We currently use xinetd to spawn a process per tcp socket connection. I have been able to convert this over so that a container is spawned per connection. The remaining problem is that I’m not sure how to eliminate xinetd from the host os. I would like a load balancing to spawn the containers per connection That way the ingress network could balance these containers across nodes. (Our application requires one process/container per connection. We can’t just put a thread pool in each process and balance across those.)
Any recommendations? Does anything like this exist? I would be happy on developing contributing such a feature. If so would people have recommendations on which areas of the code to start looking at?