Hello. We are having trouble understanding whether a Flink (1.14) job with a Kafka sink configured with EoS is able to recover in the case where the Kafka transaction expires. The documentation states that data loss occurs in this scenario, but it’s not clear whether the job is able to recover (with this data loss) of whether it aborts. This happened to us, where after a transaction timeout the job was unable to continue with ProducerFencedException. We are trying to understand if this is the root cause or if we’re still missing something else.

Hi, do you have any insights about this? I'm experiencing the same exception @Marcelo Miranda