Kubernetes集群生产环境搭建全过程

摘要：本文详细讲解如何搭建高可用的集群，以下简称由三台服务器组成集群命名为，，，用来代替集群搭建首先搭建集群为集群的核心组成部分，负责所有集群配置信息和服务信息的存储，所以必须要保证高可用，此处采用的静态服务发现，即在启动的时候，确定的。

本文详细讲解如何搭建高可用的Kubernetes集群，以下简称k8s

由三台服务器(CentOS 7.0)组成master集群,命名为m1，m2，m3，ip用m1 m2 m3来代替

首先搭建etcd集群,etcd为k8s集群的核心组成部分，负责所有集群配置信息和服务信息的存储，所以必须要保证高可用，此处采用etcd的静态服务发现，即在etcd启动的时候，确定etcd node的ip。

yum安装etcd yum install -y etcd

分别在三台机器启动etcd进程(实际操作中需要将m1 m2 m3替换成实际的ip地址)
m1:

etcd -name infra1 -initial-advertise-peer-urls http://m1:2380 -listen-peer-urls http://m1:2380 -listen-client-urls http://m1:2379,http://127.0.0.1:2379 -advertise-client-urls http://m1:2379 -initial-cluster-token etcd-cluster-1 -initial-cluster infra1=http://m1:2380,infra2=http://m2:2380,infra3=http://m3:2380 -initial-cluster-state new

m2:

etcd -name infra2 -initial-advertise-peer-urls http://m2:2380 -listen-peer-urls http://m2:2380 -listen-client-urls http://m2:2379,http://127.0.0.1:2379 -advertise-client-urls http://m2:2379 -initial-cluster-token etcd-cluster-1 -initial-cluster infra1=http://m1:2380,infra2=http://m2:2380,infra3=http://m3:2380 -initial-cluster-state new

m3:

etcd -name infra3 -initial-advertise-peer-urls http://m3:2380 -listen-peer-urls http://m3:2380 -listen-client-urls http://m3:2379,http://127.0.0.1:2379 -advertise-client-urls http://m3:2379 -initial-cluster-token etcd-cluster-1 -initial-cluster infra1=http://m1:2380,infra2=http://m2:2380,infra3=http://m3:2380 -initial-cluster-state new

集群启动后，选取一台机器执行 etcdctl cluster-health

如果出现3个类似 member cbfa6350b369c3a is healthy 的字样，说明etcd集群部署成功。

实战中采取了systemd来进行管理
新建文件 /usr/lib/systemd/system/etcd.service
内容如下：

[Unit]
Description=Etcd Server
After=network.target
After=network-online.target
Wants=network-online.target

[Service]
Type=notify
WorkingDirectory=/root
ExecStart=etcd -name infra1 -initial-advertise-peer-urls http://m1:2380 -listen-peer-urls http://m1:2380 -listen-client-urls http://m1:2379,http://127.0.0.1:2379 -advertise-client-urls http://m1:2379 -initial-cluster-token etcd-cluster-1 -initial-cluster infra1=http://m1:2380,infra2=http://m2:2380,infra3=http://m3:2380 -initial-cluster-state new
Restart=on-failure
LimitNOFILE=65536

[Install]
WantedBy=multi-user.target

这样可以使用 systemctl start etcd来启动服务

master搭建是通过自动化搭建脚本实现的 脚本内容如下

#!/bin/bash
echo "################ Prerequisites..."
systemctl stop firewalld
systemctl disable firewalld
yum -y install ntp
systemctl start ntpd
systemctl enable ntpd

echo "################ Installing flannel..."
#安装flannel
yum install flannel -y

echo "################ Add subnets for flannel..."
A_SUBNET=172.17.0.0/16
B_SUBNET=192.168.0.0/16
C_SUBNET=10.254.0.0/16

FLANNEL_SUBNET=$A_SUBNET
SERVICE_SUBNET=$B_SUBNET
OCCUPIED_IPs=(`ifconfig -a | grep "inet " | cut -d ":" -f 2 |cut -d " " -f 1 | grep -v "^127"`)
for ip in ${OCCUPIED_IPs[@]};do
    if [ $(ipcalc -n $ip/${A_SUBNET#*/}) == $(ipcalc -n ${A_SUBNET}) ];then
        FLANNEL_SUBNET=$C_SUBNET
        SERVICE_SUBNET=$B_SUBNET
        break
    fi
    if [ $(ipcalc -n $ip/${B_SUBNET#*/}) == $(ipcalc -n ${B_SUBNET}) ];then
        FLANNEL_SUBNET=$A_SUBNET
        SERVICE_SUBNET=$C_SUBNET
        break
    fi
    if [ $(ipcalc -n $ip/${C_SUBNET#*/}) == $(ipcalc -n ${C_SUBNET}) ];then
        FLANNEL_SUBNET=$A_SUBNET
        SERVICE_SUBNET=$B_SUBNET
        break
    fi
done

while ((1));do
    sleep 2
    etcdctl cluster-health
    flag=$?
    if [ $flag == 0 ];then
etcdctl mk /coreos.com/network/config "{"Network":""${FLANNEL_SUBNET}""}"
        break
    fi
done

echo "################ Starting flannel..."
#此处将m1 m2 m3 换成实际的ip
echo -e "FLANNEL_ETCD="http://m1:2379,http://m2:2379,http://m3:2379"
FLANNEL_ETCD_KEY="/coreos.com/network"" > /etc/sysconfig/flanneld
systemctl enable flanneld
systemctl start flanneld

echo "################ Installing K8S..."
yum -y install kubernetes
echo "KUBE_API_ADDRESS="--address=0.0.0.0"
KUBE_API_PORT="--port=8080"
KUBELET_PORT="--kubelet_port=10250"
KUBE_ETCD_SERVERS="--etcd_servers=http://m1:2379,http://m2:2379,http://m3:2379"
KUBE_SERVICE_ADDRESSES="--service-cluster-ip-range="${SERVICE_SUBNET}""
KUBE_ADMISSION_CONTROL="--admission_control=NamespaceLifecycle,NamespaceExists,LimitRanger,SecurityContextDeny,ResourceQuota"
KUBE_API_ARGS=""" > /etc/kubernetes/apiserver

echo "################ Start K8S components..."
for SERVICES in kube-apiserver kube-controller-manager kube-scheduler; do
    systemctl restart $SERVICES
    systemctl enable $SERVICES
    systemctl status $SERVICES
done

node的搭建也是通过自动化部署脚本实现的，脚本内容如下：

#!/bin/bash

echo "################ Prerequisites..."
#关闭firewall 开启ntp时间同步
systemctl stop firewalld
systemctl disable firewalld
yum -y install ntp
systemctl start ntpd
systemctl enable ntpd

#安装kubernetes所需要的几个软件
yum -y install kubernetes docker flannel bridge-utils


#此处使用了一个vip 命名为vip 实际部署时需要替换为你的集群的vip 使用此ip的服务有 kube-master（8080） registry（5000） skydns（53）

echo "################ Configuring nodes..."
echo "################ Configuring nodes > Find Kube master..."
KUBE_REGISTRY_IP="vip"
KUBE_MASTER_IP="vip"
echo "################ Configuring nodes > Configuring Minion..."
echo -e "KUBE_LOGTOSTDERR="--logtostderr=true"
KUBE_LOG_LEVEL="--v=0"
KUBE_ALLOW_PRIV="--allow_privileged=false"
KUBE_MASTER="--master=http://${KUBE_MASTER_IP}:8080"" > /etc/kubernetes/config
echo "################ Configuring nodes > Configuring kubelet..."
#取每个node机器的eth0的ip作为标识
KUBE_NODE_IP=`ifconfig eth0 | grep "inet " | awk "{print $2}"`

#api_servers 使用master1 master2 master3的ip数组形式
echo -e "KUBELET_ADDRESS="--address=0.0.0.0"
KUBELET_PORT="--port=10250"
KUBELET_HOSTNAME="--hostname_override=${KUBE_NODE_IP}"
KUBELET_API_SERVER="--api_servers=http://m1:8080,http://m2:8080,http://m3:8080"
KUBELET_ARGS="--cluster-dns=vip --cluster-domain=k8s --pod-infra-container-image=${KUBE_REGISTRY_IP}:5000/pause:latest"" > /etc/kubernetes/kubelet


#flannel读取etcd配置信息 为本机的docker0分配ip 保证node集群子网互通
echo "################ Configuring flannel..."
echo -e "FLANNEL_ETCD="http://m1:2379,http://m2:2379,http://m3:2379"
FLANNEL_ETCD_KEY="/coreos.com/network"" > /etc/sysconfig/flanneld


echo "################ Accept private registry..."
echo "OPTIONS="--selinux-enabled --insecure-registry ${KUBE_REGISTRY_IP}:5000"
DOCKER_CERT_PATH=/etc/docker" > /etc/sysconfig/docker

echo "################ Start K8S Components..."
systemctl daemon-reload
for SERVICES in kube-proxy flanneld; do
    systemctl restart $SERVICES
    systemctl enable $SERVICES
    systemctl status $SERVICES
done

echo "################ Resolve interface conflicts..."
systemctl stop docker
ifconfig docker0 down
brctl delbr docker0

echo "################ Accept private registry..."
echo -e "OPTIONS="--selinux-enabled --insecure-registry ${KUBE_REGISTRY_IP}:5000"
DOCKER_CERT_PATH=/etc/docker" > /etc/sysconfig/docker

for SERVICES in docker kubelet; do
    systemctl restart $SERVICES
    systemctl enable $SERVICES
    systemctl status $SERVICES
done

至此，Kubernetes master和node的搭建就结束了。