Thresholds & Reports

Under Reviewv0.1.0-alpha

---

SLO Thresholds

Defined in load-tests/config/profiles.js and applied uniformly to every profile:

Metric	Threshold	Meaning
http_req_failed	rate < 0.01	HTTP error rate must stay below 1%
http_req_duration	p(95) < 350ms	95th-percentile response time under 350 ms

When either threshold is crossed, k6 exits with code 99. This makes it directly usable as a CI gate — the pipeline step fails if SLOs are violated.

// config/profiles.js
export const THRESHOLDS = {
  http_req_failed:   ['rate<0.01'],
  http_req_duration: ['p(95)<350'],
};

---

Stock Error Note

400 Insufficient Stock responses count as HTTP failures and increment http_req_failed. The setup() function seeds stock per profile to prevent this:

Profile	Stock per variant
smoke	100
load	5 000
stress / spike	50 000
soak	10 000

If you see unexpected stock errors, either re-run setup() by resetting the database, or manually increase quantities for the affected variants.

---

Reports

After every run, handleSummary automatically writes a timestamped JSON file:

load-tests/reports/<profile>_<YYYY-MM-DD>.json

The file contains the full k6 metrics object — counters, rates, trends, and histogram buckets — and can be ingested by Grafana, DataDog, or any JSON-aware observability tool.

// main.js — handleSummary
export function handleSummary(data) {
  const timestamp = new Date().toISOString().split('T')[0];
  return {
    'stdout': textSummary(data, { indent: '  ', enableColors: true }),
    [`load-tests/reports/${profile}_${timestamp}.json`]: JSON.stringify(data, null, 2),
  };
}

The reports/ directory is git-ignored. Commit only reports you explicitly want to version.

---

CI Integration

k6 exits with code 99 when thresholds are crossed. Wire it into your pipeline:

# GitHub Actions example
- name: Run load test
  run: k6 run -e PROFILE=load load-tests/main.js
  env:
    BASE_URL: ${{ secrets.STAGING_URL }}

The step will fail if either SLO threshold is violated, blocking the merge or deploy.

Metrics Grouping in Dashboards

Every HTTP call in the test suite is tagged with a name: param:

// helpers/api.js example
http.get(`${baseUrl}/api/products`, { tags: { name: 'GET /api/products' } });

In k6 Cloud, Grafana, or any Prometheus-compatible backend, filter by this tag to get per-route latency breakdowns instead of aggregate totals.