I have a Kafka cluster with 40 partitions and I published a few million messages. I have a pool of consumers and for the most part the consumer lag < 100.
After my experiment, I stopped publishing any new message and left consumers to drain.
But, I don't see consumer lag come down to zero. It is struck between 0-200 for different partitions. Any reason they wont' be zero?
I am using confluent Kafka python for producer and consumer.