[Request]: container cluster create: kubeadm config patches, extra mounts, and node labels

オープン
#2,498 コメント 1 件 リアクション 2 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
おおむね明確
活発さ
活発
技術スタック
kubernetes, swift

調査の方向性

Start with the container cluster create implementation and the related k8s plugin work in #2043; the issue does not name source files or tests. Compare the existing create options with the requested kubeadm patches, node mounts, labels, and readiness wait, then determine which configuration approach the project supports. Done means the selected create-time options are implemented and documented, with coverage for their behavior.

索引モデルが issue の本文から書いたものです。

説明

Feature or enhancement request details

I've been trying container cluster create as a drop-in for Kind in a tool that spins up throwaway clusters to compare Kubernetes schedulers. Most of it maps over nicely (--node-image, --workers, write-config, --shared-image-store), but there are three Kind config features I couldn't find a way to express:

  • kubeadm config patches (Kind's kubeadmConfigPatches)
  • extra mounts into a node (Kind's per-node extraMounts)
  • node labels (Kind's per-node labels)

The concrete case: running kube-scheduler with a custom KubeSchedulerConfiguration. With Kind, that's a file mounted into the control-plane node plus a ClusterConfiguration patch that adds --config to the scheduler:

kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
  - role: control-plane
    extraMounts:
      - hostPath: ./kube-scheduler.yaml
        containerPath: /etc/kubernetes/scheduler-config.yaml
        readOnly: true
  - role: worker
    labels:
      example/role: system
kubeadmConfigPatches:
  - |
    apiVersion: kubeadm.k8s.io/v1beta4
    kind: ClusterConfiguration
    scheduler:
      extraArgs:
        - name: config
          value: /etc/kubernetes/scheduler-config.yaml
      extraVolumes:
        - name: scheduler-config
          hostPath: /etc/kubernetes/scheduler-config.yaml
          mountPath: /etc/kubernetes/scheduler-config.yaml
          readOnly: true
          pathType: File

Workaround today (container 1.7.1): after cluster create, container exec the config file into the node and rewrite /etc/kubernetes/manifests/kube-scheduler.yaml so kubelet restarts the static pod with --config. It works — /configz shows the custom profiles and the scheduler is Ready again in ~25s — but it restarts a control-plane component after bootstrap and relies on kubeadm's manifest layout. Labels are easy enough to add afterwards with kubectl label node, but mounts and patches really want to happen at create time.

What would help, either of:

  1. The --config <path> option mentioned under "Future Work" in #2043 — ideally accepting Kind's kind.x-k8s.io/v1alpha4 Cluster schema (or a documented subset of it), which would make porting existing Kind setups close to free.
  2. Or individual flags, something like:
    container cluster create --name lab \
      --kubeadm-config-patch ./scheduler-patch.yaml \
      --mount ./kube-scheduler.yaml:/etc/kubernetes/scheduler-config.yaml:ro \
      --workers count=1,label=example/role=system
    

A smaller related nicety: a --wait <duration> for node readiness, like kind create cluster --wait 180s, so scripts don't need a separate kubectl wait.

Related: #2043 (k8s plugin), #2468 (network/MTU options for cluster create), #2382 (skipping kube-proxy).

Code of Conduct
  • I agree to follow this project's Code of Conduct
主要言語
Swift
スター
51.9k
フォーク
1.8k
平均マージ
2日 3時間
マージ済み PR(30日)
27

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。

apple/container のほかの issue

apple/container の issue をすべて見る

似ている issue

Swift 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。