In the current implementation, when the Scheduling and Preempting stages fail, both scheduler_pod_scheduling_attempts{unschedulable} and scheduler_pod_preempting_attempts{nominatedFailure} will be recorded, leading to unclear semantics for these metrics.
In this MR:
when Scheduling and Preempting both fail: scheduler_pod_scheduling_attempts{unschedulable} += 1; scheduler_pod_preempting_attempts{nominatedFailure} += 1
when Scheduling fails but Preempting succeeds: scheduler_pod_scheduling_attempts{scheduled} += 1; scheduler_pod_preempting_attempts{nominated} += 1
In the current implementation, when the Scheduling and Preempting stages fail, both
scheduler_pod_scheduling_attempts{unschedulable}
andscheduler_pod_preempting_attempts{nominatedFailure}
will be recorded, leading to unclear semantics for these metrics.In this MR:
scheduler_pod_scheduling_attempts{unschedulable} += 1
;scheduler_pod_preempting_attempts{nominatedFailure} += 1
scheduler_pod_scheduling_attempts{scheduled} += 1
;scheduler_pod_preempting_attempts{nominated} += 1