I have the following setup:
- docker container application sitting in K8s pod(s) exposed internally via TCP service
- Ingress Nginx routing incoming requests to this service
This setup works perfectly fine - I can call the application by accessing /app/<endpoint+params> for example via command line or a browser.
The problem is when I want to use our browser-based orchestration app based on Java. Making API requests from it results in occasional Connection Reset error right after the app returns 200 (OK). I can confirm the app is finishing the request with no issue, but the orchestrator seems not being able to confirm it. This problem occurs only for requests that take longer than a few minutes.
I tried adding some annotations to my ingress configuration .yaml file but nothing seems to work. I have gone through plenty of stackoverflow issues and I kind of understood it is related to closed socket connection that the orchestrator want to access again however I'm not sure what and where I should add to my setup to prevent this socket to close.
I will appreciate your help guys, thanks!