I am doing a vanilla install of the latest free5gc and I noticed that three pods look like they are connected to the container network, but they are not. I can ping between all the free5gc pods except the AMF, SMF and UPF. Each of these pods gets an IP address but if I exec into one of these pods I can’t ping anything else in the cluster. All the pods say they are running but I had to disable the wait-nrf container to get the AMF to go ready. I discovered after the fact that the wait-nrf never completes because eth0 on the amf-pod is not actually connected to anything (it seems). Has anyone seen this?
Ok so I have a work around… It turns out that the latest version of free5gc has an extra route in the AMF and SMF that causes packet traffic that should be going to the cluster network to go to the N2 network instead. I removed the route and all the cluster services are talking. I’m not sure why the extra route was added but I assume it was required to send some packets to a network beyond the directly connected N2.