Using kubeplay to deploy a kubernetes cluster offline

kubeplay is a tool for offline deployment of kuberneres clusters based on kubespray

Features

• Includes all dependencies and can be installed offline with a single command
• Support amd64 and arm64 CPU architectures
• kubeadm-generated certificate validity adjusted to 10 years
• De-dockerized deployment, seamless migration to containerd as container runtime
• Offline installation of rpm/deb packages that the platform depends on (e.g. storage client) for toB privatization scenarios
• Multi-cluster deployment, with support for deploying kubernetes clusters as Job Pods within kubernetes clusters
• Build offline installers using GitHub Actions, no membership required, 100% open source 100% free

Component Versions

Support OS

compose

Use nerdctl compose on the deployment tool run node to start the nginx and registry containers, which provide offline resource download and image distribution services, respectively.

kubespray

Use the kubernetes community’s kubespray as a cluster deployment feature, and the resources relied on during deployment are fetched from the compose node.

Deployment

Download

Download the corresponding offline installer from the GitHub release page k8sli/kubeplay/releases

1
2
3

sha256sum                                    # Installation package sha256sum checksum file
kubeplay-v0.1.0-alpha.1-linux-amd64.tar.gz   # For amd64 CPU architecture
kubeplay-v0.1.0-alpha.1-linux-arm64.tar.gz   # Suitable for arm64 CPU architecture

Unpacking

1
2
3

$ tar -xpf kubeplay-v0.1.0-alpha.1-linux-amd64.tar.gz
$ cd kubeplay
$ vi config.yaml

Configuration

The config.yaml configuration file is divided into the following sections

• compose: nginx and registry deployment node information
• kubespray: kubespray deployment configuration
• invenory: kubernetes cluster node ssh login information
• default: some default parameters

compose

1
2
3
4
5
6
7

compose:
  # Compose bootstrap node ip, default is local internal ip
  internal_ip: 172.20.0.25
  # Nginx http server bind port for download files and packages
  nginx_http_port: 8080
  # Registry domain for CRI runtime download images
  registry_domain: kube.registry.local

kubespray

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

kubespray:
  # Kubernetes version by default, only support v1.20.6
  kube_version: v1.21.3
  # For deploy HA cluster you must configure a external apiserver access ip
  external_apiserver_access_ip: 127.0.0.1
  # Set network plugin to calico with vxlan mode by default
  kube_network_plugin: calico
  #Container runtime, only support containerd if offline deploy
  container_manager: containerd
  # Now only support host if use containerd as CRI runtime
  etcd_deployment_type: host
  # Settings for etcd event server
  etcd_events_cluster_setup: true
  etcd_events_cluster_enabled: true

inventory

inventory is the ssh login configuration for kubernetes cluster nodes, supporting yaml, json, and ini formats.

• yaml format

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

  # Cluster nodes inventory info inventory: all: vars: ansible_port: 22 ansible_user: root ansible_ssh_pass: Password # ansible_ssh_private_key_file: /kubespray/config/id_rsa hosts: node1: ansible_host: 172.20.0.21 node2: ansible_host: 172.20.0.22 node3: ansible_host: 172.20.0.23 node4: ansible_host: 172.20.0.24 children: kube_control_plane: hosts: node1: node2: node3: kube_node: hosts: node1: node2: node3: node4: etcd: hosts: node1: node2: node3: k8s_cluster: children: kube_control_plane: kube_node: gpu: hosts: {} calico_rr: hosts: {}

• json format

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

  inventory: | { "all": { "vars": { "ansible_port": 22, "ansible_user": "root", "ansible_ssh_pass": "Password" }, "hosts": { "node1": { "ansible_host": "172.20.0.21" }, "node2": { "ansible_host": "172.20.0.22" }, "node3": { "ansible_host": "172.20.0.23" }, "node4": { "ansible_host": "172.20.0.24" } }, "children": { "kube_control_plane": { "hosts": { "node1": null, "node2": null, "node3": null } }, "kube_node": { "hosts": { "node1": null, "node2": null, "node3": null, "node4": null } }, "etcd": { "hosts": { "node1": null, "node2": null, "node3": null } }, "k8s_cluster": { "children": { "kube_control_plane": null, "kube_node": null } }, "gpu": { "hosts": {} }, "calico_rr": { "hosts": {} } } } }

• ini format

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

  inventory: | [all:vars] ansible_port=22 ansible_user=root ansible_ssh_pass=Password #ansible_ssh_private_key_file=/kubespray/config/id_rsa [all] kube-control-01 ansible_host=172.20.0.21 kube-control-02 ansible_host=172.20.0.23 kube-control-03 ansible_host=172.20.0.22 kube-node-01 ansible_host=172.20.0.24 [bastion] # bastion-01 ansible_host=x.x.x.x ansible_user=some_user [kube_control_plane] kube-control-01 kube-control-02 kube-control-03 [etcd] kube-control-01 kube-control-02 kube-control-03 [kube_node] kube-control-01 kube-control-02 kube-control-03 kube-node-01 [calico_rr] [k8s_cluster:children] kube_control_plane kube_node calico_rr

default

The following default parameters are not recommended to be changed without special requirements, just leave them as default. The ntp_server parameter is automatically replaced by the internal_ip value in the compose; the registry_ip and offline_resources_url parameters are automatically generated according to the parameters in the compose without modification.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22

default:
  # NTP server ip address or domain, default is internal_ip
  ntp_server:
    - internal_ip
  # Registry ip address, default is internal_ip
  registry_ip: internal_ip
  # Offline resource url for download files, default is internal_ip:nginx_http_port
  offline_resources_url: internal_ip:nginx_http_port
  # Use nginx and registry provide all offline resources
  offline_resources_enabled: true
  # Image repo in registry
  image_repository: library
  # Kubespray container image for deploy user cluster or scale
  kubespray_image: "kubespray"
  # Auto generate self-signed certificate for registry domain
  generate_domain_crt: true
  # For nodes pull image, use 443 as default
  registry_https_port: 443
  # For push image to this registry, use 5000 as default, and only bind at 127.0.0.1
  registry_push_port: 5000
  # Set false to disable download all container images on all nodes
  download_container: false

Deploy Cluster

1

$ bash install.sh

Add Node

1

$ bash install.sh add-node $NODE_NAMES

Delete Node

1

$ bash install.sh remove-node $NODE_NAME

Remove Cluster

1

$ bash install.sh remove-cluster

Remove all components

1

$ bash install.sh remove

Unfinished follow-up additions

TODO

Refenrece https://blog.k8s.li/deploy-k8s-by-kubeplay.html