Hi, we recently switched from FlinkKafkaConsumer t...
# troubleshooting
d
Hi, we recently switched from FlinkKafkaConsumer to KafkaSource and we believe we've been hit by something like https://issues.apache.org/jira/browse/KAFKA-4682 (more details in thread).
• We run an old version of kafka 1.0 • We have topics that can go idle for months • We had the KafkaSource configured with
OffsetsInitializer.committedOffsets()
. The job is supposed to start from a savepoint or from bootstrap state with kafka offsets preset.
The job after running for several days did fail then attempted a restart but failed and entered a crashloop with
org.apache.kafka.clients.consumer.NoOffsetForPartitionException: Undefined offset with no reset policy for partitions: [eqiad.mediawiki.page-delete-0]
A quick fix with
OffsetsInitializer.committedOffsets(OffsetResetStrategy.LATEST)
unblocked the job. (But this is not exactly what we want, we would have preferred to fail if attempting to start the job from a bootstrap state without offsets preset).
And finally my question: with
OffsetResetStrategy.LATEST
on an idle stream (idle long enough such that offsets are no longer available from the broker) are there chances that some messages might be lost when the source wakes up?