This message was deleted.

Turns out, the managednodegroup has its own field for version. Updating that along with the cluster version took care of it