Details
-
Bug
-
Resolution: Fixed
-
Major - P3
-
3.5.11
-
None
-
Fully Compatible
-
ALL
-
Platforms 2017-09-11, Platforms 2017-10-02
Description
The cumulative time metrics such as _totalSpentExecuting and _totalSpentScheduled are accumulated in thread-local variables and are only rolled up to the global variables every few seconds, so that they increment in jumps every few seconds instead of continuously. This negates their usefulness for measuring the instantaneous number of operations in those states and instantaneous latency. They should either be incremented on every operation, or should be rolled up much more frequently.