[
https://issues.apache.org/jira/browse/KAFKA-14750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17710259#comment-17710259
]
Sergei Morozov commented on KAFKA-14750:
----------------------------------------
[~sagarrao] thank you for digging into this and providing the details. I would
agree that a solution that will cause an overhead is not the best for this
corner case. I'm not well familiar with the Kafka internals, so I don't
understand why this issue is reproducible only during a mass topic deletion
(which makes it a corner case). As a general idea, if a client fails to
retrieve the position, and we know that at this point the topic may no longer
exist, could we just refresh the cache on the client side via the admin client
and only fail if the topic still exists?
> Sink connector fails if a topic matching its topics.regex gets deleted
> ----------------------------------------------------------------------
>
> Key: KAFKA-14750
> URL: https://issues.apache.org/jira/browse/KAFKA-14750
> Project: Kafka
> Issue Type: Bug
> Components: KafkaConnect
> Affects Versions: 3.3.1
> Reporter: Sergei Morozov
> Assignee: Sagar Rao
> Priority: Major
>
> Steps to reproduce:
> # In {{{}config/connect-standalone.properties{}}}, set:
> {code:bash}
> plugin.path=libs/connect-file-3.3.1.jar
> {code}
> # In {{{}config/connect-file-sink.properties{}}}, remove the {{topics=}} line
> and add this one:
> {code:bash}
> topics.regex=connect-test-.*
> {code}
> # Start zookeeper:
> {code:bash}
> bin/zookeeper-server-start.sh config/zookeeper.properties
> {code}
> # Start the brokers:
> {code:bash}
> bin/kafka-server-start.sh config/server.properties
> {code}
> # Start the file sink connector:
> {code:bash}
> bin/connect-standalone.sh config/connect-standalone.properties
> config/connect-file-sink.properties
> {code}
> # Create topics for the sink connector to subscribe to:
> {code:bash}
> for i in {0..2}; do
> for j in $(seq $(($i * 100)) $(( ($i + 1) * 100 - 1 ))); do
> bin/kafka-topics.sh \
> --bootstrap-server localhost:9092 \
> --create \
> --topic connect-test-$j
> done &
> done
> wait
> {code}
> # Wait until all the created topics are assigned to the connector. Check the
> number of partitions to be > 0 in the output of:
> {code:bash}
> bin/kafka-consumer-groups.sh \
> --bootstrap-server localhost:9092 \
> --group connect-local-file-sink \
> --describe --members
> {code}
> # Delete the created topics:
> {code:bash}
> for i in {0..2}; do
> for j in $(seq $(($i * 100)) $(( ($i + 1) * 100 - 1 ))); do
> bin/kafka-topics.sh \
> --bootstrap-server localhost:9092 \
> --delete \
> --topic connect-test-$j
> echo Deleted topic connect-test-$j.
> done &
> done
> wait
> {code}
> # Observe the connector fail with the following error:
> {quote}org.apache.kafka.common.errors.TimeoutException: Timeout of 60000ms
> expired before the position for partition connect-test-211-0 could be
> determined
> {quote}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
