Recommended way for CNI plugin logging

[MrHohn]

Hello CNI maintainers,

I found a similar question that was closed last year: https://github.com/containernetworking/plugins/issues/502, but wanted to ask this in a slightly generic way.

During the process of debugging the CNI plugins in our kubernetes environment (either the commonly used plugins or a few new ones), we found it not so intuitive that the plugin logs aren't always available. When the CNI plugin errors out, there might be more logs available. But in a working case, we don't seem to see any logs during the process (e.g. add network for a pod).

The question being - is there a recommended way for CNI plugin logging (e.g is printing to stdout/stderr sufficient)? Should we expect no logs from the plugin unless the execution fails? Or is there a way to surface informational logs from the plugin if needed?

(I'm also curious if this varies between different calling processes - e.g. dockershim, containerd and others).

Thanks!

cc @sypakine