KafkaProducer: Consistent region reset can trigger addtional reset
Opened this issue · 2 comments
Deleted user commented
On consistent region reset, the KafkaProducer is forcibly closed. When transactions are used, an on-going transaction is aborted before.
Following exceptions due to this handling can be observed:
17 Dec 2019 ... onRecordProduceException:130] - Producer record 170 could finally not be produced for topic partition 'mySinglePartitionCCTopic-0' with 1 producer generations: org.apache.kafka.common.KafkaException: Failing batch since transaction was aborted
or
17 Dec 2019 ... onRecordProduceException:126] - Producer record 19 could not be produced for topic partition 'mySinglePartitionCCTopic-0' with 1 producer generations: org.apache.kafka.common.KafkaException: Producer is closed forcefully.
These exceptions are handled like all Kafka related exceptions are handled when in consistent region: A consistent region reset is initiated.
That's why a CR reset may be followed by another CR reset.
Expected behaviour: Exceptions caused by reset handling must be suppressed.
Deleted user commented
This unwelcome behavior has been introduced with implementation of the Producer's status port in v2.2.0
Deleted user commented
resolved in release 3.0.1