I have a .NET Core application in the linux container running on the linux host - 2 replicas. Additionally I have a container with nginx server as a load balancer (also on a linux). This load balancer has a single upstream - just a name of the .NET core service. And this setup works - nginx pass calls to the http://serviceName and the Docker's Dnsrr direct those once to the first replica and once to the second replica. However when I am trying to ping the same address (http://serviceName) from the windows container, I am always redirected by the dns to the same IP (or my console app which is trying to get response from http://serviceName). If I create two replicas of the windows container then those containers pings/do requests to different replicas of my service. I do not understand why linux ngnix using the same address works correctly and windows don't. I suppose that windows cached the ip or something but I am not sure. Is there any solution to this problem?