kube-apiserver 高可用，keepalived + haproxy

为什么要做高可用

在生产环境中，kubernetes 集群中会多多个 master 节点，每个 master 节点上都会部署 kube-apiserver 服务，实现高可用。但是 client 访问 kube-apiserver 时，需要指定 ip 或者域名，这样会出现单点故障。官方推荐的做法是使用一个负载均衡器，将多个 kube-apiserver 服务负载均衡，实现高可用，但很多时候我们是没有这个条件的。这时候就得想想办法了，比如 nignx 转发，但是 nginx 也是单点。域名的方式，但是这种方式生效时间较长，不太适合紧急情况。所以这里介绍一种使用 keepalived + haproxy 的方式实现 kube-apiserver 的高可用。这是一共公用 IP 的方式，当主节点宕机时，VIP 会自动切换到备节点，实现高可用。

环境准备

• master1: 192.168.31.203
• master2: 192.168.31.34
• master3: 192.168.31.46
• worker1: 192.168.31.25
• VIP （虚拟IP）: 192.168.31.230

安装

sudo apt install keepalived haproxy

systemctl enable haproxy
systemctl restart haproxy

systemctl enable keepalived
# 没有配置会出现错误 不用管
systemctl restart keepalived

配置 keepalived

配置文件

编辑 keepalived 配置文件

编辑 /etc/keepalived/keepalived.conf

master1：

# 健康检查 查看 haproxy 的进程在不在
vrrp_script chk_haproxy {
  script "killall -0 haproxy"
  interval 2 # 多少秒教程一次
  weight 3 # 成功了优先级加多少
}

vrrp_instance haproxy-vip {
  state MASTER # MASTER / BACKUP 1 MASTER 2 BACKUP
  priority 100 # 优先级 强的机器高一些 三台master 分别 100 99 98
  interface enp0s3     # 网卡名称
  virtual_router_id 51 # 路由 ip 默认就好
  advert_int 1 # keepalived 之间广播频率 秒
  authentication {
    auth_type PASS
    auth_pass test_k8s
  }
  unicast_src_ip 192.168.31.203 # 自己和其他 keepalived 通信地址
  unicast_peer {
    192.168.31.34                    # master2 的 IP 地址
    192.168.31.46                     # master3 的 IP 地址
  }

  virtual_ipaddress {
    192.168.31.230 # 这里必须和其他所有的ip 在一个局域网下
  }

  track_script {
    chk_haproxy
  }
}

master2：

vrrp_script chk_haproxy {
  script "killall -0 haproxy"
  interval 2
  weight 3
}

vrrp_instance haproxy-vip {
  state BACKUP
  priority 99
  interface enp0s3
  virtual_router_id 51
  advert_int 1
  authentication {
    auth_type PASS
    auth_pass test_k8s
  }
  unicast_src_ip 192.168.31.34
  unicast_peer {
    192.168.31.203
    192.168.31.46
  }

  virtual_ipaddress {
    192.168.31.230
  }

  track_script {
    chk_haproxy
  }
}

master3：

vrrp_script chk_haproxy {
  script "killall -0 haproxy"
  interval 2
  weight 3
}

vrrp_instance haproxy-vip {
  state BACKUP
  priority 98
  interface enp0s3
  virtual_router_id 51
  advert_int 1
  authentication {
    auth_type PASS
    auth_pass test_k8s
  }
  unicast_src_ip 192.168.31.46
  unicast_peer {
    192.168.31.203
    192.168.31.34
  }

  virtual_ipaddress {
    192.168.31.230
  }

  track_script {
    chk_haproxy
  }
}

测试

重启所有几点的 keepalived ， 虚拟 ip 会在节点 master 上，因为他的优先级高。

# master 1
ip a show enp0s3
2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 08:00:27:ca:59:86 brd ff:ff:ff:ff:ff:ff
    inet 192.168.31.203/24 metric 100 brd 192.168.31.255 scope global dynamic enp0s3
       valid_lft 41983sec preferred_lft 41983sec
    inet 192.168.31.230/32 scope global enp0s3
       valid_lft forever preferred_lft forever
    inet6 fe80::a00:27ff:feca:5986/64 scope link
       valid_lft forever preferred_lft forever

现在我们关掉 master1 的 haproxy 或者 keepalived

systemctl stop haproxy
# 再查看网络信息 发现虚拟ip 没了
ip a show enp0s3
2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 08:00:27:ca:59:86 brd ff:ff:ff:ff:ff:ff
    inet 192.168.31.203/24 metric 100 brd 192.168.31.255 scope global dynamic enp0s3
       valid_lft 41925sec preferred_lft 41925sec
    inet6 fe80::a00:27ff:feca:5986/64 scope link
       valid_lft forever preferred_lft forever

# 在优先级第二高的 master IP 上看下网络
ip a show enp0s3
2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 08:00:27:11:af:4f brd ff:ff:ff:ff:ff:ff
    inet 192.168.31.34/24 metric 100 brd 192.168.31.255 scope global dynamic enp0s3
       valid_lft 41857sec preferred_lft 41857sec
    inet 192.168.31.230/32 scope global enp0s3
       valid_lft forever preferred_lft forever
    inet6 fe80::a00:27ff:fe11:af4f/64 scope link
       valid_lft forever preferred_lft forever

# 启动 master1 的 haproxy ip就会回来

配置 haproxy

把 16443 端口的请求转发到 6443 端口 （3 master 的 kube-apiserver 对外端口）

/etc/haproxy/haproxy.cfg

global
    log /dev/log  local0 warning
    chroot      /var/lib/haproxy
    pidfile     /var/run/haproxy.pid
    maxconn     4000
    user        haproxy
    group       haproxy
    daemon

    stats socket /var/lib/haproxy/stats

defaults
    log global
    option  httplog
    option  dontlognull
    timeout connect 5000
    timeout client 50000
    timeout server 50000

frontend kube-apiserver
    bind *:16443
    mode tcp
    option tcplog
    default_backend kube-apiserver

backend kube-apiserver
    mode tcp
    option tcp-check
    balance roundrobin
    default-server inter 10s downinter 5s rise 2 fall 2 slowstart 60s maxconn 250 maxqueue 256 weight 100
    server kube-apiserver-1 192.168.31.203:6443 check
    server kube-apiserver-2 192.168.31.34:6443 check
    server kube-apiserver-3 192.168.31.46:6443 check

安装 kubernetes 集群

master1

kubeadm init --image-repository registry.aliyuncs.com/google_containers --control-plane-endpoint=192.168.31.230:16443 --v=10

master2 和 master3 加入集群

kubeadm join 192.168.31.230:16443 --token rxblci.ddh60vl370wjgtn7         --discovery-token-ca-cert-hash sha256:d712016d5b8ba4ae5c4a1bda8b6ab1944c13a04757d2c488dd0aefcfd1af0157   --certificate-key    c398d693c6ce9b664634c9b670f013da3010580c00bd444caf7d0a5a81e803f5         --control-plane --v=10

worker 加入集群

kubeadm join 192.168.31.230:16443 --token rxblci.ddh60vl370wjgtn7 \
        --discovery-token-ca-cert-hash sha256:d712016d5b8ba4ae5c4a1bda8b6ab1944c13a04757d2c488dd0aefcfd1af0157

查看集群状态

kubectl get node
NAME      STATUS     ROLES           AGE     VERSION
master1   Ready      control-plane   21m     v1.28.2
master2   Ready      control-plane   3m46s   v1.28.12
master3   Ready      control-plane   2m12s   v1.28.12
worker1   Ready      <none>          5s      v1.28.2

测试

#  关闭 master1 的 kubelet 和 apiserver
systemctl stop kubelet
sudo kill -9 $(pgrep kube-apiserver)

kubectl get node
NAME      STATUS     ROLES           AGE     VERSION
master1   NotReady   control-plane   25m     v1.28.2
master2   Ready      control-plane   7m40s   v1.28.12
master3   Ready      control-plane   6m6s    v1.28.12
worker1   Ready      <none>          3m59s   v1.28.2

# 关闭 master1 的 haproxy
systemctl stop haproxy
root@master1:/home/zhy# kubectl get node
NAME      STATUS     ROLES           AGE     VERSION
master1   NotReady   control-plane   26m     v1.28.2
master2   Ready      control-plane   9m12s   v1.28.12
master3   Ready      control-plane   7m38s   v1.28.12
worker1   Ready      <none>          5m31s   v1.28.2

# 关闭 master2 的 keepalived
kubectl get node
NAME      STATUS     ROLES           AGE     VERSION
master1   NotReady   control-plane   28m     v1.28.2
master2   Ready      control-plane   10m     v1.28.12
master3   Ready      control-plane   9m12s   v1.28.12
worker1   Ready      <none>          7m5s    v1.28.2

# 可以看到 虚拟ip 跑到了 master3 上
ip a show enp0s3
2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 08:00:27:f1:b5:ae brd ff:ff:ff:ff:ff:ff
    inet 192.168.31.46/24 metric 100 brd 192.168.31.255 scope global dynamic enp0s3
       valid_lft 41021sec preferred_lft 41021sec
    inet 192.168.31.230/32 scope global enp0s3
       valid_lft forever preferred_lft forever
    inet6 fe80::a00:27ff:fef1:b5ae/64 scope link
       valid_lft forever preferred_lft forever