* percentiles together takes about 55% as long as computing them separately.
 *
 * <p>When calling {@link ScaleAndIndex#compute} (in {@linkplain ScaleAndIndexes#compute either
 * form}), the memory requirement is 8*N bytes for the copy of the dataset plus an overhead which is
 * independent of N (but depends on the quantiles being computed). When calling {@link

		return rd
	}
	rd.Replicate = newReset && oi.ModTime.Before(resetBeforeDate)
	return rd
}

const resyncTimeInterval = time.Minute * 1

// PersistToDisk persists in-memory resync metadata stats to disk at periodic intervals
func (s *replicationResyncer) PersistToDisk(ctx context.Context, objectAPI ObjectLayer) {
	resyncTimer := time.NewTimer(resyncTimeInterval)
	defer resyncTimer.Stop()

That "wait for something else" normally refers to <abbr title="Input and Output">I/O</abbr> operations that are relatively "slow" (compared to the speed of the processor and the RAM memory), like waiting for:

* the data from the client to be sent through the network
* the data sent by your program to be received by the client through the network

	SRAIW	$-1, X5, X6			// ERROR "immediate out of range 0 to 31"
	SD	X5, 4294967296(X6)		// ERROR "constant 4294967296 too large"
	FNES	F1, (X5)			// ERROR "needs an integer register output"

	// Memory Ordering Instructions
	FENCE		X1, R				// ERROR "invalid FENCE predecessor operand"
	FENCE		R, X2				// ERROR "invalid FENCE successor operand"
	FENCE		$1, R				// ERROR "invalid FENCE predecessor operand"

  ([#33684](https://github.com/kubernetes/kubernetes/pull/33684),
  [@fraenkel](https://github.com/fraenkel))
* Flag gate faster evictions based on node memory pressure using kernel memcg
  notifications - `--experimental-kernel-memcg-notification`.
  ([#38258](https://github.com/kubernetes/kubernetes/pull/38258),
  [@derekwaynecarr](https://github.com/derekwaynecarr))
* Nodes can now report two additional address types in their status: InternalDNS

  /**
   * Returns a fresh {@link MapMakerInternalMap} with {@link MapMaker.Dummy} values but otherwise as
   * specified by the given {@code builder}. The returned {@link MapMakerInternalMap} will be
   * optimized to saved memory. Since {@link MapMaker.Dummy} is a singleton, we don't need to store
   * any values at all. Because of this optimization, {@code build.getValueStrength()} must be
   * {@link Strength#STRONG}.
   *

- Fixes a kubelet and kube-apiserver memory leak in default 1.29 configurations related to tracing. ([#126985](https://github.com/kubernetes/kubernetes/pull/126985), [@dashpole](https://github.com/dashpole)) [SIG API Machinery and Node]

* Безпека - HTTPS
* Автозапуск
* Перезапуски
* Реплікація (кількість запущених процесів)
* Пам’ять
* Попередні кроки перед запуском

## Пам’ять { #memory }

Якщо ви запускаєте один процес на контейнер, ви матимете більш-менш чітко визначений, стабільний і обмежений обсяг пам’яті, що споживається кожним із цих контейнерів (їх може бути більше одного, якщо вони репліковані).

---

要点是，这些都**不是**你必须盲目遵循的**铁律**。你可以用这些思路来**评估你自己的场景**，并决定最适合你的系统的方法，看看如何管理以下概念：

* 安全 - HTTPS
* 启动时运行
* 重启
* 复制（运行的进程数）
* 内存
* 启动前的前置步骤

## 内存 { #memory }

如果你**每个容器只运行一个进程**，那么每个容器消耗的内存将更容易定义、较为稳定且有限（如果有复制则为多个容器）。

接着，你可以在容器管理系统（例如 **Kubernetes**）的配置中设置同样的内存限制与需求。这样它就能在**可用的机器**上**复制容器**，同时考虑容器所需的内存量以及集群中机器可用的内存量。

- Fix memory leak in the job controller related to JobTrackingWithFinalizers ([#111723](https://github.com/kubernetes/kubernetes/pull/111723), [@alculquicondor](https://github.com/alculquicondor)) [SIG Apps]