[improvement](cloud) Add enable_recycler config to skip recycler dynamically

[mymeiyi]

No description provided.

[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?

[mymeiyi]

/review

[Copilot]

Pull request overview

This PR adds a mutable cloud config to dynamically disable recycler scheduling and instance recycling work.

Changes:

• Adds enable_recycler mutable config with default true.
• Gates recycler instance scanning and recycle callback processing on the new config.
• Adds unit tests for default behavior and skip behavior when disabled.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
cloud/src/common/config.h	Defines the new mutable enable_recycler config.
cloud/src/recycler/recycler.cpp	Skips instance scanning and pending recycle processing when the config is disabled.
cloud/test/recycler_test.cpp	Adds tests covering default config and disabled recycler behavior.

---

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

[github-actions]

Review result: request changes. The new config is small and focused, but the implementation does not fully satisfy a dynamic recycler disable switch because in-flight recycle rounds keep running and leasing after the config is turned off. Test coverage also does not currently prove that scanning or recycling was skipped.

Critical checkpoint conclusions:

• Goal/test proof: Partially met. The code skips new scans and pending instances when disabled, but does not stop active recycling; the added tests do not prove the guards are effective.
• Scope: The code change is small and focused.
• Concurrency/lifecycle: The recycler has scanner, worker, lease, and checker threads. The new checks are only at scanner enqueue and worker dequeue boundaries, so an already-running InstanceRecycler remains live and lease_recycle_jobs continues renewing it.
• Config behavior: A mutable config was added and is observed dynamically only at loop boundaries; disabling is not applied promptly to active recycle tasks.
• Compatibility/storage format: No incompatible format or protocol changes found.
• Parallel paths: Snapshot migrator/chain compactor are separate switches and not directly affected.
• Tests: Unit tests were added, but two do not distinguish the new disabled path from existing no-op/failure behavior. No test result was run by me in this review.
• Observability: Logs are present for skipped scans/workers, but active tasks continue without a disable-triggered stop indication.
• Transaction/persistence/data correctness: No direct new persisted metadata path; the concern is that active recycler work can continue deleting objects/KV after operators disable the switch.
• Performance: The disabled scanner path can log every interval; acceptable at default interval, but tests use a zero interval.
• User focus: No additional user-provided review focus was supplied.

[mymeiyi]

run buildall

[hello-stephen]

Cloud UT Coverage Report

Increment line coverage 100.00% (24/24) 🎉

Increment coverage report
Complete coverage report

Category	Coverage
Function Coverage	78.06% (1854/2375)
Line Coverage	64.53% (33339/51662)
Region Coverage	65.24% (16532/25341)
Branch Coverage	55.75% (8838/15852)

[mymeiyi]

/review

[github-actions]

Summary: I found one additional issue beyond the existing review threads. The PR adds a mutable recycler switch, but the scanner can sleep for the full recycle interval after observing enable_recycler=false, so turning the switch back on may not resume recycling for up to the default 3600 seconds. Existing inline threads also already cover the in-flight recycle not stopping promptly and the new tests not proving the disabled paths.

Critical checkpoints:

• Goal/test proof: The goal is to dynamically disable/enable recycler work. The current implementation only gates scanning/new starts and does not fully prove or deliver dynamic behavior; tests for disabled paths are already called out as insufficient, and re-enable latency is not covered.
• Scope/focus: The change is small, but the minimal checks are placed only in scanner/new-worker paths, leaving dynamic lifecycle gaps.
• Concurrency/lifecycle: Recycler scanner, worker, lease, and task-check threads run concurrently. The new flag is read from worker threads while updated externally, and the change lacks a wakeup/abort mechanism for prompt state transitions.
• Configuration: A new mutable config is added, but affected recycler threads do not reliably observe changes promptly.
• Parallel paths: Scanner and new recycle start paths were touched; active recycle/lease paths are still covered by existing review comments as missing dynamic-disable handling.
• Tests: Added tests do not distinguish skipped work from attempted work, per existing threads, and do not cover re-enabling behavior.
• Recycler-specific safety: No direct mark-before-delete or packed-file ordering change was introduced, but incomplete dynamic disable can still allow recycle/deletion to continue in active jobs as already noted.

Focus response: No additional user-provided review focus was specified.

[gavinchou]

LGTM

Dismissing review to unblock PR merge. All checks passed, gavinchou approved.