白话kubeadm安装K8S[集群](史上最K8S搭建教程)

前言

市面上很多k8s的安装工具，作为产品的设计者和推广者，K8S组织也知道自己的产品部署起来十分的困难，于是把开源爱好者写的工具kubeadmn收编为正规军，纳入到了自己的麾下。(关于kubeadm的博客，你可以看这篇《kubeadm探秘》)

为什么我们要用kubeadm来部署？因为kubeadm不仅直接相关的命令简单到只有两条，而且还可以放生产环境使用（这里有个前提，需要能很好的理解K8S的各个组件，处理好它们的关系，说人话就是能看得懂、玩得转）。

官方文档有中文教程，K8S最新版本1.26已经弃用了docker做自己的运行时，笔者还没有摸索出来怎么部署，这里就以老版本的1.18为例子来讲解

我的演示环境

系统：centos7.6

CPU：2核

内存：2G

最好是2核4G，20G硬盘，如果你想模拟更多的生产环境过程部署，比如jenkins、nginx、MySQL等，最好提升一下虚拟机配置，否则可能无法运行那么多的pod。

因为是演示环境，所以这里操作的是单节点。集群节点我会写备注

环境配置

[集群的node节点，这一步也需要执行]

1.  
   # yum源改为阿里云
2.  
   curl -o /etc/yum.repos.d/CentOS-Base.repo https://mirrors.aliyun.com/repo/Centos-7.repo
3.  
    
4.  
   # vim配置
5.  
   echo -e 'set paste
6.  
   set expandtab
7.  
   set ts=4' >> ~/.vimrc
8.  
    
9.  
   # 一些工具
10.  
    yum install net-tools vim telnet lsof -y
11.  
     
12.  
    # k8s用阿里云源，这样速度快一些
13.  
    echo '#k8s
14.  
    [kubernetes]
15.  
    name=Kubernetes
16.  
    baseurl=https://mirrors.aliyun.com/kubernetes/yum/repos/kubernetes-el7-x86_64
17.  
    enabled=1
18.  
    gpgcheck=0
19.  
    '>/etc/yum.repos.d/kubernetes.repo
20.  
     
21.  
    #
22.  
    setenforce 0
23.  
    sed -i 's/SELINUX=enforc.*/SELINUX=disabled/g' /etc/selinux/config
24.  
    cat <<EOF > /etc/sysctl.d/k8s.conf
25.  
    net.bridge.bridge-nf-call-ip6tables = 1
26.  
    net.bridge.bridge-nf-call-iptables = 1
27.  
    EOF
28.  
    sysctl --system
29.  
     
30.  
    # 关闭swap
31.  
    swapoff -a
32.  
    sed -ri 's/.*swap.*/#&/' /etc/fstab
33.  
    free -m |grep Swap
34.  
    # 关闭防火墙
35.  
    systemctl stop firewalld && systemctl disable firewalld

部署 docker [集群的node节点，这一步也需要执行]

1.  
   yum install -y yum-utils device-mapper-persistent-data lvm2
2.  
   yum-config-manager --add-repo https://mirrors.aliyun.com/docker-ce/linux/centos/docker-ce.repo
3.  
   yum install -y docker-ce docker-ce-cli containerd.io
4.  
   mkdir /etc/docker
5.  
   cat > /etc/docker/daemon.json <<EOF
6.  
   {
7.  
   "registry-mirrors": ["https://6ijb8ubo.mirror.aliyuncs.com"],
8.  
   "exec-opts": ["native.cgroupdriver=systemd"],
9.  
   "log-driver": "json-file",
10.  
    "log-opts": {
11.  
    "max-size": "100m"
12.  
    }
13.  
    }
14.  
    EOF
15.  
     
16.  
    systemctl daemon-reload
17.  
    systemctl restart docker.service
18.  
    systemctl enable docker.service
19.  
    docker info | grep "Cgroup Driver"

部署kubeadm

[集群的node节点，这一步也需要执行]

1.  
   yum install -y kubelet-1.18.2 kubeadm-1.18.2 kubectl-1.18.2
2.  
   systemctl enable kubelet && systemctl start kubelet
3.  
    
4.  
   #查看需要依赖的镜像版本
5.  
   kubeadm config images list

kubeadm 部署的时候默认从k8s.gcr.io拉取镜像，对于国内用户来说要么速度慢，要么无法下载，换成阿里云的镜像，如何换？笔者猜kubeadm理论上是调用了"docker pull"命令，那么就跟自己手动没什么区别。

1.  
   # 通过这个命令可以获取到需要拉取的镜像名称
2.  
   kubeadm config images list | awk -F'/' '/k8s.gcr.io/{print $2}'

1.  
   cat > ~/pull_image.sh <<EOF
2.  
   #!/bin/bash
3.  
   for imageName in \$(kubeadm config images list | awk -F'/' '/k8s.gcr.io/{print \$2}') ; do
4.  
   docker pull registry.cn-hangzhou.aliyuncs.com/谷歌_containers/\${imageName}
5.  
   docker tag registry.cn-hangzhou.aliyuncs.com/谷歌_containers/\${imageName} k8s.gcr.io/\${imageName}
6.  
   docker rmi registry.cn-hangzhou.aliyuncs.com/谷歌_containers/\${imageName}
7.  
   done
8.  
   EOF

看下已经拉取下来的镜像，k8s.gcr.io开头的都是刚刚拉取的镜像文件

kubeadm初始化 master

1.  
   # kubeadm config print init-defaults > /opt/kubeadm-config.yaml
2.  
   kubeadm init --pod-network-cidr=10.244.0.0/16 --kubernetes-version=v1.18.2 2>&1 | tee kubeadm-init.log

这执行真一步的时候我遇到了报错,

1.  
   [init] Using Kubernetes version: v1.18.2
2.  
   [preflight] Running pre-flight checks
3.  
   [WARNING Firewalld]: firewalld is active, please ensure ports [6443 10250] are open or your cluster may not function correctly
4.  
   [WARNING SystemVerification]: this Docker version is not on the list of validated versions: 23.0.1. Latest validated version: 19.03
5.  
   [preflight] Pulling images required for setting up a Kubernetes cluster
6.  
   [preflight] This might take a minute or two, depending on the speed of your internet connection
7.  
   [preflight] You can also perform this action in beforehand using 'kubeadm config images pull'
8.  
   error execution phase preflight: [preflight] Some fatal errors occurred:
9.  
   [ERROR ImagePull]: failed to pull image k8s.gcr.io/kube-apiserver:v1.18.2: output: Error response from daemon: Get "https://k8s.gcr.io/v2/": dial tcp 142.250.157.82:443: i/o timeout
10.  
    , error: exit status 1
11.  
    [ERROR ImagePull]: failed to pull image k8s.gcr.io/kube-controller-manager:v1.18.2: output: Error response from daemon: Get "https://k8s.gcr.io/v2/": net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)
12.  
    , error: exit status 1
13.  
    [ERROR ImagePull]: failed to pull image k8s.gcr.io/kube-scheduler:v1.18.2: output: Error response from daemon: Get "https://k8s.gcr.io/v2/": net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers)
14.  
    , error: exit status 1
15.  
    [ERROR ImagePull]: failed to pull image k8s.gcr.io/kube-proxy:v1.18.2: output: Error response from daemon: Get "https://k8s.gcr.io/v2/": dial tcp 142.250.157.82:443: i/o timeout
16.  
    , error: exit status 1
17.  
    [preflight] If you know what you are doing, you can make a check non-fatal with `--ignore-preflight-errors=...`
18.  
    To see the stack trace of this error execute with --v=5 or higher

如果你到这里也报错了，请看上面截图的输出，kube.*组件版本号是不是也是v1.18.20？如果是的话，我感觉这里可能是正则匹配的问题, kubeadm出错了,没关系，手动tag一下把镜像改为1.18.2即可，就可以走通了。

接着往下：

1.  
   mkdir -p $HOME/.kube
2.  
   sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
3.  
   sudo chown $(id -u):$(id -g) $HOME/.kube/config
4.  
   kubectl get pod -A # 即可看到pod启动了

查看组件状态

kubectl get componentstatuses

部署网络组件

为什么有pod状态为pending？这个是因为网络组件没有安装。

1.  
   wget https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml
2.  
   # 下载之后本地执行命令, 虚拟机无法下载的话，翻墙浏览器打开URL
3.  
   kubectl apply -f kube-flannel.yml

不能下载的话，可以复制下面的内容

1.  
   ---
2.  
   kind: Namespace
3.  
   apiVersion: v1
4.  
   metadata:
5.  
   name: kube-flannel
6.  
   labels:
7.  
   pod-security.kubernetes.io/enforce: privileged
8.  
   ---
9.  
   kind: ClusterRole
10.  
    apiVersion: rbac.authorization.k8s.io/v1
11.  
    metadata:
12.  
    name: flannel
13.  
    rules:
14.  
    - apiGroups:
15.  
    - ""
16.  
    resources:
17.  
    - pods
18.  
    verbs:
19.  
    - get
20.  
    - apiGroups:
21.  
    - ""
22.  
    resources:
23.  
    - nodes
24.  
    verbs:
25.  
    - get
26.  
    - list
27.  
    - watch
28.  
    - apiGroups:
29.  
    - ""
30.  
    resources:
31.  
    - nodes/status
32.  
    verbs:
33.  
    - patch
34.  
    - apiGroups:
35.  
    - "networking.k8s.io"
36.  
    resources:
37.  
    - clustercidrs
38.  
    verbs:
39.  
    - list
40.  
    - watch
41.  
    ---
42.  
    kind: ClusterRoleBinding
43.  
    apiVersion: rbac.authorization.k8s.io/v1
44.  
    metadata:
45.  
    name: flannel
46.  
    roleRef:
47.  
    apiGroup: rbac.authorization.k8s.io
48.  
    kind: ClusterRole
49.  
    name: flannel
50.  
    subjects:
51.  
    - kind: ServiceAccount
52.  
    name: flannel
53.  
    namespace: kube-flannel
54.  
    ---
55.  
    apiVersion: v1
56.  
    kind: ServiceAccount
57.  
    metadata:
58.  
    name: flannel
59.  
    namespace: kube-flannel
60.  
    ---
61.  
    kind: ConfigMap
62.  
    apiVersion: v1
63.  
    metadata:
64.  
    name: kube-flannel-cfg
65.  
    namespace: kube-flannel
66.  
    labels:
67.  
    tier: node
68.  
    app: flannel
69.  
    data:
70.  
    cni-conf.json: |
71.  
    {
72.  
    "name": "cbr0",
73.  
    "cniVersion": "0.3.1",
74.  
    "plugins": [
75.  
    {
76.  
    "type": "flannel",
77.  
    "delegate": {
78.  
    "hairpinMode": true,
79.  
    "isDefaultGateway": true
80.  
    }
81.  
    },
82.  
    {
83.  
    "type": "portmap",
84.  
    "capabilities": {
85.  
    "portMappings": true
86.  
    }
87.  
    }
88.  
    ]
89.  
    }
90.  
    net-conf.json: |
91.  
    {
92.  
    "Network": "10.244.0.0/16",
93.  
    "Backend": {
94.  
    "Type": "vxlan"
95.  
    }
96.  
    }
97.  
    ---
98.  
    apiVersion: apps/v1
99.  
    kind: DaemonSet
100.  
     metadata:
101.  
     name: kube-flannel-ds
102.  
     namespace: kube-flannel
103.  
     labels:
104.  
     tier: node
105.  
     app: flannel
106.  
     spec:
107.  
     selector:
108.  
     matchLabels:
109.  
     app: flannel
110.  
     template:
111.  
     metadata:
112.  
     labels:
113.  
     tier: node
114.  
     app: flannel
115.  
     spec:
116.  
     affinity:
117.  
     nodeAffinity:
118.  
     requiredDuringSchedulingIgnoredDuringExecution:
119.  
     nodeSelectorTerms:
120.  
     - matchExpressions:
121.  
     - key: kubernetes.io/os
122.  
     operator: In
123.  
     values:
124.  
     - linux
125.  
     hostNetwork: true
126.  
     priorityClassName: system-node-critical
127.  
     tolerations:
128.  
     - operator: Exists
129.  
     effect: NoSchedule
130.  
     serviceAccountName: flannel
131.  
     initContainers:
132.  
     - name: install-cni-plugin
133.  
     image: docker.io/flannel/flannel-cni-plugin:v1.1.2
134.  
     #image: docker.io/rancher/mirrored-flannelcni-flannel-cni-plugin:v1.1.2
135.  
     command:
136.  
     - cp
137.  
     args:
138.  
     - -f
139.  
     - /flannel
140.  
     - /opt/cni/bin/flannel
141.  
     volumeMounts:
142.  
     - name: cni-plugin
143.  
     mountPath: /opt/cni/bin
144.  
     - name: install-cni
145.  
     image: docker.io/flannel/flannel:v0.21.2
146.  
     #image: docker.io/rancher/mirrored-flannelcni-flannel:v0.21.2
147.  
     command:
148.  
     - cp
149.  
     args:
150.  
     - -f
151.  
     - /etc/kube-flannel/cni-conf.json
152.  
     - /etc/cni/net.d/10-flannel.conflist
153.  
     volumeMounts:
154.  
     - name: cni
155.  
     mountPath: /etc/cni/net.d
156.  
     - name: flannel-cfg
157.  
     mountPath: /etc/kube-flannel/
158.  
     containers:
159.  
     - name: kube-flannel
160.  
     image: docker.io/flannel/flannel:v0.21.2
161.  
     #image: docker.io/rancher/mirrored-flannelcni-flannel:v0.21.2
162.  
     command:
163.  
     - /opt/bin/flanneld
164.  
     args:
165.  
     - --ip-masq
166.  
     - --kube-subnet-mgr
167.  
     resources:
168.  
     requests:
169.  
     cpu: "100m"
170.  
     memory: "50Mi"
171.  
     securityContext:
172.  
     privileged: false
173.  
     capabilities:
174.  
     add: ["NET_ADMIN", "NET_RAW"]
175.  
     env:
176.  
     - name: POD_NAME
177.  
     valueFrom:
178.  
     fieldRef:
179.  
     fieldPath: metadata.name
180.  
     - name: POD_NAMESPACE
181.  
     valueFrom:
182.  
     fieldRef:
183.  
     fieldPath: metadata.namespace
184.  
     - name: EVENT_QUEUE_DEPTH
185.  
     value: "5000"
186.  
     volumeMounts:
187.  
     - name: run
188.  
     mountPath: /run/flannel
189.  
     - name: flannel-cfg
190.  
     mountPath: /etc/kube-flannel/
191.  
     - name: xtables-lock
192.  
     mountPath: /run/xtables.lock
193.  
     volumes:
194.  
     - name: run
195.  
     hostPath:
196.  
     path: /run/flannel
197.  
     - name: cni-plugin
198.  
     hostPath:
199.  
     path: /opt/cni/bin
200.  
     - name: cni
201.  
     hostPath:
202.  
     path: /etc/cni/net.d
203.  
     - name: flannel-cfg
204.  
     configMap:
205.  
     name: kube-flannel-cfg
206.  
     - name: xtables-lock
207.  
     hostPath:
208.  
     path: /run/xtables.lock
209.  
     type: FileOrCreate

这个时候各节点就正常了

加入node节点

打开之前的文件 $HOME/.kube/config，放到node节点相同的位置

1.  
   # 将主节点的内容拷贝一下，粘贴到下面的文件
2.  
   vi ~/.kube/config
3.  
    
4.  
   # 去主节点查看 kubeadm-init.log 内容，重点是最后的一行
5.  
   kubeadm join 172.16.11.23:6443 --token hwrdb3.qnjvvmgi3mose4ts --discovery-token-ca-cert-hash sha256:a26c63dab6c5936041c5aade8ef7e7d99f91ee4723b476e1fbda39153ae6a58b

FAQ

1. pod coredns 状态异常

答：检查一下firewalld状态 "systemctl status firewalld.service"，看看是不是之前的步骤中，忘记了执行关闭防火墙这一步。

2. node节点显示NotReady，并且describe之后显示：Node node06 status is now: NodeHasSufficientPID

答：检查一下，node节点要执行的步骤是否有遗漏操作，导致node节点没有docker镜像。你可以通过命令"docker images"来检查。

欢迎小伙伴就部署中碰到的问题与我交流

你可能会想看这些文章

k8s集群恢复与重启