horkasai commented on issue #12387: URL: https://github.com/apache/apisix/issues/12387#issuecomment-3022090492
**Description** Basically I have a WebSocket service that is upstreamed by APISIX, everything is working as expected until a node in my EKS cluster gets terminated/replaced by Karpenter. This node can be totally unrelated to my services and apisix, so long it is within the same cluster, the issue will occur. **The Issue** During a node termination/replacement, a handful of incoming requests will be fail to receive by my WebSocket service. Upon checking the logs, the connection was still connected between the client and the server. Moreover, using tcpdump, I can see there are a number of RST packets sent to my WebSocket endpoint from the APISIX pods, which could explains why my service was not receiving any requests from the client. I've tried removing APISIX and directly accessing my WebSocket service through AWS NLB endpoint, the issue did not occur at all. **Steps to Reproduce** 1. Have a WebSocket Service Upstreamed By Apisix 2. Send 4-5k requests per second to the service 3. Terminate a node in the same cluster (can be in a different node group to the WebSocket service) 4. Observe the service behaviour during the node termination process. -- 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: notifications-unsubscr...@apisix.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
