[fix](cloud) Fix warmup inflight count not decremented on early return #60480
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
run buildall |
gavinchou
reviewed
Feb 3, 2026
gavinchou
reviewed
Feb 3, 2026
deardeng
force-pushed
the
fix-warmup-inflight-count-leak
branch
from
0491cea to
0c66eac
Compare
During cloud tablet decommission, some tablets take unexpectedly long time (5+ minutes) to migrate because FE keeps waiting for warmup tasks to complete, even though the tasks have already failed. Root cause: In `FileCacheBlockDownloader::download_file_cache_block()`, when early return occurs (e.g., tablet not found, rowset not found, storage resource error), the `_inflight_tablets` count is not decremented. This causes: 1. `check_download_task()` always returns `done=false` for these tablets 2. FE's `checkInflightWarmUpCacheAsync()` waits until timeout (default 300s) 3. Tablet migration is blocked unnecessarily Example log showing the issue: ``` W download_file_cache_block: tablet_id=xxx rowset_id not found, rowset_id=xxx ``` After this warning, the tablet's inflight count remains, causing the 5-minute wait. 1. Extract the inflight count decrement logic into a reusable lambda `decrease_inflight_count` 2. Call `decrease_inflight_count()` in all early return paths: - When `get_tablet()` fails - When `rowset_id` is not found - When `remote_storage_resource()` fails 3. Refactor `download_done` callback to reuse `decrease_inflight_count`, eliminating code duplication 4. Use value capture for `decrease_inflight_count` in `download_done` lambda to ensure lifetime safety if the callback is ever called asynchronously 5. Add unit tests to verify inflight count is correctly decremented on failures Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
deardeng
force-pushed
the
fix-warmup-inflight-count-leak
branch
from
0c66eac to
7df33de
Compare
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 32304 ms |
ClickBench: Total hot run time: 28.29 s |
gavinchou
previously approved these changes
Feb 3, 2026
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
Contributor
Author
|
run beut |
Contributor
Author
|
run external |
Contributor
Author
|
run nonConcurrent |
Contributor
Author
|
run p0 |
Contributor
Author
|
run buildall |
github-actions
bot
removed
the
approved
TPC-H: Total hot run time: 30433 ms |
ClickBench: Total hot run time: 28.27 s |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
gavinchou
approved these changes
Feb 10, 2026
hello-stephen
approved these changes
Feb 11, 2026
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed changes
Problem
During cloud tablet decommission, some tablets take unexpectedly long time (5+ minutes) to migrate because FE keeps waiting for warmup tasks to complete, even though the tasks have already failed on BE side.
Root cause: In
FileCacheBlockDownloader::download_file_cache_block(), when early return occurs (e.g., tablet not found, rowset not found, storage resource error), the_inflight_tabletscount is not decremented. This causes:check_download_task()always returnsdone=falsefor these tabletscheckInflightWarmUpCacheAsync()waits until timeout (default 300 seconds)Example log showing the issue:
After this warning, the tablet's inflight count remains in
_inflight_tabletsmap, causing the 5-minute wait before FE times out and proceeds.Solution
Extract the inflight count decrement logic into a reusable lambda
decrease_inflight_countCall
decrease_inflight_count()in all early return paths:get_tablet()failsrowset_idis not foundremote_storage_resource()failsRefactor
download_donecallback to reusedecrease_inflight_count, eliminating code duplicationUse value capture for
decrease_inflight_countindownload_donelambda to ensure lifetime safety if the callback is ever called asynchronously in the futureAdd unit tests to verify inflight count is correctly decremented on failures
Further comments
This bug also causes a minor memory leak: entries in
_inflight_tabletsmap are never cleaned up when warmup fails, slowly accumulating over time (cleared on BE restart).Checklist(Required)