-
Type:
Task
-
Resolution: Won't Do
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Cluster Scalability
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Each operation fetches and applies in its own batches. A transaction can be cut across a batch boundary in one operation and not in the other. Resuming in the middle of such a transaction must not lose or duplicate any of its entries.
Set the fetcher and applier batch sizes to 1. While both operations are applying, run small, large and prepared transactions that write to A and B. Pause one operation's fetcher partway through a transaction. Then make it resume, once with a recipient stepdown and once with a donor error.
Assert:
- no duplicate _ids
- every entry of each transaction is applied exactly once
- each operation's buffer contains only its own collection's entries