[
https://issues.apache.org/jira/browse/KAFKA-7108?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16525286#comment-16525286
]
Guozhang Wang commented on KAFKA-7108:
--------------------------------------
I'm in favor of not extending the production.exception.handler for this failure
scenario, because if exactly-once is turned on, then we should not give users
options to handle such errors but always fail and stop the world immediately to
preserve the semantics (we should probably try-catch and stop in a more
graceful way than today though). For sending records, if they gets thrown from
the producer, users can still decide whether to retry / ignore based on their
logic that does not violate EOS.
> "Exactly-once" stream breaks production exception handler contract
> ------------------------------------------------------------------
>
> Key: KAFKA-7108
> URL: https://issues.apache.org/jira/browse/KAFKA-7108
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Affects Versions: 1.1.0
> Reporter: Anna O
> Priority: Major
> Labels: exactly-once
>
> I have a stream configured with "default.production.exception.handler" that
> is supposed to log the error and continue. When I set "processing.guarantee"
> to "exactly_once" it appeared that retryable NotEnoughReplicasException that
> passed the production exception handler was rethrown by the
> TransactionManager wrapped with KafkaException and terminated the stream
> thread:
> _org.apache.kafka.common.KafkaException: Cannot execute transactional method
> because we are in an error stateat
> org.apache.kafka.clients.producer.internals.TransactionManager.maybeFailWithError(TransactionManager.java:784)
> ~[kafka-clients-1.1.0.jar:?]_
> _at
> org.apache.kafka.clients.producer.internals.TransactionManager.sendOffsetsToTransaction(TransactionManager.java:250)
> ~[kafka-clients-1.1.0.jar:?]_
> _at
> org.apache.kafka.clients.producer.KafkaProducer.sendOffsetsToTransaction(KafkaProducer.java:617)
> ~[kafka-clients-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamTask.commitOffsets(StreamTask.java:357)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamTask.access$000(StreamTask.java:53)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamTask$1.run(StreamTask.java:316)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamsMetricsImpl.measureLatencyNs(StreamsMetricsImpl.java:208)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamTask.commit(StreamTask.java:307)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamTask.commit(StreamTask.java:297)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.AssignedTasks$1.apply(AssignedTasks.java:67)
> ~[kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.AssignedTasks.applyToRunningTasks(AssignedTasks.java:357)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.AssignedTasks.commit(AssignedTasks.java:347)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.TaskManager.commitAll(TaskManager.java:403)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamThread.maybeCommit(StreamThread.java:994)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamThread.runOnce(StreamThread.java:811)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamThread.runLoop(StreamThread.java:750)
> [kafka-streams-1.1.0.jar:?]_
> _at
> org.apache.kafka.streams.processor.internals.StreamThread.run(StreamThread.java:720)
> [kafka-streams-1.1.0.jar:?]_
> _Caused by: org.apache.kafka.common.errors.NotEnoughReplicasException:
> Messages are rejected since there are fewer in-sync replicas than required._
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
