[fix](cloud) Recycle empty rowsets without resource id

[liaoxin01]

What problem does this PR solve?

Issue Number: None

Problem Summary:

When recycling a dropped tablet, empty rowsets may have num_segments == 0 and an empty resource_id. The protobuf field can be explicitly set to an empty string, so has_resource_id() returns true. The recycler then attempts to look up an accessor for the empty ID, fails the recycle round, and repeatedly leaves the dropped tablet/index metadata behind.

This PR skips object deletion for every rowset with zero segments because it has no segment objects to remove. Rowsets with segments still require a non-empty resource ID.

Release note

Fix cloud recycler failing to remove dropped tablet metadata when an empty rowset has no resource ID.

Check List (For Author)

• Test

  • Regression test
  • Unit Test
  • Manual test
  • No need to test or manual test.

• Behavior changed:

  • No.
  • Yes. Empty rowsets are skipped during object deletion because they have no segment objects; non-empty rowsets still require a valid resource ID.

• Does this need documentation?

  • No.
  • Yes.

Check List (For Reviewer who merge this PR)

• Confirm the release note
• Confirm test cases
• Confirm document
• Add branch pick label

[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]

Pull request overview

This PR fixes a cloud recycler failure mode where dropped tablet recycling can get stuck if a rowset protobuf has num_segments == 0 and resource_id is present but empty (so has_resource_id() is true). The recycler now skips object deletion for empty rowsets (since there are no segment objects to remove) and still enforces a non-empty resource ID for non-empty rowsets.

Changes:

• Skip rowsets with num_segments == 0 during dropped tablet object deletion to avoid accessor lookup on empty resource IDs.
• Tighten the resource ID validation for non-empty rowsets by checking resource_id().empty() (covers both “unset” and “set-to-empty” cases).
• Add a unit test covering the explicitly-empty resource_id + zero-segment rowset scenario.

Reviewed changes

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

File	Description
cloud/src/recycler/recycler.cpp	Skip empty rowsets before accessor lookup; require non-empty resource_id for non-empty rowsets.
cloud/test/recycler_test.cpp	Add unit test validating recycling succeeds when rowset has 0 segments and empty resource_id.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[liaoxin01]

/review

[liaoxin01]

run buildall

[wyxxxcat]

LGTM

[github-actions]

I found one blocking gap in the fix.

Critical checkpoints:

• Goal/test: The normal internal_get_rowset path is fixed and covered by a unit test, but the same empty-rowset shape can still reach recycle_tablet via restore-job rowset KVs, so the dropped-tablet cleanup goal is incomplete.
• Scope/focus: The change is small, but a functionally parallel path in the same function was not updated.
• Concurrency/lifecycle: No new concurrency or lifetime management is introduced. The failure is retry/lifecycle related: restore-job recycle rounds can repeatedly fail before KV removal.
• Config/compatibility/protocol: No config, format, or protocol change.
• Transaction/persistence: Normal empty-rowset KV cleanup proceeds after object deletion; restore-job rowset KVs can still be left behind when the restore loop returns -1.
• Data/object deletion: Skipping zero-segment normal rowsets is consistent with the lack of segment objects; packed-file ordering concerns were reviewed and I did not find a new issue for this path.
• Tests: Added coverage exercises only the normal rowset response injection; restore-job rowset coverage is missing. I did not run unit tests; git diff --check for the two PR files passed.
• User focus: No additional user-provided review focus.

fixed

[github-actions]

PR approved by anyone and no changes requested.

[hello-stephen]

Cloud UT Coverage Report

Increment line coverage 100.00% (6/6) 🎉

Increment coverage report
Complete coverage report

Category	Coverage
Function Coverage	77.32% (1889/2443)
Line Coverage	64.38% (33944/52726)
Region Coverage	64.77% (17454/26948)
Branch Coverage	53.93% (9339/17316)