Hi All,

I am using Standalone Spark.

I am using dynamic memory allocation. Despite giving max executors, min executors and initial executors, my  streaming job is taking all executors available in the cluster. Could anyone please suggest what can be wrong here?

Please note source is Kafka.

I feel this can be avoided by setting max cores per application. But why this is happening if max executors is set and also what can be other best ways to avoid that.