(docs) Add JavaDocs for org.apache.kafka.common.security.oauthbearer.secured

[justinrlee]

Single-line change to build.gradle to render javadocs for new org.apache.kafka.common.security.oauthbearer.secured package (part of KIP-768)

cc @junrao

[justinrlee]

(Also, I'm not sure if we'd want a similar PR to the 3.1 branch)

[junrao]

@justinrlee : Thanks for the PR. LGTM. Do you know why there are many test failures for java 17?

[justinrlee]

I'm really not sure, have never done this before. When I do the build locally, I get a bunch of what look like completely test failures. I'm not sure how my PR would cause those failures but I don't know enough to troubleshoot, tbh.

[junrao]

retest this please

[junrao]

test this please

[junrao]

ok to test

[junrao]

retest this please

[junrao]

@justinrlee : It doesn't seem other PRs have similar problems. Could you trigger another build by adding some minor changes to the PR (e.g. white space) to see if they are transient?

[justinrlee]

Just submitted two commits (add space, and then undo). Will monitor this.

[showuon]

The latest build results look good. The most failed tests also failed in trunk build. (ref: https://ci-builds.apache.org/job/Kafka/job/kafka/job/trunk/731/#showFailuresLink).

    Build / ARM / org.apache.kafka.streams.integration.NamedTopologyIntegrationTest.shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing
    Build / JDK 17 and Scala 2.13 / integration.kafka.server.FetchRequestBetweenDifferentIbpTest.testControllerNewIBP()
    Build / JDK 17 and Scala 2.13 / kafka.admin.LeaderElectionCommandTest.[1] Type=Raft, Name=testPreferredReplicaElection, Security=PLAINTEXT
    Build / JDK 17 and Scala 2.13 / org.apache.kafka.streams.integration.NamedTopologyIntegrationTest.shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing
    Build / JDK 11 and Scala 2.13 / org.apache.kafka.connect.integration.RebalanceSourceConnectorsIntegrationTest.testDeleteConnector
    Build / JDK 11 and Scala 2.13 / org.apache.kafka.streams.integration.NamedTopologyIntegrationTest.shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing
    Build / JDK 8 and Scala 2.12 / kafka.controller.ControllerIntegrationTest.testPartitionReassignmentToBrokerWithOfflineLogDir()
    Build / JDK 8 and Scala 2.12 / org.apache.kafka.streams.integration.NamedTopologyIntegrationTest.shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing
    Build / JDK 17 and Scala 2.13 / kafka.server.DynamicBrokerReconfigurationTest.testThreadPoolResize()
    Build / JDK 17 and Scala 2.13 / kafka.server.DynamicBrokerReconfigurationTest.testThreadPoolResize()
    Build / JDK 11 and Scala 2.13 / kafka.server.DynamicBrokerReconfigurationTest.testThreadPoolResize()
    Build / JDK 11 and Scala 2.13 / kafka.server.DynamicBrokerReconfigurationTest.testThreadPoolResize()

The most frequent flaky tests are under investigation and working on the fix:

1. shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing is fixing in this PR: MINOR: fix flaky shouldRemoveOneNamedTopologyWhileAnotherContinuesProcessing #11827
2. testThreadPoolResize is fixing in this PR: KAFKA-13672 Flaky test kafka.server.DynamicBrokerReconfigurationTest.testThreadPoolResize() [WIP] #11821
3. LeaderElectionCommandTest is fixing in this PR: KAFKA-8785: fix request timeout by waiting for metadata cache up-to-date #11681

The remaining are some flaky tests happened not so frequently.

PS. I've noticed the many flaky tests issue, and are working with the community to make it better. At least, the frequent failed tests should be fixed soon!

[junrao]

@justinrlee : Thanks for following up on this. LGTM

[junrao]

@justinrlee : It would be useful to submit a similar PR to 3.1 too. Thanks.

[ijuma]

Just to double check, @kirktrue this are intended to be public API?

[kirktrue]

Yes.

Users of the OAuth code will not necessarily code against the classes in this package, but they will need to specify the class names from this package in their config. So it's helpful that the documentation be included for reference for config options and a description of the code behavior.

[ijuma]

If that's the case, then we should only have that one public class in this package. The rest should either be non public or in other packages. Is that how it is?

[kirktrue]

No, all of the classes (public and non-public) are in the package together. Separating them out should be trivial from a code standpoint but please advise on what is needed to change that the proper way from a process standpoint (KIP, Jira, etc.).

[junrao]

@kirktrue : We could just make the changes to expose the proper public API and send an update to the original KIP-768 voting thread so that people are aware of the minor public changes if any.

[kirktrue]

I have filed KAFKA-13725 to address the package layout bug.