Hi All, I understand that we get a NotLeaderForPartitionException when the client tries to publish a message to a server which is not the Leader for a given partition. I also understand that once leadership does change the client will learn that and start publishing messages to the right server. I am however seeing that sometimes even after leadership has changed the client continues to send messages to the wrong server resulting in it continuously getting the NotLeaderForPartitionException.
I am not sure how to reproduce this problem because we have only seen this sporadically in certain production deployments. When we get into this situation one step which usually works is to restart zookeeper, kafka and the client. Obviously this is not acceptable. Is there a workaround that I could implement which would help to remediate this situation. For example would re-creating the KafkaProducer help in this scenario? Thanks, -Moiz
