bobhan1 opened a new pull request, #61036:
URL: https://github.com/apache/doris/pull/61036

   ## Proposed changes
   
   This PR adds visibility into "hidden" time sinks in SyncRowset that are not 
covered by existing RPC metrics.
   
   ### Problem
   
   Query profile shows `SyncRowsetTime: 3sec828ms` but all RPC detail metrics 
are 0, making it difficult to diagnose where the time is spent.
   
   ### Solution
   
   Added 3 new profile metrics under `SyncRowsetTime`:
   
   - **SyncRowsetBthreadScheduleWaitTime**: bthread scheduling delay (from task 
creation to execution start)
   - **SyncRowsetMetaLockWaitTime**: total wait time for `_meta_lock` 
(std::shared_mutex) acquisitions
   - **SyncRowsetSyncMetaLockWaitTime**: total wait time for `_sync_meta_lock` 
(bthread::Mutex) acquisitions
   
   These metrics track lock wait times in:
   - `CloudTablet::sync_rowsets()`
   - `CloudTablet::sync_if_not_running()`
   - `CloudMetaMgr::sync_tablet_rowsets_unlocked()`
   
   ### Checklist
   
   - [x] Compiles and builds successfully
   - [ ] Existing UT passed (need verification)
   - [ ] Manual testing done
   
   ## Types of changes
   
   - [ ] Bugfix (non-breaking change which fixes an issue)
   - [ ] New feature (non-breaking change which adds functionality)
   - [ ] Breaking change (fix or feature that would cause existing 
functionality to not work as expected)
   - [ ] Docs Update / Doc Refactor
   - [ ] Code Refactor (internal modification that does not change 
functionality)
   - [ ] Test Case
   - [ ] CI (CD related, like github action, k8s yaml, etc.)
   
   ## Further comments
   
   Helps diagnose cases where high SyncRowsetTime is caused by:
   - Lock contention between concurrent scans on the same tablet
   - bthread worker pool saturation


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to