Kafka(2.5.0) and Kafka consumer(io.projectreactor.kafka:reactor-kafka:1.3.3).
On the production environment unexpectedly our consumer started read from beginning. What the reason of such behavior could be? How better can I diagnose the exact reason?
The consumer group wasn't changed. Nobody could reset offset by hand. auto.offset.reset = 'earliest' but we commited offsets a lot before as it happened.
Kafka cluster consist of 3 nodes.
The kafka consumer was developed using exactly once delivery semantics ( KafkaReceiver.create(...).receiveExactlyOnce(...) ). Topic configuration: cleanup.policy=compact
On the node where pod with consumer worked, there were some network issues but I don't understand how it could make the consumer reset offset.