Apache Spark uses Hadoop underneath and is doing very well. We use it in our production server for all of our machine learning algorithms
Not at all. The ecosystem has matured significantly and it's a crucial part of data infrastructure at pretty much every major tech company I'm aware of. Definitely not as hyped as it was 6-8 years ago, but as far as I'm concerned that's a good thing.
Now's probably the time to get into Hadoop. The hype train is gone, but all the companies that follow the hype train now have lots of Hadoop infrastructure that needs maintaining.
The hype it had few years ago was taken over by Machine Learning and AI fields in job perspective. The technology however is still good.