Splunk has 3 main ways to work with Hadoop:
1) Hadoop Connect move files, based on your search to HDFS. And can copy files from HDFS to Splunk Indexers.
2) Hunk (aka Splunk Analytics for Hadoop) does not move files to and from Splunk, it generates Hadoop jobs (MapReduce jobs) and process the data inside Hadoop. With Hunk all you see are the results of your MapReduce job.
3) Hadoop Data Roll archive the Splunk raw data from the indexers to HDFS.