This message was deleted.

@brief-dog-58395 can you provide a few more details while look into this behavior/attempt a repro? 1. Your Rancher Desktop version 2. macOS/Windows versions that you tested on 3. On Windows, are you using the new tunneling network, or the default legacy stack?

To test the issue, in one terminal run:

docker run --rm -it -p 4444:4444 alpine sh -c 'apk add nmap-ncat; ncat -k -l 4444'

to start server listening on port 4444 then in another terminal run do:

docker run --rm -it alpine sh
 apk add nmap-ncat
 ncat -v -z host.docker.internal 4444
 ncat -v -z 192.168.5.2 4444

host.docker.internal takes 1 second 192.168.5.2 takes 0 second.

Thanks for the info, I can confirm that there isn't a delay on Windows 11 with the networking tunnel enabled (Preferences => WSL => Networking => Enable Network Tunnel) Can you give that a try and let us know if you experience the same?

I will, thanks. Anything on Mac side that can be done? I'm using extra_hosts to add static host in docker-compose as workaround

Unfortunately, not that I'm aware of.. I understand that we will need to wait for Lima to implement

user-v2

networking and that should behave similarly to the Network Tunnel feature in Windows.

Thank you for looking into it!

So, r we saying that "network tunneling" is the way to go for performance reasons? So for what use cases do I NOT want to use networking tunneling?