[fix](checker) Avoid false-positive leaked delete bitmaps for unexpired job tmp rowsets

[mymeiyi]

The delete bitmap inverted check reported a delete bitmap as leaked when its owning rowset was not found among the committed rowsets / pending delete bitmaps of the tablet. However, a compaction job writes delete bitmaps keyed by its tmp output rowset id before the job commits and makes the rowset visible. During that window the delete bitmap exists while the rowset is not yet in the committed list, so it was wrongly flagged as leaked.

Add InstanceChecker::collect_unexpired_job_tmp_rowsets() to gather, per tablet, the rowset ids of job tmp rowsets that are still alive from the recycler's point of view, and skip them during the inverted check. The same calculate_tmp_rowset_expired_time() threshold as the recycler is reused (now exposed via recycler.h) so a delete bitmap is never reported as leaked while its tmp rowset is still alive from the recycler's view.

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

[mymeiyi]

/review

[github-actions]

Review conclusion: request changes.

Critical checkpoints:

• Goal/test coverage: The PR aims to avoid false leaked-delete-bitmap reports for still-live job tmp rowsets and adds focused unit coverage for unexpired, expired, and non-job tmp rowsets. The main path is covered, but the new overflow path is not safely handled.
• Scope/focus: The change is small and localized to the Cloud recycler checker.
• Concurrency/lifecycle: No new shared mutable state or long-lived lifecycle issue was found; this is a synchronous checker scan.
• Configuration/compatibility: No new config or storage/protocol compatibility change.
• Parallel paths: The checker now reuses the recycler tmp-rowset expiration helper, which is appropriate.
• Error handling/data correctness: One issue: when the collection cap is reached, the checker continues with incomplete tmp-rowset state and can report live delete bitmaps as leaked.
• Testing: Positive/negative tests were added, but the cap/overflow case that causes false positives is missing.
• Observability/performance: Logging exists for the cap, but logging does not prevent an incorrect checker result.

User focus: No additional user-provided focus points were listed.

[github-actions]

PR approved by at least one committer and no changes requested.

[mymeiyi]

run buildall

[hello-stephen]

Cloud UT Coverage Report

Increment line coverage 71.82% (79/110) 🎉

Increment coverage report
Complete coverage report

Category	Coverage
Function Coverage	78.40% (1916/2444)
Line Coverage	64.89% (34282/52828)
Region Coverage	65.32% (17636/26998)
Branch Coverage	54.00% (9372/17354)

dismiss