-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: WiredTiger
-
None
-
Storage Engines - Foundations
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Writes to disaggregated (layered) tables have always required a commit
timestamp. WT-18585 added a check that enforces it. MongoDB has call sites that
violate the requirement, so the check currently defaults to off
(debug_mode=(disagg_commit_ts_optional=true)); WT-18608 tracks turning it on.
Declare each violating site explicitly with
RecoveryUnit::allowOneUntimestampedWrite() (begin_transaction(no_timestamp=true)),
or give it a real timestamp where that is straightforward, so the global opt-out
can be dropped.
Declaring only makes a violation per-site and greppable; it does not make the
write correct. Replacing the remaining declarations with real timestamps is
tracked separately.
- blocks
-
SERVER-134926 Replace no_timestamp declarations on disaggregated writes with real commit timestamps
-
- Open
-
- is related to
-
WT-18585 Untimestamped write commits are accepted while step_down_timestamp is set
-
- Closed
-
-
WT-18608 Enable the disaggregated commit timestamp check once the MongoDB failures are fixed
-
- In Code Review
-
- related to
-
SERVER-134926 Replace no_timestamp declarations on disaggregated writes with real commit timestamps
-
- Open
-