Kubernetes作为容器集群系统,通过健康检查+重启策略实现了Pod故障自我修复能力,通过调度算法实现将Pod分布式部署,并保持预期副本数,根据Node失效状态自动在其他Node拉起Pod,实现了应用层的高可用性。
针对Kubernetes集群,高可用性还应包含以下两个层面的考虑:Etcd数据库的高可用性和Kubernetes Master组件的高可用性。而Etcd我们已经采用3个节点组建集群实现高可用,本节将对Master节点高可用进行说明和实施。
Master节点扮演着总控中心的角色,通过不断与工作节点上的Kubelet进行通信来维护整个集群的健康工作状态。如果Master节点故障,将无法使用kubectl工具或者API做任何集群管理。
Master节点主要有三个服务kube-apiserver、kube-controller-mansger和kube-scheduler,其中kube-controller-mansger和kube-scheduler组件自身通过选择机制已经实现了高可用,所以Master高可用主要针对kube-apiserver组件,而该组件是以HTTP API提供服务,因此对他高可用与Web服务器类似,增加负载均衡器对其负载均衡即可,并且可水平扩容。
根据服务器整体规划和测试资源的限制,这次我们高可用涉及的三台服务器部署(橙色字体):
二进制单master部署可以参考:https://www.cnblogs.com/huanglingfa/p/13773234.html
角色 |
IP |
组件 |
k8s-master-1 |
192.168.10.160 |
kube-apiserver,kube-controller-manager,kube-scheduler,etcd |
k8s-master-2 |
192.168.10.166 |
kube-apiserver,kube-controller-manager,kube-scheduler |
k8s-node-1 |
192.168.10.161 |
kubelet,kube-proxy,docker etcd |
k8s-node-2 |
192.168.10.162 |
kubelet,kube-proxy,docker,etcd |
K8s-lb-master |
192.168.10.164 ,192.168.10.168 (VIP) |
Nginx L4 ,keepalived |
K8s-lb-backup |
192.168.10.165 |
Nginx L4, keepalived |
多Master架构图:
基础优化
1、时间同步 echo "#time sync by fage at 2020-7-22" >>/var/spool/cron/root && echo "*/5 * * * * /usr/sbin/ntpdate ntp1.aliyun.com >/dev/null 2>&1" >>/var/spool/cron/root && systemctl restart crond.service 2、关闭防火墙和selinux systemctl stop firewalld && systemctl disable firewalld && setenforce 0 && sed -i s#SELINUX=enforcing#SELINUX=disable#g /etc/selinux/config 3、更改主机名 hostname k8s-master-2 echo "k8s-master-2" >/etc/hostname hostname k8s-lb-A echo "k8s-lb-A" >/etc/hostname hostname k8s-lb-B echo " k8s-lb-B" >/etc/hostname 4、更改hosts文件 cat >/etc/hosts <<EOF 127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 ::1 localhost localhost.localdomain localhost6 localhost6.localdomain6 192.168.10.160 k8s-master-1 192.168.10.161 k8s-node-1 192.168.10.162 k8s-node-2 192.168.10.163 k8s-node-3 192.168.10.164 k8s-lb-A 192.168.10.165 k8s-lb-B 192.168.10.166 k8s-master-2 EOF 5、节点node要禁用swap设备 不禁用要配置声明 swapoff -a sed -i "s@/dev/mapper/centos-swap swap@#/dev/mapper/centos-swap swap@g" /etc/fstab 6、将桥接的IPv4流量传递到iptables的链: #注意nginx不做 cat > /etc/sysctl.d/k8s.conf << EOF net.ipv4.ip_forward = 1 net.bridge.bridge-nf-call-ip6tables = 1 net.bridge.bridge-nf-call-iptables = 1 EOF sysctl --system #设置时区 cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime systemctl restart crond.service
提示:如果可以的话集群master与node做master连接node的ssh 免密,操作起来会比较方便
一、安装Docker
#直接复制master-1上的docker文件到master-2上 cd /usr/bin/ scp -r /usr/lib/systemd/system/docker.service root@192.168.10.166:/usr/lib/systemd/system/ scp -r containerd containerd-shim docker dockerd docker-init docker-proxy runc root@192.168.10.166:/usr/bin/ scp -r /etc/docker root@192.168.10.166:/etc/ systemctl daemon-reload && systemctl start docker &&systemctl enable docker #或者使用yum镜像安装 wget https://mirrors.aliyun.com/docker-ce/linux/centos/docker-ce.repo -O /etc/yum.repos.d/docker-ce.repo yum -y install docker-ce-18.06.1.ce-3.el7 systemctl enable docker && systemctl start docker docker --version #docker镜像加速器 mkdir -p /etc/docker cat > /etc/docker/daemon.json << EOF { "registry-mirrors": ["https://b9pmyelo.mirror.aliyuncs.com"] } EOF systemctl daemon-reload systemctl restart docker.service docker info
二、部署Master-2 Node(192.168.10.166)
Master-2 与已部署的Master1所有操作一致。所以我们只需将Master-1所有K8s文件拷贝过来,再修改下服务器IP和主机名启动即可。
1. 创建etcd证书目录
在Master-2创建etcd证书目录:
mkdir -p /opt/etcd/ssl
2. 拷贝文件(Master1操作)
拷贝Master-1上所有K8s文件和etcd证书到Master-2:
scp -r /opt/kubernetes root@192.168.10.166:/opt scp -r /opt/cni/ root@192.168.10.166:/opt scp -r /opt/etcd/ssl root@192.168.10.166:/opt/etcd scp /usr/lib/systemd/system/kube* root@192.168.10.166:/usr/lib/systemd/system scp /usr/bin/kubectl root@192.168.10.166:/usr/bin
3. 删除证书文件(master-2操作)
删除kubelet证书和kubeconfig文件:
rm -f /opt/kubernetes/cfg/kubelet.kubeconfig rm -f /opt/kubernetes/ssl/kubelet*
4. 修改配置文件IP和主机名(master-2操作)
修改apiserver、kubelet和kube-proxy配置文件为本地IP:
vi /opt/kubernetes/cfg/kube-apiserver.conf ... --bind-address=192.168.10.166 --advertise-address=192.168.10.166 ... vi /opt/kubernetes/cfg/kubelet.conf --hostname-override=k8s-master-2 vi /opt/kubernetes/cfg/kube-proxy-config.yml hostnameOverride: k8s-master-2
5. 启动设置开机启动(master-2操作)
systemctl daemon-reload systemctl start kube-apiserver systemctl start kube-controller-manager systemctl start kube-scheduler systemctl start kubelet systemctl start kube-proxy systemctl enable kube-apiserver systemctl enable kube-controller-manager systemctl enable kube-scheduler systemctl enable kubelet systemctl enable kube-proxy
6. 查看集群状态(k8s-master-2查看)
kubectl get cs NAME STATUS MESSAGE ERROR scheduler Healthy ok controller-manager Healthy ok etcd-1 Healthy {"health":"true"} etcd-2 Healthy {"health":"true"} etcd-0 Healthy {"health":"true"}
7. 批准kubelet证书申请(Master1操作)
kubectl get csr NAME AGE SIGNERNAME REQUESTOR CONDITION node-csr-JYNknakEa_YpHz797oKaN-ZTk43nD51Zc9CJkBLcASU 12m kubernetes.io/kube-apiserver-client-kubelet kubelet-bootstrap Pending kubectl certificate approve node-csr-JYNknakEa_YpHz797oKaN-ZTk43nD51Zc9CJkBLcASU kubectl get node NAME STATUS ROLES AGE VERSION k8s-master Ready <none> 34h v1.18.6 k8s-master2 Ready <none> 13m v1.18.6 k8s-node1 Ready <none> 33h v1.18.6 k8s-node2 Ready <none> 33h v1.18.6
三、 部署Nginx负载均衡器
kube-apiserver高可用架构图:
- Nginx是一个主流Web服务和反向代理服务器,这里用四层实现对apiserver实现负载均衡。
- Keepalived是一个主流高可用软件,基于VIP绑定实现服务器双机热备,在上述拓扑中,Keepalived主要根据Nginx运行状态判断是否需要故障转移(偏移VIP),例如当Nginx主节点挂掉,VIP会自动绑定在Nginx备节点,从而保证VIP一直可用,实现Nginx高可用。
1. 安装软件包(k8s-lb主/备)
yum install epel-release -y yum install nginx keepalived -y
2. Nginx配置文件(k8s-lb主/备一样)
cat > /etc/nginx/nginx.conf << "EOF" user nginx; worker_processes auto; error_log /var/log/nginx/error.log; pid /run/nginx.pid; include /usr/share/nginx/modules/*.conf; events { worker_connections 1024; } # 四层负载均衡,为两台Master apiserver组件提供负载均衡 stream { log_format main '$remote_addr $upstream_addr - [$time_local] $status $upstream_bytes_sent'; access_log /var/log/nginx/k8s-access.log main; upstream k8s-apiserver { server 192.168.10.160:6443; # Master1 APISERVER IP:PORT server 192.168.10.166:6443; # Master2 APISERVER IP:PORT } server { listen 6443; proxy_pass k8s-apiserver; } } http { log_format main '$remote_addr - $remote_user [$time_local] "$request" ' '$status $body_bytes_sent "$http_referer" ' '"$http_user_agent" "$http_x_forwarded_for"'; access_log /var/log/nginx/access.log main; sendfile on; tcp_nopush on; tcp_nodelay on; keepalive_timeout 65; types_hash_max_size 2048; include /etc/nginx/mime.types; default_type application/octet-stream; server { listen 80 default_server; server_name _; location / { } } } EOF
3. keepalived配置文件(k8s-lb-master)
cat > /etc/keepalived/keepalived.conf << EOF global_defs { notification_email { acassen@firewall.loc failover@firewall.loc sysadmin@firewall.loc } notification_email_from fage@qq.com smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id NGINX_MASTER } vrrp_script check_nginx { script "/etc/keepalived/check_nginx.sh" } vrrp_instance VI_1 { state MASTER interface eth0 virtual_router_id 51 # VRRP 路由 ID实例,每个实例是唯一的 priority 100 # 优先级,备服务器设置 90 advert_int 1 # 指定VRRP 心跳包通告间隔时间,默认1秒 authentication { auth_type PASS auth_pass 1111 } # 虚拟IP virtual_ipaddress { 192.168.10.168/24 } track_script { check_nginx } } EOF
- vrrp_script:指定检查nginx工作状态脚本(根据nginx状态判断是否故障转移)
- virtual_ipaddress:虚拟IP(VIP)
检查nginx状态脚本:
cat > /etc/keepalived/check_nginx.sh << "EOF" #!/bin/bash count=$(ps -ef |grep nginx |egrep -cv "grep|$$") if [ "$count" -eq 0 ];then exit 1 else exit 0 fi EOF chmod +x /etc/keepalived/check_nginx.sh
4. keepalived配置文件(k8s-lb-backup)
cat > /etc/keepalived/keepalived.conf << EOF global_defs { notification_email { acassen@firewall.loc failover@firewall.loc sysadmin@firewall.loc } notification_email_from fage@qq.com smtp_server 127.0.0.1 smtp_connect_timeout 30 router_id NGINX_BACKUP } vrrp_script check_nginx { script "/etc/keepalived/check_nginx.sh" } vrrp_instance VI_1 { state BACKUP interface eth0 virtual_router_id 51 # VRRP 路由 ID实例,每个实例是唯一的 priority 90 advert_int 1 authentication { auth_type PASS auth_pass 1111 } virtual_ipaddress { 192.168.10.168/24 } track_script { check_nginx } } EOF
上述配置文件中检查nginx运行状态脚本:
cat > /etc/keepalived/check_nginx.sh << "EOF" #!/bin/bash count=$(ps -ef |grep nginx |egrep -cv "grep|$$") if [ "$count" -eq 0 ];then exit 1 else exit 0 fi EOF chmod +x /etc/keepalived/check_nginx.sh
注:keepalived根据脚本返回状态码(0为工作正常,非0不正常)判断是否故障转移。
5. 启动并设置开机启动(k8s-lb两个节点都操作)
systemctl daemon-reload systemctl start nginx && systemctl enable nginx && systemctl status nginx systemctl start keepalived && systemctl enable keepalived && systemctl status keepalived
6. 查看keepalived工作状态
ip a | grep 192 inet 192.168.10.164/24 brd 192.168.10.255 scope global noprefixroute eth0 inet 192.168.10.168/24 scope global secondary eth0
可以看到,在eth0网卡绑定了192.168.10.168 虚拟IP,说明工作正常。
7. Nginx+Keepalived高可用测试
关闭主节点Nginx,测试VIP是否漂移到备节点服务器。
在Nginx Master执行 pkill nginx
在Nginx Backup,ip addr命令查看已成功绑定VIP。
8. 访问负载均衡器测试
找K8s集群中任意一个节点,使用curl查看K8s版本测试,使用VIP访问:
curl -k https://192.168.10.168:6443/version { "major": "1", "minor": "18", "gitVersion": "v1.18.6", "gitCommit": "2e7996e3e2712684bc73f0dec0200d64eec7fe40", "gitTreeState": "clean", "buildDate": "2020-05-20T12:43:34Z", "goVersion": "go1.13.9", "compiler": "gc", "platform": "linux/amd64" }
可以正确获取到K8s版本信息,说明负载均衡器搭建正常。该请求数据流程:curl -> vip(nginx) -> apiserver
通过查看Nginx日志也可以看到转发apiserver IP:
tail -f /var/log/nginx/k8s-access.log
192.168.10.164 192.168.10.160:6443 - [30/May/2020:11:15:10
+0800] 200 422
192.168.10.164 192.168.10.166:6443 - [30/May/2020:11:15:26
+0800] 200 422
到此还没结束,还有下面最关键的一步。
四、修改所有Worker Node连接LB VIP
试想下,虽然我们增加了Master-2和负载均衡器,但是我们是从单Master架构扩容的,也就是说目前所有的Node组件连接都还是Master-1,如果不改为连接VIP走负载均衡器,那么Master还是单点故障。
因此接下来就是要改所有Node组件配置文件,由原来192.168.10.160修改为192.168.10.168(VIP)
角色 |
IP |
k8s-master1 |
192.168.10.160 |
k8s-master2 |
192.168.10.166 |
k8s-node1 |
192.168.10.161 |
k8s-node2 |
192.168.10.162 |
也就是通过kubectl get node命令查看到的节点。
在上述所有Worker Node执行:
sed -i 's#192.168.10.160:6443#192.168.10.168:6443#' /opt/kubernetes/cfg/* systemctl restart kubelet systemctl restart kube-proxy
检查节点状态:
kubectl get node NAME STATUS ROLES AGE VERSION k8s-master Ready <none> 34h v1.18.6 k8s-master2 Ready <none> 18m v1.18.6 k8s-node1 Ready <none> 33h v1.18.6 k8s-node2 Ready <none> 33h v1.18.6
至此,一套完整的 Kubernetes 高可用集群就部署完成了!
PS:如果是在公有云上,一般都不支持keepalived,那么可以直接用它们的负载均衡器产品(内网就行,还免费~),架构与上面一样,直接负载均衡多台Master kube-apiserver即可!