| Attached is a script showing inconsistency in how killCursors is not counted. Put sharding-histogram-miscounts.js in jstests. This script is asserts that the killCursors command is being counted 0 times. If you run with the core executor, it will fail as expected, i.e.:
python ./buildscripts/resmoke.py --executor=core ./jstests/sharding-histogram-miscounts.js
|
This fails with the expected message:
[thread1] Error: [99] != [0] are not equal : Miscounted commands :
|
However, running the following will incorrectly succeed:
python ./buildscripts/resmoke.py --executor=sharded_collections_jscore_passthrough ./jstests/sharding-histogram-miscounts.js
|
However, if the test is run with the sharded_collections_jscore_passthrough executor and repeated multiple times, it will strangely fail
python ./buildscripts/resmoke.py --repeat=30 --executor=sharded_collections_jscore_passthrough ./jstests/sharding-histogram-miscounts.js
|
This fails with the a message similar to this one:
[js_test:sharding-histogram-miscounts] 2016-07-01T14:23:30.530-0400 assert: [700] != [0] are not equal : Miscounted commands
|
My current conjecture (based on logging from mongod) is that the kill cursors command are these additional commands. I'm unsure why these are counted since the script does call close on the cursors.
|