Chaos Testing for QA Engineers: Break Your System Before Users Do

Your system works when everything is perfect. But production is never perfect. Chaos testing deliberately breaks things to verify your system recovers gracefully.

🎭 Want to master this with real projects? Join the Playwright Automation Mastery course at The Testing Academy.

Chaos Engineering for QA Engineers

Traditional QA: “Does the system work?” Chaos testing: “Does the system work when things go wrong?”

Chaos Experiments QA Can Run

Experiment	What Breaks	Expected Behavior
Kill a service	Service dependency	Graceful degradation, error message
Network latency (3s)	Response time	Loading indicator, no timeout crash
Database connection drop	Data layer	Retry logic, queue messages
Disk full	Storage	Alert fires, writes rejected cleanly
DNS failure	External services	Cached response or fallback
CPU spike (100%)	Compute	Auto-scaling triggers, no crash

Chaos Testing with Playwright

test('app handles API timeout gracefully', async ({ page }) => {
  // Simulate 10-second delay on payment API
  await page.route('**/api/payment/**', async route => {
    await new Promise(r => setTimeout(r, 10000));
    route.abort('timedout');
  });

  await page.goto('/checkout');
  await page.getByRole('button', { name: 'Pay' }).click();

  // Should show timeout message, not crash
  await expect(page.getByText('Payment service unavailable')).toBeVisible();
  await expect(page.getByRole('button', { name: 'Retry' })).toBeVisible();
});

test('app handles 503 from dependency', async ({ page }) => {
  await page.route('**/api/inventory/**', route => {
    route.fulfill({ status: 503, body: 'Service Unavailable' });
  });

  await page.goto('/products');
  // Should show fallback, not blank page
  await expect(page.getByText('Inventory temporarily unavailable')).toBeVisible();
});

Steady State Hypothesis

// Before chaos: verify normal behavior
test.beforeEach(async ({ page }) => {
  await page.goto('/');
  await expect(page.getByText('Dashboard')).toBeVisible();
  // System is in known good state
});

// During chaos: inject failure
// After chaos: verify recovery
test('system recovers after service restart', async ({ page }) => {
  // Kill dependency (via API or Docker)
  await fetch('http://localhost:9090/chaos/kill-service/inventory');

  // Verify degraded behavior
  await page.goto('/products');
  await expect(page.getByText('Inventory unavailable')).toBeVisible();

  // Restart service
  await fetch('http://localhost:9090/chaos/restart-service/inventory');

  // Verify recovery
  await page.reload();
  await expect(page.getByText('products found')).toBeVisible();
});

Tools

• Chaos Monkey (Netflix): Randomly kills production instances
• Litmus (CNCF): Kubernetes chaos experiments
• Gremlin: Enterprise chaos platform
• Toxiproxy (Shopify): Simulate network conditions
• Playwright page.route(): Application-level chaos (timeout, 503, slow responses)