query results are printed to the console, organized by workload name.

All metrics returned are from server-side reports. This means that latencies
and error rates are from the perspective of the service itself and not of an
individual client (or aggregate set of clients). Rates and latencies are
calculated over a time interval of 1 minute.
`,
		Example: `  # Retrieve workload metrics for productpage-v1 workload

		uerr = errors.Unwrap(leafErr)
		depth++
		if depth == unwrapErrsDepth {
			// If we have reached enough depth we
			// do not further recurse down, this
			// is done to avoid any unnecessary
			// latencies this might bring.
			break
		}
	}
	if uerr == nil {
		leafErr = err
	}
	return leafErr
}

// One log message per error.

	m.Set(driveAvailabilityErrorsTotal, float64(disk.Metrics.TotalErrorsAvailability), labels...)
	m.Set(driveWaitingIO, float64(disk.Metrics.TotalWaiting), labels...)

	// Append the api label for the drive API latencies.
	labels = append(labels, "api", "")
	lastIdx := len(labels) - 1
	for apiName, latency := range disk.Metrics.LastMinute {
		labels[lastIdx] = "storage." + apiName

// heuristic, they may be put into the same cluster and it can introduce
// artificial dependencies and incur great performance loss.  In this example,
// Node_Y becomes dependent on IteratorGetNext and the latencies add up if
// Node_X and Node_Y are in the same cluster.
//
// IteratorGetNext -> Node_X -> Stage
//
// Unstage -> Node_Y
//
Status ClusterScopingPassImpl::ScopingForPipelineStages() {