I noticed that the latency monitoring seems to only account for the time between when a worker receives a task and when it starts the task.
However, it would be great if we could also track the latency from when a task is published. The task-sent event is disabled by default, but if enabled on a cluster, it could be valuable in understanding where scheduling latencies are occurring.
I noticed that the latency monitoring seems to only account for the time between when a worker receives a task and when it starts the task.
However, it would be great if we could also track the latency from when a task is published. The
task-sentevent is disabled by default, but if enabled on a cluster, it could be valuable in understanding where scheduling latencies are occurring.