FWIW, attached_to is not meant to be set by puppet (or other cfg mgmt tool) but it's constructed from virtual names (e.g. eth0.1 -> eth0), see https://github.com/brandonweeks/foreman/blob/56671b5f0cfa6dd185cb75d40445300e3b6668f5/app/services/fact_parser.rb#L126 for more details. Anyway, thanks for the patch.

Thanks for the context! Knowing that it's obvious that the problem we were running into is our our virtual interface is named "bond0," do you think there any other improvements that should be made for handling virtual interfaces that don't have the suffix pattern?

I think your patch fixed it for all interfaces, if I understood it correctly it makes attached_to nil instead of empty string. The other way to fix it would be to just fix the validation to accept empty strings in this attribute for bonds and bridges but this seems to work just fine :-)