# 动态资源分配

LLMS index: [llms.txt](/llms.txt)

---

<!--
reviewers:
- klueska
- pohly
title: Dynamic Resource Allocation
content_type: concept
weight: 20
aliases:
- /docs/concepts/scheduling-eviction/dynamic-resource-allocation/
-->

<!-- overview -->



















<div class="feature-state-notice feature-stable" title=" 特性门控： DynamicResourceAllocation">
  <span class="feature-state-name">特性状态：</span>
  <span class="feature-state-details">
    <span class="feature-state-stage">GA</span> since Kubernetes v1.35
    </span>
</div>


<div class="feature-stable">
  
  <details>
    <summary>More information about this feature</summary>
    <p>This is a stable feature in Kubernetes, and has been since version v1.35. It was first available in the v1.30 release. You can no longer disable or opt out of this feature or behavior (it is locked); if you explicitly set a value for the associated feature gate <tt>DynamicResourceAllocation</tt>, Kubernetes ignores it but does not report any error.</p>
  </details>

  
  </div>


<!--
This section introduces _dynamic resource allocation (DRA)_ in Kubernetes.
-->
本节介绍 Kubernetes 中的**动态资源分配（Dynamic Resource Allocation，DRA）**。

<!--
title: Dynamic Resource Allocation
id: dra
full_link: /docs/concepts/scheduling-eviction/dynamic-resource-allocation/
short_description: >
  A Kubernetes feature for requesting and sharing resources, like hardware
  accelerators, among Pods.

aka:
- DRA
tags:
- extension
-->
<!--
A Kubernetes feature that lets you request and share resources among Pods.
These resources are often attached
<a class='glossary-tooltip' title='直接或间接挂接到集群节点上的所有资源，例如 GPU 或电路板。' data-bs-toggle='tooltip' data-bs-placement='top' href='/zh-cn/docs/reference/glossary/?all=true#term-device' target='_blank' aria-label='devices'>devices</a> like hardware
accelerators.
-->
<p><p>DRA 是 kubernetes 提供的一项特性，允许你在多个 Pod 之间请求和共享资源。<br>
这些资源通常是挂接的<a class='glossary-tooltip' title='直接或间接挂接到集群节点上的所有资源，例如 GPU 或电路板。' data-bs-toggle='tooltip' data-bs-placement='top' href='/zh-cn/docs/reference/glossary/?all=true#term-device' target='_blank' aria-label='设备'>设备</a>，例如硬件加速器。</p></p>
<!--
With DRA, device drivers and cluster admins define device _classes_ that are
available to _claim_ in workloads. Kubernetes allocates matching devices to
specific claims and places the corresponding Pods on nodes that can access the
allocated devices.
-->
<p>借助 DRA，设备驱动和集群管理员可以定义设备的<strong>类别</strong>，这些类别可供工作负载中的 Pod 申领。
Kubernetes 会将匹配的设备分配给特定的申领，并将相应的 Pod 调度到能够访问这些已分配设备的节点上。</p>

<!--
Allocating resources with DRA offers a similar experience to
[dynamic volume provisioning](/docs/concepts/storage/dynamic-provisioning/),
in which you use PersistentVolumeClaims to _claim_ storage capacity from storage classes,
and request the claimed capacity for use in your Pods.
-->
使用 DRA 分配资源提供了与[动态卷制备](/zh-cn/docs/concepts/storage/dynamic-provisioning/)类似的体验：
你使用 PersistentVolumeClaims 从存储类中**申领**存储容量，
并在 Pod 中请求使用该已申领的容量。

<!-- body -->

<!--
### Benefits of DRA {#dra-benefits}
-->
### DRA 的优势 {#dra-benefits}

<!--
DRA provides a flexible way to categorize, request, and use devices in your cluster.
Using DRA provides benefits like the following:

* **Flexible device filtering**: use common expression language (CEL) to perform
  fine-grained filtering for specific device attributes.
* **Device sharing**: share the same resource with multiple containers or Pods
  by referencing the corresponding resource claim.
* **Device configuration**: attach vendor-specific device configurations to your resource claim, enabling per-workload device configuration, rather than today's per-node device configuration
* **Centralized device categorization**: device drivers and cluster admins can
  use device classes to provide app operators with hardware categories that are
  optimized for various use cases. For example, you can create a cost-optimized
  device class for general-purpose workloads, and a high-performance device
  class for critical jobs.
* **Simplified Pod requests**: with DRA, app operators don't need to specify
  device quantities in Pod resource requests. Instead, the Pod references a
  resource claim, and the device configuration in that claim applies to the Pod.
-->
DRA 提供了一种灵活的方式来分类、请求和使用集群中的设备。
使用 DRA 具有以下优势：

* **灵活的设备过滤：** 使用公共表达式语言（CEL）对特定设备属性进行细粒度过滤。
* **设备共享：** 通过引用相应的资源申领，与多个容器或 Pod 共享同一资源。
* **设备配置：** 将特定于供应商的设备配置附加到你的资源申领，
  启用按工作负载的设备配置，而非如今的按节点设备配置。
* **集中式设备分类：** 设备驱动和集群管理员可以使用 DeviceClass
  为应用运维人员提供针对各种用例优化的硬件类别。
  例如，你可以为通用工作负载创建成本优化的 DeviceClass，
  为关键作业创建高性能的 DeviceClass。
* **简化的 Pod 请求：** 通过 DRA，应用运维人员无需在 Pod 资源请求中指定设备数量。
  相反，Pod 引用一个资源申领，该申领中的设备配置将应用于该 Pod。

<!--
These benefits provide significant improvements in the device allocation
workflow when compared to
[device plugins](/docs/concepts/extend-kubernetes/compute-storage-net/device-plugins/),
which require per-container device requests, don't support device sharing, and
don't support expression-based device filtering.
-->
与需要按容器请求设备、不支持设备共享且不支持基于表达式的设备过滤的
[设备插件](/zh-cn/docs/concepts/extend-kubernetes/compute-storage-net/device-plugins/)相比，
这些优势在设备分配工作流方面提供了显著改进。

<!--
### Types of DRA users {#dra-user-types}
-->
### DRA 用户类型 {#dra-user-types}

<!--
The workflow of using DRA to allocate devices involves the following types of users:
-->
使用 DRA 分配设备的工作流涉及以下类型的用户：

<!--
* **Device owner**: responsible for devices. Device owners might be commercial
  vendors, the cluster operator, or another entity. To use DRA, devices must
  have DRA-compatible drivers that do the following:

  * Create ResourceSlices that provide Kubernetes with information about
    nodes and resources.
  * Update ResourceSlices when resource capacity in the cluster changes.
  * Configure devices according to the claim, and attach them to containers via Container Device Interface (CDI).
  * Optionally, create DeviceClasses that workload operators can use to
    claim devices.
-->
* **设备所有者（Device owner）：** 负责设备。设备所有者可以是商业供应商、集群运维人员或其他实体。
  要使用 DRA，设备必须具备兼容 DRA 的驱动，这些驱动需要完成以下工作：

  * 创建 ResourceSlice，为 Kubernetes 提供关于节点和资源的信息。
  * 当集群中的资源容量发生变化时更新 ResourceSlice。
  * 根据申领配置设备，并通过容器设备接口（Container Device Interface，CDI）
    将它们附加到容器。
  * 可选地，创建工作负载运维人员可用于申领设备的 DeviceClass。

<!--
* **Cluster admin**: responsible for configuring clusters and nodes,
  attaching devices, installing drivers, and similar tasks. To use DRA,
  cluster admins do the following:

  * Attach devices to nodes.
  * Install device drivers that support DRA.
  * Optionally, create DeviceClasses that workload operators can use to claim devices.
-->
* **集群管理员（Cluster admin）：** 负责配置集群和节点、附加设备、安装驱动等类似任务。
  要使用 DRA，集群管理员需要完成以下工作：

  * 将设备附加到节点。
  * 安装支持 DRA 的设备驱动。
  * 可选地，创建工作负载运维人员可用于申领设备的 DeviceClass。

<!--
* **Workload operator**: responsible for deploying and managing workloads in the
  cluster. To use DRA to allocate devices to Pods, workload operators do the following:

  * Create ResourceClaims or ResourceClaimTemplates to request specific
    configurations within DeviceClasses.
  * Deploy workloads that use specific ResourceClaims or ResourceClaimTemplates.
-->
* **工作负载运维人员（Workload operator）：** 负责在集群中部署和管理工作负载。
  要使用 DRA 为 Pod 分配设备，工作负载运维人员需要完成以下工作：

  * 创建 ResourceClaim 或 ResourceClaimTemplate，以在 DeviceClass 中请求特定配置。
  * 部署使用特定 ResourceClaim 或 ResourceClaimTemplate 的工作负载。

<!--
## Limitations
-->
## 限制

<!--
* The Kubernetes scheduler doesn't support
  [preemption](/docs/concepts/scheduling-eviction/pod-priority-preemption/) for
  DRA resources. This means that an existing Pod that's running on a node and is
  using DRA resources can't be preempted by a higher-priority Pod that also needs
  DRA resources. The high-priority Pod will remain in a pending state until the device
  becomes available, which happens when the conflicting Pod terminates or is
  manually deleted.
-->
* Kubernetes 调度器不支持为 DRA 资源进行
  [抢占](/zh-cn/docs/concepts/scheduling-eviction/pod-priority-preemption/)。
  这意味着在节点上运行并使用 DRA 资源的现有 Pod，
  不能被同样需要 DRA 资源的更高优先级 Pod 抢占。
  高优先级 Pod 将保持挂起状态，直到设备可用 —— 这发生在冲突 Pod 终止或被手动删除时。

## 接下来

<!--
- [Set Up DRA in a Cluster](/docs/tasks/configure-pod-container/assign-resources/set-up-dra-cluster/)
- [Allocate devices to workloads using DRA](/docs/tasks/configure-pod-container/assign-resources/allocate-devices-dra/)
- [Access DRA device metadata](/docs/tasks/configure-pod-container/assign-resources/access-dra-device-metadata/)
- For more information on the design, see the
  [Dynamic Resource Allocation with Structured Parameters](https://github.com/kubernetes/enhancements/tree/master/keps/sig-node/4381-dra-structured-parameters)
  KEP.
-->
- [在集群中设置 DRA](/zh-cn/docs/tasks/configure-pod-container/assign-resources/set-up-dra-cluster/)
- [使用 DRA 为工作负载分配设备](/zh-cn/docs/tasks/configure-pod-container/assign-resources/allocate-devices-dra/)
- [访问 DRA 设备元数据](/zh-cn/docs/tasks/configure-pod-container/assign-resources/access-dra-device-metadata/)
- 有关设计的更多信息，请参阅
  [使用结构化参数的动态资源分配](https://github.com/kubernetes/enhancements/tree/master/keps/sig-node/4381-dra-structured-parameters)
  KEP。

---

Section pages:

- [DRA API 对象](/zh-cn/docs/concepts/resource-management/dynamic-resource-allocation/dra-api/)
- [DRA 的工作原理](/zh-cn/docs/concepts/resource-management/dynamic-resource-allocation/how-dra-works/)
- [DRA 特性](/zh-cn/docs/concepts/resource-management/dynamic-resource-allocation/dra-features/)
- [设备污点与容忍度](/zh-cn/docs/concepts/resource-management/dynamic-resource-allocation/device-taints/)