A Kubernetes migration story

Posted Apr 6, 2017 17:29 UTC (Thu) by nix (subscriber, #2304)
Parent article: A Kubernetes migration story

NAT is evil. Why would anyone NAT the connections between their containers and their backend database? That's just *asking* for random, unreproducible, almost-undebuggable networking-related failures.

A Kubernetes migration story

Posted Apr 6, 2017 17:40 UTC (Thu) by ibukanov (subscriber, #3942) [Link]

I second this. I had an application that used UDP protocol. In a docker container it died randomly after a week of smooth running. It turned out it was caused by NAT between the host and the container that docker deployed. After I switched the container to use the host network the app run for months without any troubles.