the trouble with hadoop
Is once you get good at it most likely you can jump jobs to another company that will pay more.
The other trouble with hadoop is people wanting to use it for the wrong reasons, and not being able to utilize it effectively (e.g. writing batch jobs that are so poor they can only be processed by a single node, or an organization that wants to use hadoop so badly they just start using it even though they only have a GIGABYTE or two of data).
I had one clueless VP years ago tell me with a straight face he wanted to use HDFS to host VMware VMs.