Today we declare several threadpools with nontrivial TaskTrackingConfig, but only use the task-tracking in certain specialized autoscaling setups. I wonder whether we should take more care to avoid this extra overhead unless we're going to use the tracking.
Today we declare several threadpools with nontrivial
TaskTrackingConfig
, but only use the task-tracking in certain specialized autoscaling setups. I wonder whether we should take more care to avoid this extra overhead unless we're going to use the tracking.