-
Type: Task
-
Resolution: Unresolved
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Checkpoints
-
Storage Engines
-
3
-
StorEng - Refinement Pipeline
WT-12363 removed a couple of stats related to the checkpoint state: WT_CHECKPOINT_STATE_RESET_CURSORS and WT_CHECKPOINT_STATE_FLUSH_TIER_WAIT.
This ticket should investigate if there is a way to add them back in safely. The tricky part is that multiple callers can execute checkpoint code that might change the checkpoint state. In WT-12363, it was found that as long as the state is changed in __wt_txn_checkpoint, we can safely update it. However, the code in __checkpoint_tree needs an additional check through WT_SESSION_IS_CHECKPOINT to ensure the code is executed after calling the checkpoint session API. Indeed, this code can be triggered by the tiered server too.