Load Testing Guide

Resilience testing under realistic traffic — combining chaos with load simulation.

Why Test Under Load?

A system that handles faults at idle may fail under production traffic. Load testing during chaos experiments reveals:

• Degradation curves — how performance degrades under fault
• Recovery behavior — how quickly throughput recovers after fault removal
• Cascading failures — faults that only manifest under load (connection pool exhaustion, thread starvation)

Supported Tools

Tool	Best For	Metrics
k6	HTTP/gRPC APIs, scripted scenarios	req duration, error rate, throughput, VUs

k6 is the only supported load tool.

Integration Pattern

Load tools run as background activities alongside fault injection:

1. Start load tool (background)
2. Wait for traffic to stabilize (~30s)
3. Inject fault (chaos action)
4. Observe system under load + fault
5. Stop load tool
6. Collect load metrics

k6 Quick Start

# 1. Write a k6 script
cat > load/api-test.js << 'EOF'
import http from 'k6/http';
import { check } from 'k6';

export default function () {
    const res = http.get('http://localhost:8080/api/health');
    check(res, { 'status 200': (r) => r.status === 200 });
}
EOF

# 2. Reference it in your experiment
tumult run experiment-with-load.toon

Key Environment Variables

k6

Variable	Default	Description
TUMULT_K6_SCRIPT	(required)	Path to k6 test script
TUMULT_K6_VUS	10	Virtual users
TUMULT_K6_DURATION	30s	Test duration
TUMULT_K6_BINARY	k6	k6 binary path
TUMULT_OTEL_ENDPOINT	(none)	OTLP endpoint for trace correlation