参考文档：https://cooting.cn/archives/166.html

1.找到故障节点

先在ha上摘除故障节点
删除节点

kubectl get node

删除故障的master节点

kubectl delete node d0-office-master003

登陆其他可用的master节点的etcd节点

kubectl -n kube-system exec -it etcd-d0-office-master001 sh

定义别名alias

alias etcdctlold='etcdctl --endpoints=https://192.168.1.101:2379,https://192.168.1.102:2379,https://192.168.1.103:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/server.crt --key=/etc/kubernetes/pki/etcd/server.key'

查看集群状态获取etcd的id

etcdctlold endpoint status

查看集群节点

etcdctlold member list

删除废弃节点(此次我们删掉https://192.168.1.103:2379对应的节点)

etcdctlold member remove 75abeddc78aef692

------------------------------------------------------到这里master节点已删除完毕，为确保没有问题，我们再次查看

重新定义alias

alias etcdctlnew='etcdctl --endpoints=https://192.168.1.101:2379,https://192.168.1.102:2379 --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/server.crt --key=/etc/kubernetes/pki/etcd/server.key'

查看集群状态

etcdctlnew endpoint status

查看集群节点

etcdctlnew member list

注意：删除master期间不可修改集群组件和配置，不可重启服务（如果有hpa需要先停掉），不要让master节点长期处于偶数个