A media company is migrating its on-premises legacy Hadoop cluster with its associated data processing scripts and workflow to an Amazon EMR environment running the latest Hadoop release. The…

Question

Options

• ADeploy the existing Oracle Java Archive as a custom bootstrap action and run the job on the EMR
• BCompile the Java program for the desired Hadoop version and run it using a CUSTOM_JAR step
• CSubmit the Java program as an Apache Hive or Apache Spark step for the EMR cluster.
• DUse SSH to connect the master node of the EMR cluster and submit the Java program using the

• A
  13% (4)
• B
  6% (2)
• C
  3% (1)
• D
  78% (25)

Explanation

The requirement is to reuse existing Java code written for an on-premises Hadoop cluster without significant changes. SSH-ing into the EMR master node and submitting the Java program via the hadoop jar command (hadoop jar myjob.jar ...) is the most direct reuse path - it mirrors exactly how jobs were submitted on the on-premises cluster and requires no modification to the Java code. Option B (CUSTOM_JAR step) is a valid EMR approach but requires packaging the job as an EMR step and potentially recompiling it for the specific Hadoop version bundled in EMR, adding friction. Option A (deploying as a bootstrap action) is incorrect because bootstrap actions run during cluster setup, not as on-demand processing jobs. Option C (submitting as Hive or Spark) would require rewriting the Java MapReduce code in a different framework.

No community discussion yet for this question.