I hope this is the right forum– I apologize if not.
I attended the YARN deep dive presentation yesterday, and the question, “what is the best/easiest development environment for developing and debugging YARN applications, and how can that be set up,” was answered during the Q&A portion.
The answer was, “it would be easiest to use the sandbox.”
My question is, in what context would we use the sandbox? Do we install eclipse and do our Application Master development on the sandbox VM itself, or do we install eclipse and the source on the host machine and use the remote debugging capabilities to attach to the application manager running on the sandbox (presumably using the HADOOP_OPTS flags)? Or did I misunderstand the answer and is some other strategy recommended?
If this is the proper strategy, what is the right Maven dependency to add to my Eclipse projects to make them aware of the sandbox version of HDP and how do I make Eclipse find the right source code version for the sandbox, so that I can step into Hadoop library source code?