multiple outputs considers hdfs sink to be a dataset and tries to validate the hdfs path as a dataset name.
HDFSSink can now be used alongside other sinks in a Hydrator pipeline.
I don't think we need to be aliasing any of the sinks, but instead just name them as their configured 'referenceName'. I believe that would solve this issue.
We are planning to override/suffix (in ETLMapReduce code) the alias set by the user to uniquely identify the source/sink.