使用 kubeadm 创建高可用 Kubernetes 及外部 etcd 集群

博客地址：使用 kubeadm 创建高可用 Kubernetes 及外部 etcd 集群

前言

Kubernetes 的官方中文文档内容全面，表达清晰，有大量示例和解析

无论任何情况下都推荐先花几个小时通读官方文档，来了解配置过程中的可选项，以及可能会遇到哪些问题

本文基于官方文档中 入门 - 生产环境 一章来整理部署流程

Kubernetes 文档 | Kubernetes

架构

• OS: Debian 12
• CGroup Driver: systemd
• Container Runtime: containerd
• CNI: Calico
• Kubernetes: v1.32.0

注意
所有节点服务器都需要关闭 swap

• Other
  • 说明
    • 该服务器运行 K8S 外部应用，包括 Nginx、Nexus 等
    • 该服务器运行的所有业务通过 docker-compose 管理
    • 与 K8S 自身配置相关的步骤说明中的“所有节点”不包括该服务器
  • Server
    • vCPU: 2
    • Memory: 4G
  • Network: 192.168.1.100 2E:7E:86:3A:A5:20
  • Port:
    • 8443/tcp: 向集群提供 Kubernetes APIServer 负载均衡
• Etcd
  • Server
    • vCPU: 1
    • Memory: 1G
  • Network
    • Etcd-01: 192.168.1.101 2E:7E:86:3A:A5:21
    • Etcd-02: 192.168.1.102 2E:7E:86:3A:A5:22
    • Etcd-03: 192.168.1.103 2E:7E:86:3A:A5:23
  • Port:
    • 2379/tcp: etcd HTTP API
    • 2380/tcp: etcd peer 通信
• Master
  • Server
    • vCPU: 4
    • Memory: 8G
  • Network
    • Master-01: 192.168.1.104 2E:7E:86:3A:A5:24
    • Master-02: 192.168.1.105 2E:7E:86:3A:A5:25
    • Master-03: 192.168.1.106 2E:7E:86:3A:A5:26
  • Port:
    • 179/tcp: Calico BGP
    • 4789/udp: VXLAN
    • 5473/tcp: Calico 管理界面
    • 6443/tcp: Kubernetes APIServer
    • 6666/tcp: Calico Felix
    • 10250/tcp: kubelet API
• Node
  • Server
    • vCPU: 4
    • Memory: 8G
  • Network
    • Node-01: 192.168.1.107 2E:7E:86:3A:A5:27
    • Node-02: 192.168.1.108 2E:7E:86:3A:A5:28
    • Node-03: 192.168.1.109 2E:7E:86:3A:A5:29
  • Port:
    • 179/tcp: Calico BGP
    • 10250/tcp: kubelet API

配置基础环境

说明
所有节点

apt update
apt upgrade
apt install curl apt-transport-https ca-certificates gnupg2 software-properties-common vimcurl -fsSL https://mirrors.ustc.edu.cn/kubernetes/core:/stable:/v1.32/deb/Release.key | gpg --dearmor -o /etc/apt/keyrings/kubernetes-apt-keyring.gpg
chmod 644 /etc/apt/keyrings/kubernetes-apt-keyring.gpg
echo "deb [signed-by=/etc/apt/keyrings/kubernetes-apt-keyring.gpg] https://mirrors.ustc.edu.cn/kubernetes/core:/stable:/v1.32/deb/ /" | tee /etc/apt/sources.list.d/kubernetes.listcurl -fsSL https://mirrors.ustc.edu.cn/docker-ce/linux/debian/gpg -o /etc/apt/keyrings/docker.asc
chmod a+r /etc/apt/keyrings/docker.asc
echo "deb [signed-by=/etc/apt/keyrings/docker.asc] https://mirrors.ustc.edu.cn/docker-ce/linux/debian bookworm stable" | tee /etc/apt/sources.list.d/docker.listapt updateapt install containerd.iomkdir -p /etc/containerd
containerd config default | tee /etc/containerd/config.tomlsystemctl restart containerdapt install kubelet kubeadm kubectl
apt-mark hold kubelet kubeadm kubectl

开启 ipv4 转发

编辑 /etc/sysctl.conf，找到下方配置并取消注释

net.ipv4.ip_forward=1

执行 sysctl -p 应用配置

创建 crictl 配置

cat << EOF > /etc/crictl.yaml
runtime-endpoint: unix:///run/containerd/containerd.sock
image-endpoint: unix:///run/containerd/containerd.sock
timeout: 10
debug: false
EOF

如果需要通过代理服务器访问容器仓库，需要为 containerd 配置代理服务

mkdir -p /etc/systemd/system/containerd.service.d
cat << EOF > /etc/systemd/system/containerd.service.d/http-proxy.conf
[Service]
Environment="HTTP_PROXY=http://username:password@proxy-server-ip:port"
Environment="HTTPS_PROXY=http://username:password@proxy-server-ip:port"
Environment="NO_PROXY=localhost,127.0.0.0/8,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16"
EOFsystemctl daemon-reload
systemctl restart containerd.service

已知问题

使用 systemd 作为 CGroup Driver 且使用 containerd 作为 CRI 运行时

需要修改 /etc/containerd/config.toml，添加如下配置

相关文章：配置 systemd cgroup 驱动 | Kubernetes

[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]...[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]SystemdCgroup = true

执行 systemctl restart containerd

或参照另一篇文章的解决方案

相关文章：Why does etcd fail with Debian/bullseye kernel? - General Discussions - Discuss Kubernetes

cat /etc/default/grub# Source:
# GRUB_CMDLINE_LINUX_DEFAULT="quiet"# Modify:
GRUB_CMDLINE_LINUX_DEFAULT="quiet systemd.unified_cgroup_hierarchy=0"

执行 update-grub 并重启

配置 etcd 节点

将 kubelet 配置为 etcd 的服务管理器

说明
所有 etcd 节点

apt update
apt install etcd-clientmkdir -p /etc/systemd/system/kubelet.service.dcat << EOF > /etc/systemd/system/kubelet.service.d/kubelet.conf
apiVersion: kubelet.config.k8s.io/v1beta1
kind: KubeletConfiguration
authentication:anonymous:enabled: falsewebhook:enabled: false
authorization:mode: AlwaysAllow
cgroupDriver: systemd
address: 127.0.0.1
containerRuntimeEndpoint: unix:///var/run/containerd/containerd.sock
staticPodPath: /etc/kubernetes/manifests
EOFcat << EOF > /etc/systemd/system/kubelet.service.d/20-etcd-service-manager.conf
[Service]
Environment="KUBELET_CONFIG_ARGS=--config=/etc/systemd/system/kubelet.service.d/kubelet.conf"
ExecStart=
ExecStart=/usr/bin/kubelet $KUBELET_CONFIG_ARGS
Restart=always
EOFsystemctl daemon-reload
systemctl restart kubelet

为 kubeadm 创建配置文件

说明
Etcd-01 节点，由该节点向其他节点分发证书及配置
该节点同时作为 CA

生成 CA

kubeadm init phase certs etcd-ca

生成如下文件

• /etc/kubernetes/pki/etcd/ca.crt
• /etc/kubernetes/pki/etcd/ca.key

为方便接下来的步骤操作，先将 etcd 节点信息导出为环境变量

export HOST0=192.168.1.101
export HOST1=192.168.1.102
export HOST2=192.168.1.103export NAME0="etcd-01"
export NAME1="etcd-02"
export NAME2="etcd-03"

为 etcd 成员生成 kubeadm 配置

HOSTS=(${HOST0} ${HOST1} ${HOST2})
NAMES=(${NAME0} ${NAME1} ${NAME2})for i in "${!HOSTS[@]}"; doHOST=${HOSTS[$i]}
NAME=${NAMES[$i]}mkdir -p /tmp/${HOST}cat << EOF > /tmp/${HOST}/kubeadmcfg.yaml
---
apiVersion: "kubeadm.k8s.io/v1beta4"
kind: InitConfiguration
nodeRegistration:name: ${NAME}
localAPIEndpoint:advertiseAddress: ${HOST}
---
apiVersion: "kubeadm.k8s.io/v1beta4"
kind: ClusterConfiguration
etcd:local:serverCertSANs:- "${HOST}"peerCertSANs:- "${HOST}"extraArgs:- name: initial-clustervalue: ${NAMES[0]}=https://${HOSTS[0]}:2380,${NAMES[1]}=https://${HOSTS[1]}:2380,${NAMES[2]}=https://${HOSTS[2]}:2380- name: initial-cluster-statevalue: new- name: namevalue: ${NAME}- name: listen-peer-urlsvalue: https://${HOST}:2380- name: listen-client-urlsvalue: https://${HOST}:2379- name: advertise-client-urlsvalue: https://${HOST}:2379- name: initial-advertise-peer-urlsvalue: https://${HOST}:2380
EOF
done

为每个成员创建证书

kubeadm init phase certs etcd-server --config=/tmp/${HOST2}/kubeadmcfg.yaml
kubeadm init phase certs etcd-peer --config=/tmp/${HOST2}/kubeadmcfg.yaml
kubeadm init phase certs etcd-healthcheck-client --config=/tmp/${HOST2}/kubeadmcfg.yaml
kubeadm init phase certs apiserver-etcd-client --config=/tmp/${HOST2}/kubeadmcfg.yaml
cp -R /etc/kubernetes/pki /tmp/${HOST2}/
# Clear useless cert
find /etc/kubernetes/pki -not -name ca.crt -not -name ca.key -type f -deletekubeadm init phase certs etcd-server --config=/tmp/${HOST1}/kubeadmcfg.yaml
kubeadm init phase certs etcd-peer --config=/tmp/${HOST1}/kubeadmcfg.yaml
kubeadm init phase certs etcd-healthcheck-client --config=/tmp/${HOST1}/kubeadmcfg.yaml
kubeadm init phase certs apiserver-etcd-client --config=/tmp/${HOST1}/kubeadmcfg.yaml
cp -R /etc/kubernetes/pki /tmp/${HOST1}/
find /etc/kubernetes/pki -not -name ca.crt -not -name ca.key -type f -deletekubeadm init phase certs etcd-server --config=/tmp/${HOST0}/kubeadmcfg.yaml
kubeadm init phase certs etcd-peer --config=/tmp/${HOST0}/kubeadmcfg.yaml
kubeadm init phase certs etcd-healthcheck-client --config=/tmp/${HOST0}/kubeadmcfg.yaml
kubeadm init phase certs apiserver-etcd-client --config=/tmp/${HOST0}/kubeadmcfg.yaml# Clear ca key from member
find /tmp/${HOST2} -name ca.key -type f -delete
find /tmp/${HOST1} -name ca.key -type f -delete

将证书移动到对应的成员服务器

scp -r /tmp/${HOST2}/pki root@${HOST2}:/etc/kubernetes/
scp /tmp/${HOST2}/kubeadmcfg.yaml root@${HOST2}:~/scp -r /tmp/${HOST1}/pki root@${HOST1}:/etc/kubernetes/
scp /tmp/${HOST1}/kubeadmcfg.yaml root@${HOST1}:~/mv /tmp/${HOST0}/kubeadmcfg.yaml ~/rm -rf /tmp/${HOST2}
rm -rf /tmp/${HOST1}
rm -rf /tmp/${HOST0}

此时在三台 etcd 节点中的文件结构均应如下

/root
└── kubeadmcfg.yaml
---
/etc/kubernetes/pki
├── apiserver-etcd-client.crt
├── apiserver-etcd-client.key
└── etcd├── ca.crt├── ca.key # 仅 CA 节点既 etcd-01├── healthcheck-client.crt├── healthcheck-client.key├── peer.crt├── peer.key├── server.crt└── server.key

创建静态 Pod 清单

说明
所有 etcd 节点

kubeadm init phase etcd local --config=/root/kubeadmcfg.yaml

检查集群运行情况

将 ${HOST0} 替换为想要检查的节点 ip

ETCDCTL_API=3 etcdctl \
--cert /etc/kubernetes/pki/etcd/peer.crt \
--key /etc/kubernetes/pki/etcd/peer.key \
--cacert /etc/kubernetes/pki/etcd/ca.crt \
--endpoints https://${HOST0}:2379 endpoint health

使用 kubeadm 创建高可用集群

说明

配置过程中需要完全重置控制平面节点的配置时，需要有至少一台能够访问集群的节点，在该节点上按如下流程操作

kubectl delete pods,nodes,namespaces,deployments,services --all --all-namespaces --force
kubectl delete -f tigera-operator.yaml --force
kubectl delete -f custom-resources.yaml --force
kubeadm reset --cleanup-tmp-dir -f
rm -rf /etc/cni/net.d/*
rm -rf ~/.kube
systemctl restart kubelet containerd

为 kube-apiserver 创建负载均衡

说明
本文中负载均衡使用 Nginx

Nginx 配置

http {...
}stream {upstream apiserver {server 192.168.1.104:6443 weight=5 max_fails=3 fail_timeout=30s; # Master-01server 192.168.1.105:6443 weight=5 max_fails=3 fail_timeout=30s; # Master-02server 192.168.1.106:6443 weight=5 max_fails=3 fail_timeout=30s; # Master-03}server {listen 8443;proxy_pass apiserver;}
}

为控制平面节点配置外部 etcd 节点

说明
任一 etcd 节点与主控制平面节点，本文中为 Etcd-01 与 Master-01

从集群中任一 etcd 节点复制到主控制平面节点

scp /etc/kubernetes/pki/etcd/ca.crt /etc/kubernetes/pki/apiserver-etcd-client.crt /etc/kubernetes/pki/apiserver-etcd-client.key root@192.168.1.104:~

在主控制平面节点中将文件移动到指定位置

mkdir -p /etc/kubernetes/pki/etcd
mv ~/ca.crt /etc/kubernetes/pki/etcd/
mv ~/apiserver-etcd-client.crt /etc/kubernetes/pki/
mv ~/apiserver-etcd-client.key /etc/kubernetes/pki/

创建 kubeadm-config.yaml，内容如下

• controlPlaneEndpoint: 负载均衡服务器
• etcd
  • external
    • endpoints: etcd 节点列表
• networking
  • podSubnet: pod ip cidr

---
apiVersion: kubeadm.k8s.io/v1beta4
kind: ClusterConfiguration
kubernetesVersion: v1.32.0
controlPlaneEndpoint: 192.168.1.100:8443
etcd:external:endpoints:- https://192.168.1.101:2379- https://192.168.1.102:2379- https://192.168.1.103:2379caFile: /etc/kubernetes/pki/etcd/ca.crtcertFile: /etc/kubernetes/pki/apiserver-etcd-client.crtkeyFile: /etc/kubernetes/pki/apiserver-etcd-client.key
networking:dnsDomain: cluster.localpodSubnet: 10.244.0.0/24serviceSubnet: 10.96.0.0/16

初始化主控制平面

说明
主控制平面节点，本文中为 Master-01

• --upload-certs: 将控制平面间的共享证书上传到 kubeadm-certs Secret
  • kubeadm-certs Secret 和解密密钥将在两小时后失效
  • 如果要重新上传证书并生成新的解密密钥，需要在已加入集群的控制平面节点上执行 kubeadm init phase upload-certs --upload-certs

kubeadm init --config kubeadm-config.yaml --upload-certs

等待运行完成后应输出类似如下内容

...Your Kubernetes control-plane has initialized successfully!To start using your cluster, you need to run the following as a regular user:mkdir -p $HOME/.kubesudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/configsudo chown $(id -u):$(id -g) $HOME/.kube/configAlternatively, if you are the root user, you can run:export KUBECONFIG=/etc/kubernetes/admin.confYou should now deploy a pod network to the cluster.
Run "kubectl apply -f [podnetwork].yaml" with one of the options listed at:https://kubernetes.io/docs/concepts/cluster-administration/addons/You can now join any number of control-plane nodes running the following command on each as root:kubeadm join 192.168.1.100:8443 --token 7r34LU.iLiRgu2qHdAeeanS --discovery-token-ca-cert-hash sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08 --control-plane --certificate-key 03d66dd08835c1ca3f128cceacd1f31ac94163096b20f445ae84285bc0832d72Please note that the certificate-key gives access to cluster sensitive data, keep it secret!
As a safeguard, uploaded-certs will be deleted in two hours; If necessary, you can use
"kubeadm init phase upload-certs --upload-certs" to reload certs afterward.Then you can join any number of worker nodes by running the following on each as root:kubeadm join 192.168.1.100:8443 --token 7r34LU.iLiRgu2qHdAeeanS --discovery-token-ca-cert-hash sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08

先将控制台输出的以上内容保存，稍后将使用这些命令来将其他控制平面节点和工作节点加入集群

根据输出的提示，复制 kubeconfig 用于 kubectl

mkdir -p ~/.kube
cp /etc/kubernetes/admin.conf ~/.kube/config

应用 CNI 插件

由于该清单过大，kubectl apply 会产生如下报错，使用 kubectl create 或 kubectl replace

注意
确认 custom-resources.yaml 中 calicoNetwork 配置的 ip cidr 与集群 podSubnet 配置一致

# kubectl apply -f https://raw.githubusercontent.com/projectcalico/calico/refs/heads/release-v3.29/manifests/tigera-operator.yaml
# The CustomResourceDefinition "installations.operator.tigera.io" is invalid: metadata.annotations: Too long: may not be more than 262144 bytes
wget https://raw.githubusercontent.com/projectcalico/calico/refs/heads/release-v3.29/manifests/tigera-operator.yaml
wget https://raw.githubusercontent.com/projectcalico/calico/refs/heads/release-v3.29/manifests/custom-resources.yaml
kubectl create -f tigera-operator.yaml
kubectl create -f custom-resources.yaml

输入以下内容查看控制平面组件 pod 启动状态

kubectl get pod -A

初始化其他控制平面

说明
除主控制平面节点外的其他控制平面节点，本文中为 Master-02 Master-03
使用 kubeadm join 命令加入集群的节点会将 KubeConfig 同步到 /etc/kubernetes/admin.conf

依照上面输出的命令，分别在其他控制平面节点中执行

• --control-plane: 通知 kubeadm join 创建一个新控制平面
• --certificate-key xxx: 从集群 kubeadm-certs Secret 下载控制平面证书并使用给定的密钥解密

kubeadm join 192.168.1.100:8443 --token 7r34LU.iLiRgu2qHdAeeanS --discovery-token-ca-cert-hash sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08 --control-plane --certificate-key 03d66dd08835c1ca3f128cceacd1f31ac94163096b20f445ae84285bc0832d72

根据输出的提示，复制 kubeconfig 用于 kubectl

mkdir -p ~/.kube
cp /etc/kubernetes/admin.conf ~/.kube/config

初始化负载节点

说明
所有负载节点
使用 kubeadm join 命令加入集群的节点会将 KubeConfig 同步到 /etc/kubernetes/kubelet.conf

依照上面输出的命令，分别在负载节点中执行

kubeadm join 192.168.1.100:8443 --token 7r34LU.iLiRgu2qHdAeeanS --discovery-token-ca-cert-hash sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08