Spark Streaming

Hey there,

I'm using this spark streaming job which integrated with Kafka (and manage its offsets commitions at Kafka itself),

The problem is when I have a failure I want to repeat the work on  those offset ranges (that something went wrong with them) , therefore I catch the exception and NOT commit (with commitAsync) this range.

However I notice it keeps proceeding (without any commit made).

moreover I removed later all the commitAsync calls and I the stream keep proceeding!

I guess there might be any inner cache or something that helps the streaming job to consume the entries from Kafka.

Could you please advice?