Details
-
Bug
-
Status: Closed (View Workflow)
-
Minor
-
Resolution: Fixed
-
10.4.13, 10.5.3
-
None
Description
- Currently when the rollback segment is allocated to the new booting transaction next slot/rollback-segment is allocated using a static rseg_slot counter.
- Neither is this logic mutex protected nor is the counter atomic.
- With increase scalability, it could potentially cause the same rollback segments to get allocated to multiple transactions when as per design allocation of rollback segment should be round-robin.
- Fix proposes to use atomic for rseg_slot.