Skip to content

Load Test (JMeter)

Goal: Verify the stability and observability of the ingress path under target throughput; ensure no real devices are hit using Dry-Run and disabled consumers; cross-validate via RabbitMQ and SkyWalking metrics.

1. Test Scope & Path

  • Path: Gateway → ruoyi-robot → MQ
  • Risk control:
  • Dry-Run: all requests carry X-Dry-Run: true; the gateway returns 403 on real branches.
  • Disable consumers: downstream consumer process/thread count = 0; only validate ingress writes and throughput.
  • Peripheral isolation: tasks are mock no-ops; no actual device actions are triggered.

2. Traffic Model & Method


  • Target 100 QPS for 3 minutes; use Constant Throughput Timer to control global rate.
  • Assertions/post-processors: keep necessary fields (request headers, injected traceId, HTTP status, etc.).

3. Results · Overview



  • Error rate 0%, throughput stable; P95/P99 latency within expectations.

4. Results · Details

  • Response 200 OK; key headers/tags such as X-Which-Route, traceId are present for Trace ↔ Log cross-checks.

5. Message-Side Verification


  • Queue: robot.tempTask.q (example); requests are visible as enqueued, no consumption (consumers disabled), DLX/DLQ rules OK.

6. Gateway Service Metrics

  • Apdex, success rate, service load, instance load, endpoint load remain stable at 100 QPS, without abnormal spikes.

7. Business Service Metrics

  • Service level: average latency, P95/P99, success rate stay stable.
  • Instance level: load distribution is balanced, no obvious skew.

8. Endpoint Load & Latency


  • Top-N endpoint load/latency visualization.
  • Troubleshooting flow: locate slow/error endpoints → correlate traces → back-reference logs by traceId.