FrancisGodinho opened a new pull request, #20307: URL: https://github.com/apache/kafka/pull/20307
## Ticket - https://issues.apache.org/jira/browse/KAFKA-17207 ## Summary This PR fixes the memory leak in `ConnectDistributedRequestTimeoutIntegrationTest#testRequestTimeouts`, where the producer hangs indefinitely after the config topic is deleted. Root Cause: The hang occurs inside the Kafka producer's internal NetworkClient.poll() loop: https://github.com/apache/kafka/blob/trunk/clients/src/main/java/org/apache/kafka/clients/NetworkClient.java#L645 Selector.poll(...) blocks while trying to reconnect to the broker and send metadata requests for the deleted config topic. Since the topic is gone and the MetadataRecoveryStrategy is set to NONE, the metadata updater continues retrying, and the producer does not timeout or fail fast. This causes test resources to hang indefinitely. ### Why not fix Kafka core? While this looks like a Kafka client bug at first, it is technically consistent with Kafka's design: the system is eventually consistent and assumes topics may become available again. Therefore, hanging here is expected behavior unless the producer is explicitly closed or the topic reappears. Fixing this behavior in the Kafka producer would require rethinking the metadata handling policy or tightening timeouts, which is risky and out of scope for this test. ### Fix: I restore the deleted topic at the end of the test to allow the producer to recover and shut down cleanly. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
