hachikuji commented on pull request #11681: URL: https://github.com/apache/kafka/pull/11681#issuecomment-1042520005
> it doesn't work to put the admin client creating after cluster.waitForReadyBrokers(), because the waitForReadyBrokers only wait for all brokers registered and unfenced (ref: [here](https://github.com/apache/kafka/blob/trunk/core/src/test/java/kafka/testkit/KafkaClusterTestKit.java#L375)), but the metadata cache in broker might have not updated the cluster metadata. I think we need to explicitly wait for all brokers up via describeCluster and then create adminClient for testing. Hmm, I wonder if this is right. I checked the `QuorumController` logic and unfencing implies that the broker has caught up to the current end of the log. At least it implies that the offset reported in the heartbeat request is up to date. From what I can tell, there might still be a race between the time that this offset gets updated in `BrokerMetadataListener` and the time that the latest entries have been published to `BrokerMetadataPublisher`. Is that delay enough to explain the issue? -- 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
