Major Description
When Java consumer encounters a corrupt record on a partition it reads from, it throws:
org.apache.kafka.common.KafkaException: Received exception when fetching the next record from XYZ. If needed, please seek past the record to continue consumption.
at org.apache.kafka.clients.consumer.internals.Fetcher$PartitionRecords.fetchRecords(Fetcher.java:1125)
at org.apache.kafka.clients.consumer.internals.Fetcher$PartitionRecords.access$1400(Fetcher.java:993)
at org.apache.kafka.clients.consumer.internals.Fetcher.fetchRecords(Fetcher.java:527)
at org.apache.kafka.clients.consumer.internals.Fetcher.fetchedRecords(Fetcher.java:488)
at org.apache.kafka.clients.consumer.KafkaConsumer.pollOnce(KafkaConsumer.java:1155)
at org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1115)
(...)
Caused by: org.apache.kafka.common.errors.CorruptRecordException: Record size is less than the minimum record overhead (14)
or:
java.lang.IllegalStateException: Unexpected error code 2 while fetching data
at org.apache.kafka.clients.consumer.internals.Fetcher.parseCompletedFetch(Fetcher.java:936)
at org.apache.kafka.clients.consumer.internals.Fetcher.fetchedRecords(Fetcher.java:485)
at org.apache.kafka.clients.consumer.KafkaConsumer.pollOnce(KafkaConsumer.java:1155)
at org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1115)
(...)
1. Could you consider throwing CorruptRecordException from parseCompletedFetch() when error == Errors.CORRUPT_MESSAGE?
2. Seeking past the corrupt record means losing data. I've noticed that the record is often correct on a follower ISR, and manual change of the partition leader to the follower node solves the issue in case partition is used by a single consumer group. Couldn't Kafka server discover such situations and recover corrupt records from logs available on other ISRs somehow?