On my machine, reading some cooling devices cur_state throws a EINVAL error, so none of the thermal or cooling device metrics are gathered at all. I feel like it may need some error handling. I would happily push a PR with that, something similar to #705 or #300. If that sounds reasonable, I need some advice from the maintainers about what is the best way to handle errors — provide a default value for the metric that causes an error, skip the device with such a metric at all, or something else.