英伟达发布 NVKE 预览版 在自有 GPU 上托管 Kubernetes

原标题:Nvidia Kubernetes Engine (NVKE)
62AI 研判分

英伟达推出 Kubernetes Engine(NVKE)预览版,提供可安装在自有 GPU 基础设施上的托管 Kubernetes 管理平面,每个工作负载集群拥有独立控制面、etcd 与节点池。该产品目前处于预览阶段,官方明确不建议用于生产环境,API 与行为可能随版本变化。

报道正文

本文目录3 个章节

NVIDIA Kubernetes Engine (NVKE)

Run a managed Kubernetes service on your own NVIDIA GPU infrastructure. NVIDIA Kubernetes® Engine (NVKE) provides the management plane that platform teams install to create and operate Kubernetes clusters for their customers. Each workload cluster has its own Kubernetes control plane, etcd and node pools. You operate the NVKE management plane; your customers use their clusters through the Kubernetes API and kubectl.

Preview: NVKE is available as a preview. APIs, configuration and behavior can change between releases, and we do not recommend NVKE for production workloads. See Preview status and known limitations.

Get started · Documentation · Contribute · Get help

Getting started

Your goalStart hereWhat you get
Evaluate without GPU hardwareLocal quickstartA development management plane and a workload cluster on simulated machines. No NVIDIA account is required.
Install on your infrastructurePlan your installationThe prerequisites and sequence for preparing nodes, bootstrapping a management cluster and installing its services.
Build or contributeBuild from sourceThe NVKE binaries and the contributor development loop.

The local quickstart runs on a Linux host or VM, amd64 or arm64, with at least 6 CPUs, 12 GiB of free memory and 100 GiB of free disk. It requires Docker, virtual TPM support and the tools listed in the guide. Follow its setup through workload-cluster creation and confirm the cluster's nodes report Ready. On macOS, use the local site guide, which runs the evaluation inside a Colima virtual machine.

For a real installation, you supply the machines, networking, node images and registries, and configure access to NVIDIA Infra Controller (NICo). NICo is the sole compatible infrastructure provider at preview. The platform and version reference lists validated combinations and pinned dependencies.

What NVKE does

  • Creates independent workload clusters. Request clusters through the NVKE API or nvkectl; NVKE provisions machines through NICo and forms Kubernetes clusters with their own control planes.
  • Manages workload cluster lifecycle. Create, scale and delete node pools, upgrade Kubernetes and replace nodes. See the node pool guide for the manual steps and preview limits on replacement and upgrades.
  • Establishes node identity. Nodes attest through SPIFFE/SPIRE using their TPM before receiving join credentials. Encryption of etcd data is available per cluster.
  • Connects builds to deployment. Its companion project, NVIDIA Kubernetes Recipes (NVKR), builds node images and NVKE components from pinned inputs. NVKE deploys those artifacts; NVKR owns their build and release evidence.

The preview needs explicit operator configuration: API authentication is off and the API uses plain HTTP after installation. Configure API authentication and protected access before admitting callers. The management cluster does not upgrade, scale or replace itself, and there is no management-plane upgrade or backup/restore procedure. FDE node images are for evaluation and provide no data-at-rest confidentiality. Read the known limitations before choosing an evaluation or installation path.

Overview

The diagram shows the NVKE product family, how it sits on NICo and your GPU machines, and which components talk to each other.

NVKE product family. NVKR builds and signs artifacts and publishes them to a registry. The operator installs the management cluster's components, the NVKE management plane, CAPNICo and NVKE observability, from the charts and images in the registry. The management plane requests machines through CAPNICo, which calls NICo, and drives each workload cluster. NICo provisions NVIDIA GPU machines from the node OS image, and those machines become the workload cluster's Kubernetes nodes, each running nvke-agent, under the cluster's Kubernetes control plane. The accelerated runtime produced by AICR and NVKE observability are optional in the workload cluster. The customer uses kubectl against the control plane.

NVKE is not a Kubernetes distribution and not a Cluster API flavour. It is a self-hosting factory for Kubernetes clusters: a throwaway seed cluster stands up a management cluster, and the management plane on it then provisions and lifecycle-manages many isolated workload clusters on demand.

An installation has two kinds of Kubernetes cluster:

  • The management cluster runs the NVKE management plane and infrastructure controllers. nvkectl forms it using a temporary seed cluster; installing the management-plane services is a separate step.
  • Workload clusters run customer applications. The management plane creates and operates them through NICo, Cluster API and the node agent.

Read What is NVKE for the operator workflow and Architecture for the implementation and data flows.

(注:更多技术实现细节与完整 API 文档请查阅原项目 README)