rke安装k8s集群

本是古典何须时尚 2022-12-23 07:05 244阅读 0赞

# #

# 1 节点规划 #

构建集群的第一步是将拥有的服务器按节点功能进行划分，下面是节点规划情况。

<table style="width:500px;"> 
 <tbody> 
  <tr> 
   <td>IP</td> 
   <td>角色</td> 
  </tr> 
  <tr> 
   <td>192.168.120.11</td> 
   <td>部署节点，master，api-server，etcd，scheduler，controller-manager&nbsp;</td> 
  </tr> 
  <tr> 
   <td>192.168.120.12</td> 
   <td>master，api-server，etcd，scheduler，controller-manager&nbsp;</td> 
  </tr> 
  <tr> 
   <td>192.168.120.13</td> 
   <td>master，api-server，etcd，scheduler，controller-manager&nbsp;</td> 
  </tr> 
  <tr> 
   <td>192.168.120.14</td> 
   <td>work，kubelet，kube-proxy</td> 
  </tr> 
  <tr> 
   <td>192.168.120.15</td> 
   <td>work，kubelet，kube-proxy</td> 
  </tr> 
 </tbody> 
</table>

规划说明：

1.  单独选择了一台机器`192.168.120.11`作为部署节点。如果机器数不多，可以将部署节点加入到 k8s 集群中。本次部署该机器作为master节点
2.  为了保证可用性，择了三台机器部署 k8s master 组件。如果有条件，可以将 etcd 和 master 中的其他组件分开部署，这样可以根据需要更灵活地控制实例个数。
3.  剩余2台机器作为 k8s worker 节点。节点个数需要根据实际情况动态调整。

# 2 环境准备 #

在完成节点规划后，需要进行环境准备工作，主要包含以下内容：

## 2.1 安装 RKE ##

\-需要在部署节点（192.168.120.11）上安装 RKE 二进制包，具体安装方法可参考 [download-the-rke-binary][]，下载后，可以查看rke可以安装的k8s的版本是多少：

#查看rke的版本
    [root@localhost rke]# ./rke -v
    rke version v1.0.14
    
    #查看rke支持安装的k8s版本
    [root@localhost rke]# ./rke config --system-images --all
    INFO[0000] Generating images list for version [v1.16.15-rancher1-2]: 
    rancher/coreos-etcd:v3.3.15-rancher1
    rancher/rke-tools:v0.1.65
    rancher/k8s-dns-kube-dns:1.15.0
    rancher/k8s-dns-dnsmasq-nanny:1.15.0
    rancher/k8s-dns-sidecar:1.15.0
    rancher/cluster-proportional-autoscaler:1.7.1
    rancher/coredns-coredns:1.6.2
    rancher/k8s-dns-node-cache:1.15.7
    rancher/hyperkube:v1.16.15-rancher1
    rancher/coreos-flannel:v0.12.0
    rancher/flannel-cni:v0.3.0-rancher6
    rancher/calico-node:v3.13.4
    rancher/calico-cni:v3.13.4
    rancher/calico-kube-controllers:v3.13.4
    rancher/calico-ctl:v3.13.4
    rancher/calico-pod2daemon-flexvol:v3.13.4
    weaveworks/weave-kube:2.6.4
    weaveworks/weave-npc:2.6.4
    rancher/pause:3.1
    rancher/nginx-ingress-controller:nginx-0.35.0-rancher1
    rancher/nginx-ingress-controller-defaultbackend:1.5-rancher1
    rancher/metrics-server:v0.3.4
    INFO[0000] Generating images list for version [v1.17.14-rancher1-1]: 
    rancher/coreos-etcd:v3.4.3-rancher1
    rancher/rke-tools:v0.1.66
    rancher/k8s-dns-kube-dns:1.15.0
    rancher/k8s-dns-dnsmasq-nanny:1.15.0
    rancher/k8s-dns-sidecar:1.15.0
    rancher/cluster-proportional-autoscaler:1.7.1
    rancher/coredns-coredns:1.6.5
    rancher/k8s-dns-node-cache:1.15.7
    rancher/hyperkube:v1.17.14-rancher1
    rancher/coreos-flannel:v0.12.0
    rancher/flannel-cni:v0.3.0-rancher6
    rancher/calico-node:v3.13.4
    rancher/calico-cni:v3.13.4
    rancher/calico-kube-controllers:v3.13.4
    rancher/calico-ctl:v3.13.4
    rancher/calico-pod2daemon-flexvol:v3.13.4
    weaveworks/weave-kube:2.6.4
    weaveworks/weave-npc:2.6.4
    rancher/pause:3.1
    rancher/nginx-ingress-controller:nginx-0.35.0-rancher2
    rancher/nginx-ingress-controller-defaultbackend:1.5-rancher1
    rancher/metrics-server:v0.3.6
    INFO[0000] Generating images list for version [v1.15.12-rancher2-3]: 
    rancher/coreos-etcd:v3.3.10-rancher1
    rancher/rke-tools:v0.1.58
    rancher/k8s-dns-kube-dns:1.15.0
    rancher/k8s-dns-dnsmasq-nanny:1.15.0
    rancher/k8s-dns-sidecar:1.15.0
    rancher/cluster-proportional-autoscaler:1.3.0
    rancher/coredns-coredns:1.3.1
    rancher/k8s-dns-node-cache:1.15.7
    rancher/hyperkube:v1.15.12-rancher2
    rancher/coreos-flannel:v0.12.0
    rancher/flannel-cni:v0.3.0-rancher6
    rancher/calico-node:v3.13.4
    rancher/calico-cni:v3.13.4
    rancher/calico-kube-controllers:v3.13.4
    rancher/calico-ctl:v3.13.4
    rancher/calico-pod2daemon-flexvol:v3.13.4
    weaveworks/weave-kube:2.6.4
    weaveworks/weave-npc:2.6.4
    rancher/pause:3.1
    rancher/nginx-ingress-controller:nginx-0.32.0-rancher1
    rancher/nginx-ingress-controller-defaultbackend:1.5-rancher1
    rancher/metrics-server:v0.3.3

## 2.2 配置 SSH 免密登录 ##

由于 RKE 通过 SSH tunnel 安装部署 k8s 集群，需要配置 RKE 所在节点到 k8s 各节点的 SSH 免密登录。如果 RKE 所在节点也需要加入到 k8s 集群中，需要配置到本机的 SSH 免密登录。

#创建普通用户
    [root@localhost rke]# useradd rancher
    [root@localhost rke]# echo "password" | passwd --stdin rancher
    [root@localhost rke]# usermod rancher -G docker 
    
    #配置执行rke up的用户到rancher@各个主机上免密。
    [root@localhost rke]# ssh-copy-id -i rancher@192.168.120.11
    [root@localhost rke]# ssh-copy-id -i rancher@192.168.120.12
    [root@localhost rke]# ssh-copy-id -i rancher@192.168.120.13
    [root@localhost rke]# ssh-copy-id -i rancher@192.168.120.14
    [root@localhost rke]# ssh-copy-id -i rancher@192.168.120.15

## 2.3安装 docker ##

由于 RKE 通过 docker 镜像`rancher/hyperkube`启动 k8s 组件，因此需要在 k8s 集群的各个节点（192.168.120.11 ~ 192.168.120.15 这 5 台机器）上安装 docker。注意docker和k8s的匹配版本。

并且rancher用户要有docker的权限，上述我们已经把rancher用户加到docker的用户组中去了。

## 2.4 关闭 swap，防火墙等 ##

k8s 1.8 开始要求关闭系统的 swap，如果不关闭，默认配置下 kubelet 将无法启动。这里需要关闭所有 k8s worker 节点的 swap。

# 3 安装步骤 #

## 3.1 准备cluster.yml文件 ##

[cluster.yml文件官网配置][cluster.yml]

nodes:
      - address: 192.168.120.11
        user: rancher
        role:
          - controlplane
          - etcd
        labels:
          isNode: true
      - address: 192.168.120.12
        user: rancher
        role:
          - controlplane
          - etcd
        labels:
          isNode: true
      - address: 192.168.120.13
        user: rancher
        role:
          - controlplane
          - etcd
        labels:
          isNode: true
      - address: 192.168.120.14
        user: rancher
        role:
          - worker
        labels:
          isNode: true
      - address: 192.168.120.15
        user: rancher
        role:
          - worker
        labels:
          isNode: true
    kubernetes_version: "v1.17.14-rancher1-1"
    private_registries:
      - url: 192.168.120.16
        user: admin
        password: 123456
    cluster_name: test
    services:
      kube-api:
        service_cluster_ip_range: 100.66.0.0/16
        extra_args:
          enable-admission-plugins: "ServiceAccount,NamespaceLifecycle,MutatingAdmissionWebhook,ValidatingAdmissionWebhook,LimitRanger,PersistentVolumeLabel,DefaultStorageClass,ResourceQuota,DefaultTolerationSeconds"
      kube-controller:
        cluster_cidr: 100.86.0.0/16
        service_cluster_ip_range: 100.66.0.0/16    
      kubelet:
        cluster_dns_server: 100.66.0.10
      kubeproxy:
        extra_args:
          proxy-mode: ipvs
          masquerade-all: true
    network:
      plugin: flannel
    dns:
      provider: coredns
      upstreamnameservers:
      - 255.255.255.55
      - 255.255.254.55
    ingress:
      provider: none

## 3.2在rke目录下生成文件 ##

rke up --config cluster.yml

![20201208173358654.png][]

## 3.3 kubectl自动补全 ##

其他的master节点要执行kubectl指令，需要将rke目录拷贝到master主机上

kubectl --kubeconfig=/opt/rke/kube_config_cluster.yml completion bash

[download-the-rke-binary]: https://rancher.com/docs/rke/latest/en/installation/#download-the-rke-binary
[cluster.yml]: https://rancher.com/docs/rke/latest/en/example-yamls/
[20201208173358654.png]: /images/20221120/617a6e473ac648d298275cac32b4e8cd.png