二进制部署 on Liangweidong's blog

K8s 二进制部署：软件准备 / 证书 / 工具 / 组件全流程

Thu, 15 May 2025 00:00:00 +0800

写于 2020-03，背景：K8s 1.18 刚 GA，二进制部署仍是国内大厂主流（比 kubeadm 灵活、可定制）。本文以 3 master 高可用集群 + Ubuntu 22.04 为模板，覆盖从软件准备到所有组件 systemd 启动的完整流程。

一、为什么还要学二进制部署

部署方式	优点	缺点
kubeadm	一行命令搞定，官方推荐	黑盒多，问题排查难
二进制	完全可控，组件协作透明	步骤多，容易踩坑
k8s-operator	适合自建 PaaS	上手成本高

结论：生产用 kubeadm 部署，但学习用二进制才能真正理解 K8s。

二、集群规划

3 master + N worker 高可用集群：

主机名	IP	角色
master1	`192.168.139.133`	master + worker（混合）
master2	`192.168.139.134`	master + worker
master3	`192.168.139.135`	master + worker
VIP	`192.168.139.150`	Keepalived 漂移的虚拟 IP
worker1~N	自定	纯 worker

网络规划：

名称	网段
Node 网络	`192.168.139.0/24`
Service 网络	`10.96.0.0/12`
Pod 网络	`172.218.0.0/16`

三、软件准备

所有节点创建统一目录：

1

mkdir -p /k8s/softs

3.1 下载 K8s Server 包

1
2


# 1.28.5 完整 server 包（含所有 master/worker 组件二进制）
wget https://dl.k8s.io/v1.28.5/kubernetes-server-linux-amd64.tar.gz

3.2 下载 etcd

1

wget https://github.com/etcd-io/etcd/releases/download/v3.5.11/etcd-v3.5.11-linux-amd64.tar.gz

3.3 下载 cfssl（证书签发工具）

1
2
3
4
5
6
7
8


wget https://github.com/cloudflare/cfssl/releases/download/v1.6.4/cfssl_1.6.4_linux_amd64
wget https://github.com/cloudflare/cfssl/releases/download/v1.6.4/cfssljson_1.6.4_linux_amd64
wget https://github.com/cloudflare/cfssl/releases/download/v1.6.4/cfssl-certinfo_1.6.4_linux_amd64

cp cfssl_1.6.4_linux_amd64 /usr/local/bin/cfssl
cp cfssljson_1.6.4_linux_amd64 /usr/local/bin/cfssljson
cp cfssl-certinfo_1.6.4_linux_amd64 /usr/local/bin/cfssl-certinfo
chmod +x /usr/local/bin/cfssl /usr/local/bin/cfssljson /usr/local/bin/cfssl-certinfo

四、系统配置（所有节点）

4.1 禁用 swap

1
2
3


free -m
swapoff -a
sed -ri 's/.*swap.*/#&/' /etc/fstab

4.2 禁用防火墙

1
2
3
4
5
6


# Ubuntu
sudo systemctl stop ufw
sudo systemctl disable ufw
# CentOS 7
systemctl stop firewalld
systemctl disable firewalld

4.3 关闭 SELinux（CentOS）

1
2


setenforce 0
sed -i 's#SELINUX=enforcing#SELINUX=disabled#g' /etc/selinux/config

4.4 时间同步

1
2
3


timedatectl set-timezone Asia/Shanghai
apt install -y ntpdate
echo "0 */1 * * * ntpdate time1.aliyun.com" >> /var/spool/cron/crontabs/root

4.5 加载 IPVS 模块

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17


apt install ipvsadm ipset sysstat conntrack -y
cat >> /etc/modules-load.d/ipvs.conf <<EOF
ip_vs
ip_vs_rr
ip_vs_wwr
ip_vs_sh
nf_conntrack
ip_tables
ip_set
xt_set
ipt_set
ipt_rpfilter
ipt_REJECT
ipip
EOF
systemctl restart systemd-modules-load.service
lsmod | grep -e ip_vs -e nf_conntrack

4.6 内核参数

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


cat << EOF > /etc/sysctl.d/k8s.conf
net.ipv4.ip_forward = 1
net.bridge.bridge-nf-call-iptables = 1
net.bridge.bridge-nf-call-ip6tables = 1
fs.may_detach_mounts = 1
vm.overcommit_memory = 1
vm.panic_on_oom = 0
fs.inotify.max_user_watches = 89100
fs.file-max = 52706963
fs.nr_open = 52706963
net.netfilter.nf_conntrack_max = 2310720
net.ipv4.tcp_keepalive_time = 600
net.ipv4.tcp_keepalive_probes = 3
net.ipv4.tcp_keepalive_intvl = 15
net.ipv4.tcp_max_tw_buckets = 36000
net.ipv4.tcp_tw_reuse = 1
net.core.somaxconn = 16384
EOF
sysctl --system

4.7 ulimit 调优

1
2
3
4
5
6
7
8
9


ulimit -SHn 65535
cat >> /etc/security/limits.conf <<EOF
* soft nofile 655360
* hard nofile 131072
* soft nproc 655350
* hard nproc 655350
* soft memlock unlimited
* hard memlock unlimited
EOF

五、证书签发（master1 操作）

K8s 集群里几乎所有通信都走 TLS，证书是入门第一道坎。

5.1 etcd 证书

 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


mkdir -p /k8s/softs/pki && cd /k8s/softs/pki

# 写入 ca 配置
cat > ca-config.json << EOF
{
 "signing": {
 "default": { "expiry": "876000h" },
 "profiles": {
 "kubernetes": {
 "usages": ["signing", "key encipherment", "server auth", "client auth"],
 "expiry": "876000h"
 }
 }
 }
}
EOF

# etcd CA
cat > etcd-ca-csr.json << EOF
{
 "CN": "etcd",
 "key": { "algo": "rsa", "size": 2048 },
 "names": [{ "C": "CN", "ST": "Beijing", "L": "Beijing", "O": "etcd", "OU": "Etcd Security" }],
 "ca": { "expiry": "876000h" }
}
EOF

cfssl gencert -initca etcd-ca-csr.json | cfssljson -bare /etc/etcd/ssl/etcd-ca

# etcd server 证书（注意 hostname 和 IP）
cat > etcd-csr.json << EOF
{
 "CN": "etcd",
 "key": { "algo": "rsa", "size": 2048 },
 "names": [{ "C": "CN", "ST": "Beijing", "L": "Beijing", "O": "etcd", "OU": "Etcd Security" }]
}
EOF

cfssl gencert \
 -ca=/etc/etcd/ssl/etcd-ca.pem \
 -ca-key=/etc/etcd/ssl/etcd-ca-key.pem \
 -config=ca-config.json \
 -hostname=127.0.0.1,master1,master2,master3,192.168.139.133,192.168.139.134,192.168.139.135 \
 -profile=kubernetes \
 etcd-csr.json | cfssljson -bare /etc/etcd/ssl/etcd

5.2 K8s CA 与 apiserver 证书

 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


# K8s CA（10 年有效）
cat > ca-csr.json << EOF
{
 "CN": "kubernetes",
 "key": { "algo": "rsa", "size": 2048 },
 "names": [{ "C": "CN", "ST": "Beijing", "L": "Beijing", "O": "kubernetes", "OU": "K8s CA" }],
 "ca": { "expiry": "876000h" }
}
EOF

cfssl gencert -initca ca-csr.json | cfssljson -bare /etc/kubernetes/pki/ca

# apiserver 证书（多个 hostname/IP）
cat > apiserver-csr.json << EOF
{
 "CN": "kube-apiserver",
 "key": { "algo": "rsa", "size": 2048 },
 "names": [{ "C": "CN", "ST": "Beijing", "L": "Beijing", "O": "kubernetes", "OU": "K8s" }]
}
EOF

cfssl gencert \
 -ca=/etc/kubernetes/pki/ca.pem \
 -ca-key=/etc/kubernetes/pki/ca-key.pem \
 -config=ca-config.json \
 -hostname=10.96.0.1,192.168.139.133,192.168.139.134,192.168.139.135,192.168.139.150,127.0.0.1,kubernetes,kubernetes.default,kubernetes.default.svc,kubernetes.default.svc.cluster,kubernetes.default.svc.cluster.local \
 -profile=kubernetes \
 apiserver-csr.json | cfssljson -bare /etc/kubernetes/pki/apiserver

5.3 前置代理证书（aggregator）

1
2
3
4
5
6
7
8


# front-proxy CA + client
cfssl gencert -initca front-proxy-ca-csr.json | cfssljson -bare /etc/kubernetes/pki/front-proxy-ca
cfssl gencert \
 -ca=/etc/kubernetes/pki/front-proxy-ca.pem \
 -ca-key=/etc/kubernetes/pki/front-proxy-ca-key.pem \
 -config=ca-config.json \
 -profile=kubernetes front-proxy-client-csr.json | \
 cfssljson -bare /etc/kubernetes/pki/front-proxy-client

5.4 controller-manager / scheduler / admin / kube-proxy 证书

按角色签发，证书 CN 决定 K8s RBAC 里的身份（system:kube-controller-manager 等）：

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23


# controller-manager
cfssl gencert -ca=/etc/kubernetes/pki/ca.pem \
 -ca-key=/etc/kubernetes/pki/ca-key.pem -config=ca-config.json \
 -profile=kubernetes manager-csr.json | \
 cfssljson -bare /etc/kubernetes/pki/controller-manager

# scheduler
cfssl gencert -ca=/etc/kubernetes/pki/ca.pem \
 -ca-key=/etc/kubernetes/pki/ca-key.pem -config=ca-config.json \
 -profile=kubernetes scheduler-csr.json | \
 cfssljson -bare /etc/kubernetes/pki/scheduler

# admin
cfssl gencert -ca=/etc/kubernetes/pki/ca.pem \
 -ca-key=/etc/kubernetes/pki/ca-key.pem -config=ca-config.json \
 -profile=kubernetes admin-csr.json | \
 cfssljson -bare /etc/kubernetes/pki/admin

# kube-proxy
cfssl gencert -ca=/etc/kubernetes/pki/ca.pem \
 -ca-key=/etc/kubernetes/pki/ca-key.pem -config=ca-config.json \
 -profile=kubernetes kube-proxy-csr.json | \
 cfssljson -bare /etc/kubernetes/pki/kube-proxy

5.5 SA（ServiceAccount）密钥对

1
2


openssl genrsa -out /etc/kubernetes/pki/sa.key 2048
openssl rsa -in /etc/kubernetes/pki/sa.key -pubout -out /etc/kubernetes/pki/sa.pub

5.6 同步证书到其他 master

1
2
3
4
5
6


for NODE in master2 master3; do
 ssh $NODE "mkdir -p /etc/kubernetes/pki /etc/etcd/ssl"
 for DIR in etcd/ssl kubernetes/pki; do
 scp -r /etc/$DIR/* $NODE:/etc/$DIR/
 done
done

六、生成 kubeconfig（每个组件一个身份）

kubeconfig 包含集群地址 + 用户证书 + 上下文：

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17


KUBECONFIG_FILE=/etc/kubernetes/admin.kubeconfig
kubectl config set-cluster kubernetes \
 --certificate-authority=/etc/kubernetes/pki/ca.pem \
 --embed-certs=true --server=https://192.168.139.150:6443 \
 --kubeconfig=$KUBECONFIG_FILE

kubectl config set-credentials kubernetes-admin \
 --client-certificate=/etc/kubernetes/pki/admin.pem \
 --client-key=/etc/kubernetes/pki/admin-key.pem \
 --embed-certs=true \
 --kubeconfig=$KUBECONFIG_FILE

kubectl config set-context kubernetes-admin@kubernetes \
 --cluster=kubernetes --user=kubernetes-admin \
 --kubeconfig=$KUBECONFIG_FILE

kubectl config use-context kubernetes-admin@kubernetes --kubeconfig=$KUBECONFIG_FILE

controller-manager / scheduler / kube-proxy 各自生成一份 kubeconfig。

七、解压二进制到 /usr/local/bin

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11


# K8s 组件（master + worker 都有）
cd /k8s/softs
tar -xf kubernetes-server-linux-amd64.tar.gz \
 --strip-components=3 -C /usr/local/bin \
 kubernetes/server/bin/kube{let,ctl,-apiserver,-controller-manager,-scheduler,-proxy}

# etcd
tar -xf etcd-v3.5.11-linux-amd64.tar.gz
mv etcd-v3.5.11-linux-amd64/etcd /usr/local/bin/
mv etcd-v3.5.11-linux-amd64/etcdctl /usr/local/bin/
mv etcd-v3.5.11-linux-amd64/etcdutl /usr/local/bin/

同步到其他节点：

 1
 2
 3
 4
 5
 6
 7
 8
 9
10


# master 节点
for NODE in master2 master3; do
 scp /usr/local/bin/kube{let,ctl,-apiserver,-controller-manager,-scheduler,-proxy} $NODE:/usr/local/bin/
 scp /usr/local/bin/etcd* $NODE:/usr/local/bin/
done

# worker 节点
for NODE in worker1 worker2; do
 scp /usr/local/bin/kube{let,-proxy} $NODE:/usr/local/bin/
done

八、systemd 单元文件

8.1 kube-apiserver.service

所有 master 节点都需要，注意 --advertise-address 每个 master 不同：

 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


[Unit]
Description=Kubernetes API Server
Documentation=https://github.com/kubernetes/kubernetes
After=network.target

[Service]
ExecStart=/usr/local/bin/kube-apiserver \
 --v=2 \
 --allow-privileged=true \
 --bind-address=0.0.0.0 \
 --secure-port=6443 \
 --advertise-address=192.168.139.133 \
 --service-cluster-ip-range=10.96.0.0/12,fd00:1111::/112 \
 --service-node-port-range=30000-32767 \
 --etcd-servers=https://192.168.139.133:2379,https://192.168.139.134:2379,https://192.168.139.135:2379 \
 --etcd-cafile=/etc/etcd/ssl/etcd-ca.pem \
 --etcd-certfile=/etc/etcd/ssl/etcd.pem \
 --etcd-keyfile=/etc/etcd/ssl/etcd-key.pem \
 --client-ca-file=/etc/kubernetes/pki/ca.pem \
 --tls-cert-file=/etc/kubernetes/pki/apiserver.pem \
 --tls-private-key-file=/etc/kubernetes/pki/apiserver-key.pem \
 --kubelet-client-certificate=/etc/kubernetes/pki/apiserver.pem \
 --kubelet-client-key=/etc/kubernetes/pki/apiserver-key.pem \
 --service-account-key-file=/etc/kubernetes/pki/sa.pub \
 --service-account-signing-key-file=/etc/kubernetes/pki/sa.key \
 --service-account-issuer=https://kubernetes.default.svc.cluster.local \
 --enable-admission-plugins=NamespaceLifecycle,LimitRanger,ServiceAccount,DefaultStorageClass,DefaultTolerationSeconds,NodeRestriction,ResourceQuota \
 --authorization-mode=Node,RBAC \
 --enable-bootstrap-token-auth=true \
 --requestheader-client-ca-file=/etc/kubernetes/pki/front-proxy-ca.pem \
 --proxy-client-cert-file=/etc/kubernetes/pki/front-proxy-client.pem \
 --proxy-client-key-file=/etc/kubernetes/pki/front-proxy-client-key.pem \
 --requestheader-allowed-names=aggregator \
 --requestheader-group-headers=X-Remote-Group \
 --requestheader-extra-headers-prefix=X-Remote-Extra- \
 --requestheader-username-headers=X-Remote-User \
 --enable-aggregator-routing=true
Restart=on-failure
RestartSec=10s
LimitNOFILE=65535

[Install]
WantedBy=multi-user.target

8.2 kube-controller-manager.service

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


[Unit]
Description=Kubernetes Controller Manager
Documentation=https://github.com/kubernetes/kubernetes
After=network.target

[Service]
ExecStart=/usr/local/bin/kube-controller-manager \
 --v=2 \
 --bind-address=0.0.0.0 \
 --root-ca-file=/etc/kubernetes/pki/ca.pem \
 --cluster-signing-cert-file=/etc/kubernetes/pki/ca.pem \
 --cluster-signing-key-file=/etc/kubernetes/pki/ca-key.pem \
 --service-account-private-key-file=/etc/kubernetes/pki/sa.key \
 --kubeconfig=/etc/kubernetes/controller-manager.kubeconfig \
 --leader-elect=true \
 --use-service-account-credentials=true
Restart=on-failure
RestartSec=10s
LimitNOFILE=65535

[Install]
WantedBy=multi-user.target

8.3 kube-scheduler.service

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17


[Unit]
Description=Kubernetes Scheduler
Documentation=https://github.com/kubernetes/kubernetes
After=network.target

[Service]
ExecStart=/usr/local/bin/kube-scheduler \
 --v=2 \
 --bind-address=0.0.0.0 \
 --leader-elect=true \
 --kubeconfig=/etc/kubernetes/scheduler.kubeconfig
Restart=on-failure
RestartSec=10s
LimitNOFILE=65535

[Install]
WantedBy=multi-user.target

8.4 kubelet.service（所有 master + worker 节点）

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


[Unit]
Description=Kubernetes Kubelet
Documentation=https://github.com/kubernetes/kubernetes
After=network.target

[Service]
ExecStart=/usr/local/bin/kubelet \
 --v=2 \
 --bootstrap-kubeconfig=/etc/kubernetes/bootstrap-kubelet.kubeconfig \
 --kubeconfig=/etc/kubernetes/kubelet.kubeconfig \
 --config=/etc/kubernetes/kubelet-config.yaml \
 --cert-dir=/etc/kubernetes/pki \
 --hostname-override=<本节点主机名>
Restart=on-failure
RestartSec=10s
LimitNOFILE=65535

[Install]
WantedBy=multi-user.target

8.5 kube-proxy.service

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16


[Unit]
Description=Kubernetes Kube-Proxy Server
Documentation=https://github.com/kubernetes/kubernetes
After=network.target

[Service]
ExecStart=/usr/local/bin/kube-proxy \
 --v=2 \
 --config=/etc/kubernetes/kube-proxy-config.yaml \
 --hostname-override=<本节点主机名>
Restart=on-failure
RestartSec=10s
LimitNOFILE=65535

[Install]
WantedBy=multi-user.target

九、启动顺序与验证

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11


# 所有 master 节点
systemctl daemon-reload
systemctl enable --now kube-apiserver kube-controller-manager kube-scheduler kubelet kube-proxy
systemctl status kube-apiserver

# 验证（master1）
kubectl get componentstatuses
# NAME STATUS MESSAGE ERROR
# controller-manager Healthy ok
# scheduler Healthy ok
# etcd-0 Healthy {"health":"true"}

十、常见坑

证书 hostname 漏配：apiserver 报 “x509: certificate is not valid for any names”，重新签发并把 VIP、Service IP、k8s 域名都加进去
bootstrap token 过期：默认 24h，过期后 worker 无法 join，kubeadm token create --print-join-command 重新生成（kubeadm 部署）
etcd 启动后 apiserver 连不上：etcd 证书路径不对，配置里写 /etc/etcd/ssl/，但 cert 实际在 /etc/kubernetes/pki/etcd/，加软链
kubelet 起不来：bootstrap-kubelet.kubeconfig 不存在或 token 失效
kube-proxy 模式选错：默认 iptables，高并发场景（>5000 Service）改 IPVS 性能更好

十一、前置知识 / 下一步

前置：

K8s 架构概念（参考 2017-06-15《Kubernetes 入门》）
etcd 基本原理（参考 2020-09-15《etcd 集群部署》）
Linux 网络基础（iptables / IPVS / 网桥）

下一步：

etcd 集群部署（2020-09-15）—— K8s 数据层高可用
API Server 高可用（2021-03-15）—— Nginx 四层 + Keepalived VIP
K8s 集群插件（2021-09-15）—— CNI / CoreDNS / Metrics / Dashboard

参考资料

Kubernetes The Hard Way - Kelsey Hightower
Kubernetes 官方文档 - kubeadm 安装
cfssl 官方文档
Kubernetes 组件参数参考

K8s 生产二进制集群前置：节点规划与 Linux 系统调优

Fri, 15 Dec 2023 00:00:00 +0800

写在前面

在 Kubernetes 1.6（2017 年）之前，社区主流的"高可用生产集群"基本是 3 master + N worker 手工二进制部署路线——比起 kubeadmin/kubeadm，手工二进制能完全掌控每个组件的 flag、证书路径和 systemd 单元，是国内大厂 2018-2022 年生产环境的标配方案。

本文是整个 K8s 二进制部署系列的前置篇，只讲集群起步：网络规划、服务器角色划分、Ubuntu 22.04 系统基础优化、IPVS/iptables 内核参数准备。不涉及任何 K8s 组件安装。

适用版本：Kubernetes 1.28.5 / etcd 3.5.11 / Ubuntu 22.04 server 集群规模：3 master（etcd + controlplane + nginx/keepalived）+ 10 worker

1. 集群网络与节点规划

1.1 角色划分

角色	数量	部署组件
master	3	etcd、kube-apiserver、kube-controller-manager、kube-scheduler、kubelet、kube-proxy、nginx、keepalived
worker	N	kubelet、kube-proxy、容器运行时
VIP	2	一个内网用，一个公网用，keepalived 漂移

1.2 IP 与 VIP 段（示例拓扑）

网段	用途	说明
`<node-network>`	节点操作系统网段	物理网络，所有节点 IP 都在此
`10.96.0.0/12`	Service 网段	kube-apiserver 默认 `--service-cluster-ip-range`
`10.96.0.10`	CoreDNS Service IP	kube-dns 固定 IP
`172.218.0.0/12`	Pod 网络	kube-controller-manager 的 `--cluster-cidr`
`<vip-internal>`	内网 VIP	漂移在 master1 / master2
`<vip-public>`	公网映射 VIP	漂移在 master1 / master2

Service 段、Pod 段在生产中可以按需调整，前提是所有节点路由可达。

1.3 NodePort 端口范围

K8s 默认 Service NodePort 范围是 30000-32767，如果需要在前面用 Nginx 4 层反代 NodePort，必须在 kube-apiserver 启动参数里显式声明：

1

--service-node-port-range=30000-32767

修改 /etc/systemd/system/kube-apiserver.service 后：

1
2


systemctl daemon-reload
systemctl restart kube-apiserver.service

1.4 主机名与 hosts

1
2
3
4


# 三个 master 设置 hostname
hostnamectl set-hostname master1 # 内网示例：<node-ip> 节点
hostnamectl set-hostname master2
hostnamectl set-hostname master3

所有节点配置 /etc/hosts（内网 IP 替换为实际值）：

1
2
3
4
5
6
7


cat << "EOF" >> /etc/hosts
<node-ip> master1
<node-ip> master2
<node-ip> master3
<node-ip> worker1
<node-ip> worker2
EOF

注意：内网 IP 已用占位符 <node-ip> 替代；实际操作时换成你集群的真实 IP。

2. 安装文件清单

在 master1 上准备好所有需要分发的二进制包（典型命名）：

文件	用途
`docker-24.0.7.tgz`	Docker 静态二进制包
`cri-dockerd-0.3.9.amd64.tgz`	cri-dockerd（K8s 1.24+ 不再内置 dockershim）
`nginx-1.24.0.tar.gz`	nginx 源码（编译 `--with-stream`）
`keepalived-2.2.8.tar.gz`	keepalived 源码
`etcd-v3.5.11-linux-amd64.tar.gz`	etcd 二进制
`kubernetes-server-linux-amd64.tar.gz`	K8s 1.28.5 服务端二进制
`cfssl_1.6.4_linux_amd64`	证书签发工具
`cfssljson_1.6.4_linux_amd64`	证书 JSON 化工具
`cfssl-certinfo_1.6.4_linux_amd64`	证书查看工具
`helm-v3.13.3-linux-amd64.tar.gz`	Helm 包管理
`coredns-1.29.0.tgz`	CoreDNS chart（DNS 插件）

3. 目录约定

每台机器统一下列目录结构，方便后面所有组件统一路径：

 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


# 系统服务目录
mkdir -p /etc/systemd/system

# 二进制目录
mkdir -p /usr/local/bin

# 所有节点安装包路径
mkdir -p /data/softs

# 所有节点 k8s 配置目录
mkdir -p /data/k8scnf

# 所有节点的 k8s 证书目录
mkdir -p /etc/kubernetes/pki

# 所有节点 k8s 运行目录
mkdir -p /etc/kubernetes/manifests/
mkdir -p /etc/systemd/system/kubelet.service.d
mkdir -p /var/lib/kubelet
mkdir -p /var/log/kubernetes

# 所有节点 docker 配置目录
mkdir -p /etc/docker

# 所有 master etcd 证书目录
mkdir -p /etc/etcd/ssl

# master1, master2 的 nginx 目录
mkdir -p /usr/local/nginx/conf/{conf.d,stream}
mkdir -p /usr/local/keepalived/conf

4. 节点免密登录（master1 推送前置）

master1 上一键推送 SSH 公钥到其他节点（密码用占位符）：

1
2
3
4
5
6


apt install -y sshpass
ssh-keygen -f /root/.ssh/id_rsa -P ''

for HOST in master2 master3 worker1 worker2 worker3 worker4 worker5 worker6 worker7 worker8 worker9; do \
 sshpass -p "{{SSH_PASSWORD}}" ssh-copy-id -o StrictHostKeyChecking=no $HOST; \
done

非 22 端口的节点：

1

ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 10022 user@server"

文件批量推送到其他 master：

1

for HOST in master2 master3; do scp -r /data/softs $HOST:/data; done

5. Ubuntu 22.04 源切换

为了避免 apt install 卡顿，所有节点切到清华源（或其他国内源）：

 1
 2
 3
 4
 5
 6
 7
 8
 9
10


mv /etc/apt/sources.list /etc/apt/sources.listbak

cat > /etc/apt/sources.list << "EOF"
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ jammy main restricted universe multiverse
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ jammy-updates main restricted universe multiverse
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu/ jammy-backports main restricted universe multiverse
deb http://security.ubuntu.com/ubuntu/ jammy-security main restricted universe multiverse
EOF

apt update

6. 系统基础优化

6.1 禁用 swap

K8s 1.20+ 仍要求关闭 swap（kubelet 默认 failSwapOn: true）：

1
2
3
4


free -m
swapoff -a
sed -ri 's/.*swap.*/#&/' /etc/fstab
cat /etc/fstab

6.2 关闭防火墙

1
2
3


sudo systemctl stop ufw
sudo systemctl disable ufw
systemctl status ufw # 状态应为 inactive (dead)

6.3 关闭 systemd-resolved（关键）

Ubuntu 22.04 默认的 systemd-resolved 占用 53 端口，导致 CoreDNS 装不上：

1
2
3
4
5


sudo apt install -y unbound
sudo systemctl stop systemd-resolved
sudo systemctl disable systemd-resolved
sudo rm -rf /etc/resolv.conf
sudo touch /etc/resolv.conf

6.4 时间同步

1
2
3
4
5
6
7
8
9


timedatectl status # System clock synchronized 必须是 yes
timedatectl set-timezone Asia/Shanghai
apt install -y ntpdate
ntpdate time1.aliyun.com

# 定时同步
crontab -e
*/5 * * * * ntpdate time1.aliyun.com
crontab -l

常见报错：Waiting for cache lock: Could not get lock /var/lib/dpkg/lock-frontend 解决：用 lsof /var/lib/dpkg/lock-frontend 查到占用 PID 后 kill -9 杀掉。

6.5 ulimit 与文件描述符

 1
 2
 3
 4
 5
 6
 7
 8
 9
10


ulimit -SHn 65535

cat >> /etc/security/limits.conf << "EOF"
* soft nofile 655360
* hard nofile 131072
* soft nproc 655350
* hard nproc 655350
* soft memlock unlimited
* hard memlock unlimited
EOF

6.6 内核升级

K8s 1.28 要求内核 ≥ 4.18，Ubuntu 22.04 默认 5.15 满足，无需额外升级。

1
2


uname -r
# 5.15.0-91-generic 满足

7. IPVS 与内核网络参数

K8s kube-proxy 的 IPVS 模式比 iptables 模式在大规模集群下性能更好：

7.1 安装 IPVS

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


apt install -y ipvsadm ipset sysstat conntrack

cat >> /etc/modules-load.d/ipvs.conf << "EOF"
ip_vs
ip_vs_rr
ip_vs_wrr
ip_vs_sh
nf_conntrack
ip_tables
ip_set
xt_set
ipt_set
ipt_rpfilter
ipt_REJECT
ipip
EOF

systemctl restart systemd-modules-load.service
lsmod | grep -e ip_vs -e nf_conntrack

7.2 sysctl 优化

 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


cat << "EOF" > /etc/sysctl.d/k8s.conf
net.ipv4.ip_forward = 1
net.bridge.bridge-nf-call-iptables = 1
fs.may_detach_mounts = 1
vm.overcommit_memory = 1
vm.panic_on_oom = 0
fs.inotify.max_user_watches = 89100
fs.file-max = 52706963
fs.nr_open = 52706963
net.netfilter.nf_conntrack_max = 2310720

net.ipv4.tcp_keepalive_time = 600
net.ipv4.tcp_keepalive_probes = 3
net.ipv4.tcp_keepalive_intvl = 15
net.ipv4.tcp_max_tw_buckets = 36000
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_max_orphans = 327680
net.ipv4.tcp_orphan_retries = 3
net.ipv4.tcp_syncookies = 1
net.ipv4.tcp_max_syn_backlog = 16384
net.ipv4.ip_conntrack_max = 65536
net.ipv4.tcp_timestamps = 0
net.core.somaxconn = 16384
EOF

sysctl --system

8. 排错速查

现象	原因	解决
`apt` 卡 `dpkg lock-frontend`	后台 unattended-upgrade 占用	`lsof /var/lib/dpkg/lock-frontend` → `kill -9 PID`
CoreDNS Pod 一直 `CrashLoopBackOff`	systemd-resolved 占用 53	关闭 systemd-resolved，见 6.3
kubelet 启动失败 `failSwapOn`	swap 未关	`swapoff -a` + `/etc/fstab` 注释
NodePort 配 80 端口启动失败	不在 30000-32767 范围	kube-apiserver 加 `--service-node-port-range`
节点 `NotReady`	路由转发未开	`net.ipv4.ip_forward = 1`

9. 小结

集群前置看似都是"复制粘贴"的体力活，但每一步都是后面所有组件能起来的前提。最容易踩的坑是：

systemd-resolved 没关 → CoreDNS 起不来
swap 没禁 → kubelet 直接 failSwapOn 拒绝启动
防火墙没关 → 跨节点 Service 不通
时间不同步 → 证书签名失败、etcd 集群脑裂

下一步：容器运行时：Docker + cri-dockerd 二进制部署实战。

前置知识 / 下一步

前置：了解 Linux 基础（systemd、apt、sysctl）
本篇：所有 master/worker 节点一致的系统基线
下一步：在所有节点上部署容器运行时（Docker 24 + cri-dockerd）