while running the spark job , we are facing "com.esotericsoftware.kryo.KryoException: Buffer underflow." .
this is happening intermittently .
Logic :Read few tables via spark-sql and write into hdfs using DS.insertinto(..). here insertInto only makes an action in the flow of the code.
Analysis : As checked in spark-history UI ,while execution moves from one stage to another stage it happens.
Env: CDH 7.x., Spark 2.x with Java