-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Storage Execution
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Implement the C1 clear trigger for set-wide Replicated RecordIds : applying a new-primary noop whose (groupId, foundingOpTime) is absent or differs from the node's own drops the node's groupId table at the noop's timestamp.
Details
- Appliers currently skip all noops without reading the payload, add inspection for the new-primary noop fields.
- The noop must be processed individually so the clear commits before any later-timestamped entry applies: change the isNewPrimaryNoop case in OplogApplierBatcher from kStartNewBatch to kProcessIndividually.
- Active in steady state and during initial sync oplog application.
- The clear is idempotent and timestamped at the noop's timestamp, so rollback of a flapped term restores membership.
- Provider gating: C1 runs only on attached storage. Disagg appliers skip the noop inspection entirely, their rid honoring is architectural and unconditional.
Acceptance criteria
- Mismatch clears, match is a no-op, absent field clears.
- Batcher unit test asserts the singleton batch and unit test proves no entry of the new term applies before the clear commits.
- No C1 activity on disaggregated storage: applying an absent- or mismatched-groupId noop on the disagg provider leaves rid honoring on and touches nothing.
- depends on
-
SERVER-138106 Mint and announce the groupId at election (new-primary noop)
-
- Open
-
- is depended on by
-
SERVER-138124 Membership-based recordId emission and honoring in the oplog path
-
- Open
-
-
SERVER-138231 Logical initial sync joins the RecordIds group
-
- Open
-