-
Type:
Bug
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: Layered Tables
-
None
-
Storage Engines - Foundations
-
206.736
-
None
-
None
On a disaggregated follower, a layered cursor can reopen its stable cursor even though no newer checkpoint has arrived. When a transaction is allowed to advance the stable cursor (for example, when reading at a timestamp), the cursor advances unconditionally without first checking whether the stable checkpoint has actually moved forward.
This results in unnecessary stable cursor reopens, which are wasteful and increase the layered_curs_reopen_stable statistic with no benefit.
Regression introduced by the rework in WT-17679.
Steps to reproduce:
- On a follower, position a layered cursor on the ingest table (not the stable table).
- In a single transaction, perform two operations.
- Observe that the second operation reopens the stable cursor even though no new checkpoint has been picked up.
- is related to
-
WT-17679 Refactor layered cursor enter logic
-
- Closed
-