No more published port is mapped

Hi there,

We have a very strange problem today. We used to use Rancher 1.1.3 then 1.2.0-pre3 without any problem. Last Thursday we upgraded to the brand new 1.2.0. And then … Boooom!

Everything seems to work (I have my stacks, my hosts …Etc) but no one of my services is reachable. I check the port mapping and the hosting rule (set on “managed”). Everything is started and running.

But if I log on and random host and put a docker ps command I have this :

None of the ports I specified in my stacks definitions appears in the “PORT” column. It explains why my services are’nt reachable but I don’t understand why I now have this behaviour I never had before (My docker-compose are the same than before)

Please help

Thank you

The list docker reported was never actually used or necessarily accurate since we allow you to change the host port after the container is running and there is no way to update what Docker thinks it is. In 1.2 we set the networkMode to none for managed containers which causes it to hide them entirely. But the actual directing of the connections to a container is done with iptabes, as it was in 1.1.

I get that the list didn’t say anything.

Yet, when I mapped e.g. the port was only mapped to the local host.
After updating Rancher to v1.2.0 I can access (where this would be my external IP) from the outside and can’t bind to it from within the host.

I used to map Prometheus from it’s container’s 9090 to and then create a SSH tunnel to my local machine to access it, to not expose it to the outside.
This is currently broken, as I can access Prometheus from the outside but not from the host on itself.

Is there a better way to accomplish want I want to do, or did I oversee something, or just a bug?