Troubleshooting no more container networking after rebooting Sequoia?

[natevw]

I had been using containers for several weeks and nearly come to rely on it for some active projects. This morning I rebooted to install the update to Sequoia 15.7.5 and now my containers have no network access. Updated my install of container itself from 0.9.0 to 0.10.0 with no change.

I am able to pull images just fine, but anything within the image build or the image itself has no access to networking. E.g. container run -it --rm alpine grabs the image and starts but then:

• ping google.com fails after a few seconds with "ping: bad address 'google.com'"
• ping 8.8.8.8 just hangs after logging the outgoing "PING 8.8.8.8 (8.8.8.8): 56 data bytes"

Running alpine's ip route shows simply:

default via 192.168.64.1 dev eth0 
192.168.64.0/24 dev eth0 scope link  src 192.168.64.7 

Within the container if I do a nc -l 5000 and then from the host echo "HELLO" | nc 192.168.64.7 5000 (using the container's IP address) just hangs. Pinging the container results in "Request timeout" but I don't recall if that worked before — pinging the buildkit container's IP also fails even though the general build system seems responsive.

This has so far been true for every container I've tried with a variety of images (alpine, debian, etc.) and like I mentioned affects both the build portion and the running portion.

I don't remember nor do my notes show having to do anything beyond container system start to get things working before. Any idea what's going on or what sort of troubleshooting I could do next?

[natevw]

Aha! Turns out this was exactly what was already documented over in https://github.com/apple/container/blob/main/docs/technical-overview.md#container-ip-addresses — my bridge100 network had switched to "inet 192.168.65.1" and I needed to update the network.subnet property accordingly, and restart the container system. (Edit: also had to container builder rm -f since that had kept holding on to the old nameserver in its /etc/resolv.conf.)

Back in business now! Thanks to the project maintainers for the good documentation here.