how to determine where the Kafka consumption is made on Spark

Question

We tried two MlLib transformers for kafka consuming: one using Structured Batch Query Stream like here, and one using normal Kafka consumers. The normal consumers read into a li