- Fixes a 1.34 regression starting pods with environment variables with a value containing `$` followed by a multi-byte character ([#136490](https://github.com/kubernetes/kubernetes/pull/136490), [@AutuSnow](https://github.com/AutuSnow)) [SIG Architecture and Node]

- The EndpointSlice controller now waits for EndpointSlice and Node caches to be synced before starting. ([#94086](https://github.com/kubernetes/kubernetes/pull/94086), [@robscott](https://github.com/robscott)) [SIG Apps and Network]
- Upon successful authorization check, an impersonated user is added to the system:authenticated group.

    * Add registry action to the kubernetes-worker charm

    * Add support for kube-proxy cluster-cidr option to kubernetes-worker charm

    * Fix kubernetes-master charm starting services before TLS certs are saved

    * Fix kubernetes-worker charm failures in LXD

    * Fix stop hook failure on kubernetes-worker charm

        </field>
        <field>
          <description>
            <![CDATA[
            <p>The configuration as DOM object.</p>
            <p>By default, every element content is trimmed, but starting with Maven 3.1.0, you can add
            {@code xml:space="preserve"} to elements you want to preserve whitespace.</p>

* Properly manage security groups for loadbalancer services on OpenStack. ([#65373](https://github.com/kubernetes/kubernetes/pull/65373), [@multi-io](https://github.com/multi-io))
* Reload systemd config files before starting kubelet. ([#65702](https://github.com/kubernetes/kubernetes/pull/65702), [@mborsz](https://github.com/mborsz))

	tgtClient, err := globalTierConfigMgr.getDriver(ctx, opts.Transition.Tier)
	if err != nil {
		return err
	}

	if !opts.NoLock {
		// Acquire write lock before starting to transition the object.
		lk := er.NewNSLock(bucket, object)
		lkctx, err := lk.GetLock(ctx, globalDeleteOperationTimeout)
		if err != nil {
			return err
		}
		ctx = lkctx.Context()

* Kubelet won't evict a static pod with priority `system-node-critical` upon resource pressure. ([#74222](https://github.com/kubernetes/kubernetes/pull/74222), [@Huang-Wei](https://github.com/Huang-Wei))
* Re-issue Allocate grpc calls before starting a container that requests device-plugin resources if the cached state is missing. ([#73824](https://github.com/kubernetes/kubernetes/pull/73824), [@jiayingz](https://github.com/jiayingz))

- Starting in 1.31, `container_engine_t` was added to the list of allowed SELinux types in the baseline Pod Security Standard. ([#126165](https://github.com/kubernetes/kubernetes/pull/126165), [@haircommander](https://github.com/haircommander))

しかし、**多くのメモリを使用**している場合（たとえば**機械学習**モデルなど）、どれだけのメモリを消費しているかを確認し、**各マシンで実行するコンテナの数**を調整する必要があります（そしておそらくクラスタにマシンを追加します）。

**コンテナごとに複数のプロセス**を実行する場合、起動するプロセスの数が**利用可能なメモリ以上に消費しない**ようにする必要があります。

## 開始前の事前ステップとコンテナ { #previous-steps-before-starting-and-containers }

コンテナ（DockerやKubernetesなど）を使っている場合、主に2つのアプローチがあります。

### 複数のコンテナ { #multiple-containers }

 *       of the computation. After that result is available (or the computation fails), all objects
 *       captured by any of the steps in the pipeline are closed.
 * </ol>
 *
 * <h3>Starting a pipeline</h3>
 *
 * Start a {@code ClosingFuture} pipeline {@linkplain #submit(ClosingCallable, Executor) from a
 * callable block} that may capture objects for later closing. To start a pipeline from a {@link