Without some kind of "local" curOp on mongos, it's impossible to tell whether a failpoint has been hit. Because of this, we cannot have (non-flaky) test coverage for the case where a pinned cursor is killed at a point where the operation using the cursor won't call checkForInterrupt again.
Once we have a "local" curOp, it will be easy to modify jstests/sharding/kill_pinned_cursor.js to check this case.
- depends on
-
SERVER-18094 currentOp on a mongoS should also show current mongoS operations
-
- Closed
-
- related to
-
SERVER-38418 Add waitWithPinnedCursorDuringGetMoreBatch failpoint to v3.6.
-
- Closed
-