Add platform-specific installation guides: GKE, AKS, Huawei Cloud CCE, Tencent TKE, IBM Cloud, OpenShift, etc.

オープン
  • 何か問題がありますか？
#795 コメント 23 件 リアクション 0 件 担当者 3 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

@moezdil がすでに取り組んでいます。

2026年11月3日 から。

  • #804 @Junie06 による — マージされずにクローズ

評価

この issue はまだ評価されていません。

説明

area/docs good first issue kind/documentation kind/feature

Description

The Installation section currently contains only one platform-specific guide — HAMi on AWS — and that page mainly covers the AWS Marketplace install. There are no equivalent production deployment guides for the other major managed-Kubernetes and self-managed platforms where HAMi users commonly run GPU workloads.

Current state

docs/installation/ today:

  • online-installation.md, offline-installation.md, prerequisites.md, upgrade.md, uninstall.md, webui-installation.md — generic, platform-agnostic
  • aws-installation.md — the only platform guide (Marketplace-focused)
  • how-to-use-hami-dra.md, how-to-use-volcano-vgpu.md, how-to-use-volcano-ascend.md, configure-cdi.md — feature-specific

The generic Helm docs do not cover platform differences that matter in production: GPU driver provisioning on GPU node pools, conflicts with the platform's own NVIDIA device plugin (double registration), node taints/tolerations, image registry access/mirrors, and platform security constraints (e.g. OpenShift SecurityContextConstraints).

Proposed guides

  • Google Cloud (GKE)
  • Microsoft Azure (AKS)
  • Huawei Cloud (CCE)
  • Tencent Cloud (TKE)
  • IBM Cloud (IKS / Red Hat OpenShift on IBM Cloud)
  • Red Hat OpenShift (self-managed)
  • Other candidates by demand: Alibaba Cloud (ACK), Volcano-based platform distributions, etc.

What each guide should cover

Following the structure of aws-installation.md, but production-oriented:

  1. Prerequisites — GPU node pool creation, NVIDIA driver management options on that platform, required CLI tools and permissions.
  2. Platform-specific install — Helm values differences, nodeSelector/tolerations for GPU nodes, and how to disable/avoid the platform's default GPU device plugin so HAMi's device plugin is the only one registering GPUs.
  3. Production considerations — scheduler HA, image pull via platform registries/mirrors, offline/air-gapped install where applicable.
  4. Verification — expected pod states plus a smoke-test workload that requests a shared vGPU.
  5. Platform gotchas & troubleshooting pointers — e.g. driver pre-installed by the cloud vendor, MIG support caveats on managed control planes, SCC requirements on OpenShift.

Each new page also needs a sidebars.js entry and a Chinese mirror under i18n/zh/docusaurus-plugin-content-docs/current/.

Use Case

Users deploying HAMi on GKE / AKS / CCE / TKE / OpenShift currently have no official guidance and hit predictable, platform-specific problems (device plugin conflicts, driver provisioning, SCCs). Official per-platform guides would reduce repeated community questions and make HAMi a credible option for production deployment beyond AWS.

Related

  • #689 — RFC: Reorganize the HAMi docs
  • #656 — [LFX Mentorship] Expand HAMi GPU-sharing tutorials, labs, and troubleshooting guides
主要言語
JavaScript
スター
11
フォーク
111
平均マージ
2日 1時間
マージ済み PR（30日）
70

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

Project-HAMi/website のほかの issue

Project-HAMi/website の issue をすべて見る

似ている issue

JavaScript の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。