-
Type:
Task
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: Replication, Sharding
-
None
-
Minor Change
-
Sharding 2021-03-22
Currently when dropping a nonexistent database, we return a response that contains information about if the database previously existed or not.
- On replicaset
- On sharded cluster
This response is not reliable if a step-down happens during the operation, in fact on step-down the operation is automatically retried on the new replicaset primary. On this new node we don't have any way of assessing previous existence of the database because the old primary could have either already drop it or not.
Moreover the drop database response is not currently documented anywhere.
The goal of this ticket is to simplify the drop database response by not returning any information on the previous state of the database being dropped.
- is duplicated by
-
SERVER-54100 Remove _shardsvrDropDatabase command unused reply
-
- Closed
-
- is related to
-
SERVER-43894 Make dropping a nonexistent collection a noop
-
- Closed
-
-
DRIVERS-2118 Create collection and database management specifications
-
- Backlog
-
- related to
-
-
- Closed
-