- `apiserver_authentication_config_controller_last_config_info` metric after successfully loading the authentication configuration file.
  - `apiserver_authorization_config_controller_last_config_info` metric after successfully loading the authorization configuration file.

	startTime := time.Now()
	return func() {
		duration := time.Since(startTime)

		atomic.AddUint64(&o.operations[s], 1)
		o.latency[s].add(duration)
	}
}

// incTime will increment time on metric s with a specific duration.
func (o *osMetrics) incTime(s osMetric, d time.Duration) {
	atomic.AddUint64(&o.operations[s], 1)
	o.latency[s].add(d)
}

- Introduced new apiserver metric `apiserver_flowcontrol_current_inqueue_seats`. This metric is analogous to `apiserver_flowcontrol_current_inqueue_requests`, but tracks the total number of seats, as each request can take more than one seat. ([#119385](https://github.com/kubernetes/kubernetes/pull/119385), [@andrewsykim](https://github.com/andrewsykim))

- HPA controller exposes the following metrics from the kube-controller-manager. 
  - `metric_computation_duration_seconds`: The time(seconds) that the HPA controller takes to calculate one metric.
  - `metric_computation_total`: Number of metric computations. ([#116326](https://github.com/kubernetes/kubernetes/pull/116326), [@sanposhiho](https://github.com/sanposhiho)) [SIG Apps, Autoscaling and Instrumentation]

- Change  `preemption_victims` metric bucket from `LinearBuckets` to `ExponentialBuckets`. ([#112939](https://github.com/kubernetes/kubernetes/pull/112939), [@lengrongfu](https://github.com/lengrongfu)) [SIG Instrumentation and Scheduling]
- Extend the job `job_finished_total metric by new `reason` label and introduce a new job metric to count pod failures

	m.Set(apiTrafficSentBytes, float64(connStats.s3OutputBytes), "type", "s3")
	m.Set(apiTrafficRecvBytes, float64(connStats.s3InputBytes), "type", "s3")
	return nil
}

// Metric Descriptions for bucket level S3 metrics.
var (
	bucketAPITrafficSentBytesMD = NewCounterMD(apiTrafficSentBytes,
		"Total number of bytes received for a bucket", "bucket", "type")

	m.RLock()
	defer m.RUnlock()

	if jobID != "" {
		if job, ok := m.metrics[jobID]; ok {
			metrics.Jobs[jobID] = job.metric()
		}
		return metrics
	}

	for id, job := range m.metrics {
		metrics.Jobs[id] = job.metric()
	}
	return metrics
}

// keep job metrics for some time after the job is completed
// in-case some one wants to look at the older results.

    Potential impacts:

  /**
   * Returns the number of bytes required to encode these headers using HTTP/1.1. This is also the
   * approximate size of HTTP/2 headers before they are compressed with HPACK. This value is
   * intended to be used as a metric: smaller headers are more efficient to encode and transmit.
   */
  fun byteCount(): Long {
    // Each header name has 2 bytes of overhead for ': ' and every header value has 2 bytes of
    // overhead for '\r\n'.

- Fix etcd_object_counts metric reported by kube-apiserver ([#94818](https://github.com/kubernetes/kubernetes/pull/94818), [@tkashem](https://github.com/tkashem)) [SIG API Machinery]