Load, Stress and Soak Testing — Resilience Patterns

Load tests verify performance at expected peak; stress tests push beyond to find the breaking point and how it fails (gracefully with 503s, or collapsing); soak tests run for hours to reveal leaks and slow degradation; spike tests apply sudden bursts. Use realistic traffic mixes and data, open-model arrival rates, production-like environments, and watch resource saturation and dependency limits, not just response times.

Test types

Test	Question answered	Duration
Load	Do we meet SLOs at expected peak?	30–60 min
Stress	Where and how do we break?	Ramp until failure
Soak	Do we leak or degrade over time?	Hours–days
Spike	Do we survive sudden bursts?	Minutes
Failover under load	Does resilience work when busy?	During load test

k6 (stress to breaking point)

import http from "k6/http";
export const options = {
  scenarios: { stress: { executor: "ramping-arrival-rate", startRate: 200, timeUnit: "1s", preAllocatedVUs: 2000,
    stages: [{ target: 2000, duration: "10m" }, { target: 5000, duration: "10m" }, { target: 0, duration: "2m" }] } },
  thresholds: { "http_req_failed{status:503}": ["rate<0.5"], http_req_duration: ["p(99)<1500"] },
};
export default () => http.get(`${__ENV.BASE}/api/products?limit=20`);

Common pitfalls

Testing only the happy path at average load.
Closed-model generators hiding latency (coordinated omission).
Ignoring third-party limits in tests (use stubs).

Key points

Load, stress, soak, spike
Observe failure behavior
Realistic traffic and arrival rates

Test types

Common pitfalls

Key points

Practice questions