Currently, the Spark component in EMR supports Spark 2.0.2, 2.2.1, and 2.3.2. The software environment is as follows:

• By default, Spark is installed on a master node.
• After the log-in, run the su hadoop command then you can switch to the Hadoop user.
• The Spark software path is /usr/local/service/spark.
• The relevant logs are stored in /data/emr.

For more information, please see the community documentation. Here is mainly about how to access COS through Spark.