I’ve been watching two lectures from UC Berkeley’s CS61A course, on MapReduce in general and (in the second lecture) Hadoop specifically. If you’re looking for a good, deep kick-starter on both the concepts and the implementation, I’d recommend them. They’re a little dry (hey, they are university CS lectures), and you’ll need to focus on them as they move relatively quickly, but they reward the concentration with, I think, a really good launching point for playing with Hadoop. The lectures are up on YouTube here and here, but note there is a download link so you can watch them offline.