kafka -- /opt/kafka/bin/kafka-topics.sh \
  --bootstrap-server kafka:9092 \
  --create --topic signal-state-changes \
  --partitions 4 --replication-factor 1

Verify all three exist:

kubectl exec kafka-0 -n kafka -- /opt/kafka/bin/kafka-topics.sh \
  --bootstrap-server kafka:9092 --list

If a topic already exists that is fine — continue.

Step 4 — Point Docker at minikube

All four docker build commands below must prefix with eval $(minikube docker-env) && so images land inside minikube, not your local Docker.

Step 5 — Build and deploy all services

Run each block in sequence. Wait for the build to succeed before moving to the next.

eval $(minikube docker-env) && docker build -t sensor-simulator:latest ./SensorSimulator
kubectl apply -f k8s/sensor-simulator.yaml

eval $(minikube docker-env) && docker build -t filter-service:latest ./FilterService
kubectl apply -f k8s/filter-service.yaml

eval $(minikube docker-env) && docker build -t aggregator-service:latest ./AggregatorService
kubectl apply -f k8s/aggregator-service.yaml

eval $(minikube docker-env) && docker build -t dashboard-service:latest ./DashboardService
kubectl apply -f k8s/dashboard-service.yaml

Step 6 — Wait for all pods

Poll until all five pods in the kafka namespace show 1/1 Running:

kubectl get pods -n kafka

Expected pods: kafka-0, sensor-simulator-*, filter-service-*, aggregator-service-*, dashboard-service-*.

If any pod is in CrashLoopBackOff or Error, fetch its logs:

kubectl logs -n kafka <pod-name> --tail=40

Report what went wrong before stopping.

Step 7 — Open the dashboard

minikube service dashboard-service -n kafka

Inform the user that their browser should open automatically and that this terminal must stay open — minikube holds the tunnel in the foreground on macOS. The city map will populate within a few seconds as the first intersection events arrive.