Lewisyixin commented on issue #12692: URL: https://github.com/apache/apisix/issues/12692#issuecomment-3741468680
HI @Baoyuantop In fact, yes, this does happen. In our company, the security team continuously monitors abnormal traffic and performs IP blocking in real time. When the request rate becomes too high, we have encountered this issue. In addition, I also tested the scenario of modifying upstreams. Under high QPS, changing upstream nodes once per second can similarly impact the latency of normal traffic. I believe high-frequency upstream changes are quite common in real-world scenarios, such as rolling deployments or batch traffic draining of upstream nodes, especially when the upstream nodes are pods. The figure below shows a load test scenario with 20,000 QPS, while simultaneously changing the weight of one upstream node at a rate of 2 QPS, which means that when KVMs or pods are deployed in batches, it can have a certain impact on existing traffic. <img width="665" height="297" alt="Image" src="https://github.com/user-attachments/assets/0d0486e8-02e6-4ec4-9ffe-1cb6aaccd81d"; /> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
