gravl/k8s/production/k6-load-test-job.yaml

---
apiVersion: batch/v1
kind: Job
metadata:
  name: k6-load-test
  namespace: default
  labels:
    app: gravl
    component: load-testing
spec:
  backoffLimit: 1
  template:
    metadata:
      labels:
        app: gravl
        component: load-testing
    spec:
      containers:
      - name: k6
        image: grafana/k6:latest
        imagePullPolicy: IfNotPresent
        command:
          - k6
          - run
          - --out=json=/tmp/results.json
          - /test/load-test.js
        env:
        - name: GRAVL_API_URL
          value: "http://gravl-backend.gravl-prod:3000"
        - name: K6_VUS
          value: "10"
        - name: K6_DURATION
          value: "5m"
        volumeMounts:
        - name: test-script
          mountPath: /test
        - name: results
          mountPath: /tmp
        resources:
          requests:
            cpu: 500m
            memory: 256Mi
          limits:
            cpu: 1000m
            memory: 512Mi
      volumes:
      - name: test-script
        configMap:
          name: k6-test-script
      - name: results
        emptyDir: {}
      restartPolicy: Never
      serviceAccountName: default

---
# ConfigMap with k6 test script
apiVersion: v1
kind: ConfigMap
metadata:
  name: k6-test-script
  namespace: default
  labels:
    app: gravl
    component: load-testing
data:
  load-test.js: |
    import http from 'k6/http';
    import { check, sleep } from 'k6';
    import { Rate, Trend, Counter, Gauge } from 'k6/metrics';

    // Custom metrics
    const errorRate = new Rate('errors');
    const requestDuration = new Trend('request_duration');
    const requestCount = new Counter('requests');
    const activeConnections = new Gauge('active_connections');

    // Test configuration
    export const options = {
      vus: parseInt(__ENV.K6_VUS || '10'),
      duration: __ENV.K6_DURATION || '5m',
      thresholds: {
        'http_req_duration': [
          'p(95)<200',  // 95th percentile must be below 200ms
          'p(99)<500',  // 99th percentile must be below 500ms
        ],
        'http_req_failed': ['rate<0.1'],  // error rate must be below 10%
        'errors': ['rate<0.01'],
      },
      setupTimeout: '30s',
      teardownTimeout: '30s',
    };

    const BASE_URL = __ENV.GRAVL_API_URL || 'http://localhost:3000';

    export function setup() {
      console.log(`Starting load test against ${BASE_URL}`);
      return { start_time: new Date().toISOString() };
    }

    export default function (data) {
      activeConnections.add(1);

      // Health check endpoint
      {
        let response = http.get(`${BASE_URL}/api/health`, {
          timeout: '10s',
        });
        
        check(response, {
          'health check returns 200 or 503': (r) => r.status === 200 || r.status === 503,
          'health check has content': (r) => r.body.length > 0,
        });
        
        errorRate.add(response.status >= 500);
        requestDuration.add(response.timings.duration);
        requestCount.add(1);
      }

      sleep(1);

      // List exercises endpoint
      {
        let response = http.get(`${BASE_URL}/api/exercises`, {
          timeout: '10s',
        });
        
        check(response, {
          'exercises endpoint returns 2xx or 404': (r) => r.status >= 200 && r.status < 300 || r.status === 404,
        });
        
        errorRate.add(response.status >= 500);
        requestDuration.add(response.timings.duration);
        requestCount.add(1);
      }

      sleep(1);

      // Prometheus metrics endpoint (optional)
      {
        let response = http.get(`${BASE_URL}:3001/metrics`, {
          timeout: '5s',
          noResponseCallback: 'ignore',
        });
        
        if (response) {
          requestDuration.add(response.timings.duration);
        }
        requestCount.add(1);
      }

      sleep(1);
      activeConnections.add(-1);
    }

    export function teardown(data) {
      console.log(`\n=== Load Test Results ===`);
      console.log(`Total requests: ${requestCount.value}`);
      console.log(`Error rate: ${(errorRate.value * 100).toFixed(2)}%`);
      console.log(`Average p95 latency: ${requestDuration.value.p(95)}ms`);
      console.log(`Average p99 latency: ${requestDuration.value.p(99)}ms`);
      console.log(`Start time: ${data.start_time}`);
      console.log(`End time: ${new Date().toISOString()}`);
    }