Load testing

You may load test against Meridian - in the test environment only. This page covers the rules and how to structure a meaningful test.

Rules

1. Test environment only. mrnd_test_ keys exclusively - live keys are never for load generation.
2. Stay inside your test quota (600 req/min by default) unless you've been granted a temporary bump.
3. No destructive floods - don't hammer delete/restore cycles or delivery replays; they're heavier than reads.
4. Identify yourself - send a X-Load-Test: <name> header so our ops team can distinguish your traffic from abuse patterns.

Request a quota bump

For sustained tests above 600 req/min, contact support 3+ business days ahead with:

• Target RPS and duration
• Time window (we prefer off-peak hours in your region)
• Which endpoints you'll exercise

Approved bumps are temporary (typically 4–24h) and apply to your test environment only.

Structuring a test

meridian projects create --file fixtures/load-test.json

X-RateLimit-Limit: 600
X-RateLimit-Remaining: 41
X-RateLimit-Reset: 1726570000

Tooling

Standard tools work fine against the REST API:

k6 run - <<'EOF'
import http from 'k6/http';
export const options = { vus: 20, duration: '5m' };
export default function () {
  http.get('https://api.meridian.dev/v2/projects', {
    headers: {
      Authorization: `Bearer ${__ENV.MERIDIAN_API_KEY}`,
      'X-Load-Test': 'k6-baseline',
    },
  });
}
EOF

echo "GET https://api.meridian.dev/v2/projects" | \
  vegeta attack -rate=10/s -duration=5m \
    -header="Authorization: Bearer $MERIDIAN_API_KEY" \
    -header="X-Load-Test: vegeta-baseline" | \
  vegeta report

What to measure

• p50/p95/p99 latency per endpoint - our published targets are in Status.
• Error rate - anything above a trace of 429/5xx means you're above quota or hitting a real issue.
• Retry behavior - confirm your client's backoff doesn't stampede; good tests measure recovery, not just throughput.

Webhooks under load

Load-testing your receiver: use /v2/simulate/deliveries to fan out test events at rate, or point several endpoints at the same URL. Deliveries in test are rate-limited the same as live, so the distribution is representative.