-
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 must survive failovers on its own, and a failover that affects one operation must not affect the others.
Run N reshardings with overlapping donors and recipients, then step down:
- a shard involved in only one operation (subset)
- a shard that is a donor for A and a recipient for B, during cloning and again during applying
- the config server primary, while N coordinators are running (all)
Run these at different phases: cloning, applying, and in or near the critical section.
Assert:
- operations not involving the stepped-down node keep making progress
- the affected operations resume and complete
- every collection ends up with exactly its own documents
- retryable writes stay correct across the failovers