From what I understand, the Eclipse plugin has not been maintained as new versions of Hadoop have been released. It appears that the command to start the DataNode is missing a required argument: Usage: java DataNode [­regular | ­rollback]­ regular : Normal DataNode startup (default).­ rollback : Rollback a standard or rolling upgrade. Refer to HDFS documentation for the difference between standard and rolling upgrades. The Apache HDT (Hadoop Development Tools) project had plans to fix this, but unfortunately, it has been retired due to lack of contributions. http://hdt.incubator.apache.org/ One option to consider would be to ditch the Eclipse Plugin and leverage "mini clusters" to provide a similar development experience, but without the need to connect to an external cluster or leverage the Eclipse plugin. https://wiki.apache.org/hadoop/HowToDevelopUnitTests Another option would be to leverage the hadoop-mini-clusters project that I maintain. It simplifies the use of mini clusters by wrapping them in a common Builder pattern. https://github.com/sakserv/hadoop-mini-clusters Hope that helps. ... View more

@Bhupendra Mishra Depending on your hardware availability for the POC, I would also look at just doing the POC in the Cloud (e.g. MSFT Azure, AWS, GCP). You can leverage Cloudbreak to quickly deploy a fully fledge distributed cluster running Spark, Yarn, the whole nine yards, in the cloud in a matter of minutes. Here is the documentation on how to do so: Cloudbreak Overview - http://hortonworks.com/hadoop/cloudbreak/ Cloudbreak Docs - http://docs.hortonworks.com/HDPDocuments/HDP2/HDP-... ... View more