-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Replication
-
None
-
Fully Compatible
-
ALL
-
0
This does not necessarily throw if the error message is returned fast enough before the connection is killed.
On transition to rollback, we mark operations killed first then close the connections which calls _performPostMemberStateUpdateAction. So if the error message gets back to the client during that window, client doesn't necessarily throw but instead gets InterruptedDueToReplStateChange.
This is new behavior introduced in this commit (SERVER-41035).
So we need to change the test not to assert.throw.