333-333-333/go-load-testing

When to Use

• Validating service performance under expected and peak load
• Finding the breaking point of an endpoint
• Testing sudden traffic spikes (e.g., promotional push notifications)
• Detecting memory leaks or connection pool exhaustion (soak tests)
• Running performance smoke tests before releases

---

Critical Patterns

Test Types

| Test Type | Purpose | VUs | Duration | |-----------|---------|-----|----------| | Smoke | Verify service works under minimal load | 1-5 | 30s | | Load | Validate under expected production traffic | 50-100 | 5m | | Stress | Find breaking point | 100→500 ramp | 10m | | Spike | Test sudden traffic burst | 1→500→1 | 3m | | Soak | Find memory leaks, connection pool exhaustion | 50 | 30m+ |

Directory Structure

k6 scripts live under tests/load/{service}/:

tests/load/
  notification/
    smoke.js           # Quick smoke test (1-5 VUs)
    load.js            # Normal load test (50-100 VUs)
    stress.js          # Stress test (ramp to 500 VUs)
    spike.js           # Spike test (sudden burst)
  auth/
    smoke.js
    load.js

Thresholds

Every k6 test MUST define thresholds — the pass/fail criteria:

| Metric | Target | Meaning | |--------|--------|---------| | http_req_duration p(95) | < 200ms | 95% of requests under 200ms | | http_req_duration p(99) | < 500ms | 99% of requests under 500ms | | http_req_failed | < 1% | Less than 1% error rate | | http_reqs | > 100/s | Minimum throughput |

---

Decision Tree

Quick sanity check before deploy?
  → Smoke test (1-5 VUs, 30s)

Validating expected production traffic?
  → Load test (50-100 VUs, 5 min)

Finding the breaking point?
  → Stress test (ramp to 500 VUs)

Testing sudden traffic spike (e.g., push campaign)?
  → Spike test (1 → 500 → 1)

Detecting memory leaks or pool exhaustion?
  → Soak test (50 VUs, 30+ min)

---

Assets

| File | Description | |------|-------------| | assets/k6-smoke.js | Smoke test: minimal load sanity check | | assets/k6-load.js | Load test: expected production traffic with ramp stages | | assets/k6-stress.js | Stress test: find breaking point with increasing load | | assets/k6-spike.js | Spike test: sudden burst of traffic |

See each asset for the full script template with thresholds.

---

Commands

brew install k6                                             # Install k6
k6 run tests/load/notification/smoke.js                     # Run smoke test
k6 run tests/load/notification/load.js                      # Run load test
k6 run --out json=results.json tests/load/notification/stress.js  # With JSON output

---

Anti-Patterns

| Don't | Do | |----------|-------| | Load test without thresholds | Always define p95, p99, error rate | | Skip smoke test before deploy | Smoke is 30 seconds — always run it | | Run stress tests against production | Run against staging or a dedicated environment | | Ignore soak test results | Memory growth means a leak — fix it | | Hardcode URLs in scripts | Use k6 environment variables for base URL |