Skip to content
TheChatArchiveThe Chat Archive
Claudeclaude-opus-5· anthropicHugging Face ibragim-bad/swe_rebench_07_2026_trajectories

ISSUE DESCRIPTION [Bug] Broker may accept duplicate messages after dedup recovery race…

  1. User

    ISSUE DESCRIPTION

    [Bug] Broker may accept duplicate messages after dedup recovery race

    Issue Description

    Message deduplication recovery can be started more than once for the same topic before the first recovery replay completes.

    MessageDeduplication defines a Recovering status to prevent overlapping transitions, but the enable path starts dedup cursor replay without first changing the status to Recovering.

    Because of that, another checkStatus() call can still observe Initialized or Disabled and start a second replay for the same dedup cursor.

    One practical trigger path is topic load with existing topic-level policies:

    TimeFlow A: existing topic-level policy during loadFlow B: normal topic load chainDedup status / effect
    T1initTopicPolicy() loads existing topic policiesdedup status is Initialized or Disabled
    T2onUpdate() applies topic policies and calls checkDeduplicationStatus()first dedup recovery replay starts
    T3first replay is still running asynchronouslystatus is still Initialized or Disabled
    T4topic load continues and BrokerService explicitly calls checkDeduplicationStatus()
    T5second check also enters the enable pathsecond replay starts for the same dedup cursor
    T6first replay rebuilds producer sequence statesecond replay also rebuilds producer sequence stateoverlapping replay can advance shared replay/cursor state

    Dedup recovery rebuilds producer sequence information from the dedup cursor. If overlapping replay leaves the recovered sequence state incomplete or inconsistent, the broker may fail to identify already-published messages as duplicates.

    The impact is that duplicate messages can be accepted after topic load or policy refresh even though message deduplication is enabled.

    There is also a retry gap after recovery failure. If enabling deduplication fails transiently, the status moves to Failed, but later checks do not retry enabling even when the topic policy still requires deduplication.

    Error messages

    There may be no error message. This is a race in the dedup recovery state machine.

    Reproducing the issue

    A deterministic unit-level reproduction can be built by delaying the first dedup recovery replay and invoking checkStatus() again before the first replay completes:

    1. Keep MessageDeduplication in Initialized or Disabled.
    2. Call checkStatus() once and delay async cursor open or replay completion.
    3. Call checkStatus() again before the first recovery finishes.
    4. Observe that the second call enters the enable path again and starts another replay.

    A production-like trigger path is:

    1. Enable message deduplication.
    2. Use a topic with existing topic-level policies.
    3. Load or reload the topic.
    4. During topic load, initTopicPolicy() can invoke onUpdate(), which applies topic policies and calls checkDeduplicationStatus().
    5. The normal topic load chain later calls checkDeduplicationStatus() again.
    6. If the first recovery replay is still in progress, both checks can start overlapping dedup replay.

    Expected behavior

    • Deduplication should enter Recovering before starting recovery replay.
    • Concurrent status checks should not start another recovery while replay is in progress.
    • If enable fails transiently, a later check should retry enabling when deduplication is still required by policy.

    Actual behavior

    • Enable starts replay while status remains Initialized or Disabled.
    • Another checkStatus() can start a second replay before the first completes.
    • Failed status does not retry enable even when deduplication should still be enabled.

    Additional information

    A fix has been proposed in #25953.

    Interface

    No new interfaces are introduced.

    ADDITIONAL ADVICE

    Since you are given a git repository, you can use git commands to simplify your work. For example, if you made a mistake and want to revert your changes, you can use git checkout <file> to restore the file to its original state. You can also reset all changes in the repository using git reset --hard command. Additionally, you can use git stash and git stash pop to temporarily save your changes and restore them later. Finally, you can see the changes that you've made (which will be submitted when you call submit) using git status or git diff commands. However you don't need to use git add or git commit before submitting your changes. In fact, submit only submits currently unstaged changes because it uses git diff to get the changes that need to be submitted.

    CHECKLIST

    Before submitting your solution, please go over the following checklist and make sure you've done everything:

    • If an error was reported in the issue description, I have successfully reproduced it.
    • If an error was reported in the issue description, I have confirmed that my fix resolves the error.
    • I have thought about edge cases and made sure my fix handles them as well.
    • I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. I want you to list every bullet from this checklist and write your reasoning for why you think you did it or didn't need to.

    Repository has been uploaded and your shell is currently at the repository root. Time to solve the issue!

    (Current directory: /pulsar, current file: none) bash-$

  2. Claude Assistant

    I'll use the edit command instead.

  3. Claude Assistant

    All checks pass. Here's my checklist review:

    • If an error was reported in the issue description, I have successfully reproduced it. Done. The issue describes a race rather than an exception, and I built a temporary deterministic unit test (TempDedupRaceTest) that mocked ManagedLedger.asyncOpenCursor so the first dedup recovery stayed pending. With the original code the test failed with expected [Recovering] but found [Initialized] and a second checkStatus() call entered the enable path again (second asyncOpenCursor/replay), confirming the race. I also reproduced the "no retry after Failed" gap in the same harness.

    • If an error was reported in the issue description, I have confirmed that my fix resolves the error. Done. After the fix (status = Status.Recovering set inside the synchronized block before opening the cursor/starting replay, plus Failed added to the enable-retry condition and a dedicated result future whose failure path sets Failed), the temp test passed: only one asyncOpenCursor/replay was started for two concurrent checkStatus() calls, the status was Recovering during replay and Enabled afterwards, and a Failed enable was retried by a later checkStatus().

    • I have thought about edge cases and made sure my fix handles them as well. Done:

      • The Initialized && !shouldBeEnabled (delete cursor) path and the Enabled && !shouldBeEnabled (disable) path are untouched; Removing still blocks new transitions.
      • The previous code returned the composed future and then called future.completeExceptionally(e) on it inside exceptionally (a no-op/self-referential bug); I now use a separate result future so the caller reliably observes success/failure exactly once.
      • Failure now always transitions to Failed (via whenComplete), so the state machine never gets stuck in Recovering on error, and a subsequent check retries enabling only while the policy still requires dedup (shouldBeEnabled).
      • I deliberately kept the change scoped: Failed + !shouldBeEnabled behavior is unchanged, as the issue only asks for enable retry.
    • I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. Done. Ran MessageDuplicationTest (5), DeduplicationDisabledBrokerLevelTest (2), TopicDuplicationTest (13), BrokerMessageDeduplicationTest (3) — all passed with 0 failures. Also verified :pulsar-broker:compileJava, checkstyleMain and spotlessCheck succeed. The temporary reproduction test was deleted, so only the source fix remains in the diff.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.