Problem determining host network device about blockade HOT 5 CLOSED

A note, Ubuntu images don't seem to have the ip command available anymore. I tried with ubuntu:16:04. Use ubuntu:14.04.

from blockade.

I haven't seen that error before, but it can come from a few conditions in this method: https://github.com/dcm-oss/blockade/blob/6c79655d6550962c5385becb053270a520a93202/blockade/net.py#L67-L96

This method attempts to find the network interface device name for a container, and this is an area that has changed a few times, with different docker releases and container drivers. There must be some case we are not handling. Can you please provide:

• the output of docker version
• then launch a container (of the image you are using with blockade), and provide the output of the following two commands (substituting in your container ID):
  • docker exec $CONTAINER_ID ip link show eth0
  • ip link

from blockade.

It seems the issue is that if there is anything that causes the container to hit an error on run, this is the error that manifests. In my case, I had a typo in the volume I was attempting to mount.

So I think the problem here is that the error is misleading; or rather, some set of problems leading up to this point are not caught.

from blockade.

Interesting.. I'm surprised the run didn't fail earlier. I agree the error messages need to be better for this case. I'll keep this issue open as a reminder to address that.

Thanks for reporting this!

from blockade.

Closing this issue. In 0.3.1, blockade has a couple of fixes that address problems discussed here:

1. Vastly improved error handling around finding a container's interface
2. ip is no longer required to be installed in containers. Instead Blockade directly reads the /sys FS to find the needed information from the container.

from blockade.