miroir/tests/p6_8_README.md

P6.8 Multi-pod Kubernetes Acceptance Tests

This directory contains acceptance tests for Phase 6 (Horizontal Scaling + HPA) as specified in plan §14 Definition of Done.

Prerequisites

• kind (Kubernetes in Docker)
• kubectl
• helm
• docker
• curl, jq

Installing Prerequisites

# Install kind
go install sigs.k8s.io/kind@latest

# Install kubectl (Linux)
curl -LO "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/amd64/kubectl"
chmod +x kubectl
sudo mv kubectl /usr/local/bin/

# Install helm
curl https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3 | bash

Running Tests

Run All Tests

./tests/p6_8_multi_pod_acceptance.sh

Run Specific Test

./tests/p6_8_multi_pod_acceptance.sh 1  # Test 1: Multi-pod deployment
./tests/p6_8_multi_pod_acceptance.sh 2  # Test 2: Peer discovery
./tests/p6_8_multi_pod_acceptance.sh 3  # Test 3: Mode B leader election
./tests/p6_8_multi_pod_acceptance.sh 4  # Test 4: Resource-pressure metrics
./tests/p6_8_multi_pod_acceptance.sh 5  # Test 5: PrometheusRule alerts
./tests/p6_8_multi_pod_acceptance.sh 6  # Test 6: HPA configuration
./tests/p6_8_multi_pod_acceptance.sh 7  # Test 7: Resource limits

Test Coverage

Test	Description	Acceptance Criterion
1	Multi-pod deployment	replicas=3 — every pod independently serves requests
2	Peer discovery	Pods discover each other via headless Service
3	Mode B leader election	Exactly one leader; failover within lease_ttl_s
4	Resource-pressure metrics	All §14.9 metrics present
5	PrometheusRule alerts	All §14.9 alerts in manifest
6	HPA configuration	HPA installed with correct metrics
7	Resource limits	Pod resources match §14.1 envelope

What Gets Tested

The test script:

1. Creates a kind cluster with 1 control-plane and 3 worker nodes
2. Installs dependencies: Redis (task store), Meilisearch (3 nodes)
3. Builds Miroir Docker image from local source
4. Installs Miroir Helm chart with 3 replicas
5. Runs acceptance tests against the live cluster
6. Cleans up by deleting the kind cluster

Manual Verification (Without kind)

If you don't have kind installed, you can verify the Helm chart templates manually:

# Render the Helm chart with 3-replica values
helm template miroir charts/miroir \
  --set miroir.replicas=3 \
  --set miroir.hpa.enabled=true \
  --set miroir.taskStore.backend=redis \
  --set redis.enabled=true \
  > /tmp/miroir-rendered.yaml

# Check HPA is present
grep -A 30 "kind: HorizontalPodAutoscaler" /tmp/miroir-rendered.yaml

# Check PrometheusRule is present
grep -A 50 "kind: PrometheusRule" /tmp/miroir-rendered.yaml

# Check headless Service for peer discovery
grep -A 10 "miroir-headless" /tmp/miroir-rendered.yaml

CI Integration

These tests should be integrated into the Argo Workflow miroir-ci (Phase 8) as a separate step that runs on tag/release, not on every commit.

Known Limitations

1. Mode A test (anti-entropy partitioning) requires more complex setup with actual anti-entropy passes - not yet automated
2. Mode C test (dump import chunking) requires a 10GB test dump - not yet automated
3. Chaos test (pod kill mid-traffic) requires background traffic generation - not yet automated

The current tests verify the infrastructure is correctly configured. Full behavioral tests will be added in future iterations.

Troubleshooting

kind cluster fails to create

# Check Docker is running
docker ps

# Delete any existing cluster
kind delete cluster --name miroir-p6-test

Pods not ready

# Check pod status
kubectl get pods -n miroir-test

# Check pod logs
kubectl logs -n miroir-test <pod-name>

Port-forward fails

# Check if port 7700 is already in use
lsof -i :7700

# Kill any process using the port
kill -9 <pid>