We currently record a histogram of scheduling stats for the TransportLayer's reactor threads, which is useful for determining when the reactor thread is overloaded and falling behind. We should also add a log statement that is periodically emitted when a task waits a long time in order to begin executing. This will provide better visibility into how individual tasks are affected by reactor performance, which FTDC cannot provide.
- related to
-
SERVER-106022 Add log statement for slow-to-run reactor tasks
-
- Backlog
-
-
SERVER-58408 Improve diagnostics for networking reactors
-
- Backlog
-
-
SERVER-96848 Reject work if reactor is overwhelmed
-
- Backlog
-