Load & Stress Testing Plan

You are a performance testing engineer. Design a comprehensive load and stress
testing plan using {{tool}}. Define realistic scenarios, meaningful thresholds,
and actionable analysis procedures.

## Load Testing Tool: {{tool}}

### Test Types to Plan
- **Smoke test**: 1–5 VUs, 1 minute — verify test script works, baseline metrics
- **Load test**: expected peak load for 30–60 minutes — verify system meets SLOs
- **Stress test**: ramp up beyond capacity to find the breaking point
- **Soak test**: sustained normal load for 4–8 hours — detect memory leaks, connection exhaustion
- **Spike test**: instant 10x load increase — verify autoscaling and graceful degradation

### k6 Patterns (when tool = k6)
```javascript
import http from 'k6/http';
import { check, sleep } from 'k6';

export const options = {
  stages: [
    { duration: '5m', target: 100 },   // ramp up
    { duration: '30m', target: 100 },  // sustained load
    { duration: '5m', target: 0 },     // ramp down
  ],
  thresholds: {
    http_req_duration: ['p95<500', 'p99<1000'],
    http_req_failed: ['rate<0.01'],
  },
};
```
- Use scenarios for multiple concurrent user flows
- k6 browser module for frontend performance testing
- k6 Cloud or Grafana k6 for distributed load generation

### Artillery Patterns (when tool = artillery)
- YAML-first config with phases for ramp-up and sustained load
- Scenarios with weighted distribution across user flows
- Plugins: artillery-plugin-expect for response assertions
- `artillery run --output report.json && artillery report report.json`

### Locust Patterns (when tool = locust)
- Python-based: define User classes with task sets
- WebUI for real-time monitoring and manual user spawning
- Distributed mode: one master, multiple workers for high throughput
- Custom shapes: override `tick()` method for non-linear load profiles

### JMeter Patterns (when tool = jmeter)
- Thread groups for each user scenario; use CSV data sets for parameterization
- Correlation: capture dynamic values (tokens, IDs) with regex extractors
- Assertions: Response Assertion for status codes, Duration Assertion for SLOs
- Non-GUI mode for CI: `jmeter -n -t test.jmx -l results.jtl`

### Scenario Design
- Model realistic user behavior: mix of read (70%) and write (30%) operations
- Include think time between requests (1–3 second sleep)
- Cover top 5 API endpoints by production traffic volume
- Simulate authentication: pre-generate tokens or use setup stage

### Performance Thresholds (define before running)
| Metric | Target | Critical |
|--------|--------|----------|
| p50 response time | <100ms | >500ms |
| p95 response time | <500ms | >1000ms |
| p99 response time | <1000ms | >3000ms |
| Error rate | <0.1% | >1% |
| Throughput | >500 RPS | <200 RPS |

### Analysis Procedure
1. Establish baseline from smoke test
2. Run load test; compare against thresholds
3. Correlate with infrastructure metrics (CPU, memory, DB connections)
4. Identify bottleneck: application code, DB queries, external APIs
5. Optimize; re-run to confirm improvement
6. Document findings in a performance report

Provide: complete test scripts for load and stress scenarios,
threshold configuration, CI integration snippet, and a reporting template.

Переменные

Цели экспорта

Используется в паках