kubespray

部署生产环境就绪的kubernetes集群

项目地址

GitHub - kubernetes-sigs/kubespray: Deploy a Production Ready Kubernetes ClusterGitHub

快速开始

前置步骤：

• 关闭防火墙或配置服务器间的互信任策略

• 系统时钟同步

• ansible 节点到 k8s 节点的 ssh 免密登录

1、下载kubespray

[e-mail]:kubernetes-sigs/kubespray.git && cd kubespray && git checkout release-2.15

2、安装python3与pip3

ubuntu环境

sudo apt install python3

centos7环境

sudo yum install python3

3、安装python依赖包

python3 -m pip install -r requirements.txt

如果已经通过其他方式安装过ansible，请先执行卸载，以确保ansible使用的是python3

4、复制示例配置

cp -rfp inventory/sample inventory/mycluster

5、编辑inventory清单

cat inventory/mycluster/inventory.ini

6、查看并编辑group_vars参数

cat inventory/mycluster/group_vars/all/all.yml
cat inventory/mycluster/group_vars/k8s-cluster/k8s-cluster.yml

7、执行安装

ansible-playbook -i inventory/mycluster/hosts.yaml  --become --become-user=root cluster.yml

拉取k8s.gcr.io的镜像时可能会出错，建议预先从其他的镜像地址中先拉取好，在inventory/mycluster/hosts.yaml中配置download_always_pull: false避免重复拉取已经存在的镜像；

拉取gcr.io的镜像可以配置国内镜像源gcr_image_repo: registry.aliyuncs.com/

8、查看集群、节点以及 Pod 状态

kubectl cluster-info
kubectl get node -o wide
kubectl get pod -o wide

通过 kubespray 镜像部署

1、拉取 kubespray 代码

[e-mail]:kubernetes-sigs/kubespray.git && \
cd kubespray && \
git checkout v2.19.0

2、拉取运行镜像

docker run --rm -it -v $(pwd)/inventory/sample:/inventory -v ${HOME}/.ssh/id_rsa:/root/.ssh/id_rsa quay.io/kubespray/kubespray:v2.19.0 bash

注意：运行 kubespray 容器的节点不能与 k8s 节点相同，否则部署中途 kubespray 容器会被中止。

3、编辑 inventory 清单

cat /inventory/inventory.ini

4、查看并编辑 group_vars 参数

cat /inventory/group_vars/all/all.yml
cat /inventory/group_vars/k8s-cluster/k8s-cluster.yml

5、执行集群部署

ansible-playbook -i /inventory/inventory.ini --private-key /root/.ssh/id_rsa cluster.yml

组件安装

kube-dashboard

1、配置 dashboard

inventory/mycluster/group_vars/k8s-cluster/addons.yml

dashboard_enabled: true
dashboard_token_ttl: 43200

2、安装 dashboard

ansible-playbook -i inventory/mycluster/hosts.yaml --tags "apps" --become --become-user=root cluster.yml

3、通过 NodePort 方式暴露 dashboard 服务

kubectl edit svc kubernetes-dashboard -n kube-system

apiVersion: v1
kind: Service
metadata:
  ...
  name: kubernetes-dashboard
  namespace: kube-system
spec:
  ...
  ports:
  - nodePort: 30000
    port: 443
    protocol: TCP
    targetPort: 8443
  ...
  type: NodePort

4、创建集群管理用户

kubectl create sa admin -n kube-system
kubectl create clusterrolebinding --clusterrole='cluster-admin' --serviceaccount=kube-system:admin admin

5、通过浏览器访问 proxy 节点的 30000 端口，使用下方命令获取 token 登录 dashboard

kubectl get secret -n kube-system | grep admin | awk '{print $1}' | xargs kubectl -n kube-system get secret -o jsonpath='{.data.token}' | base64 --decode

metrics-server

1、配置 metrics-server

inventory/mycluster/group_vars/k8s-cluster/addons.yml

metrics_server_enabled: true

2、安装 metrics-server

ansible-playbook -i inventory/mycluster/hosts.yaml --tags "apps" --become --become-user=root cluster.yml

3、访问 dashboard 访问 pod 可看到运行指标

helm

1、配置 helm

inventory/mycluster/group_vars/k8s-cluster/addons.yml

helm_enabled: true

2、安装 helm

ansible-playbook -i inventory/mycluster/hosts.yaml --tags "apps" --become --become-user=root cluster.yml

3、在 [kube-master] 节点上可执行 helm 指令

local-path-provisioner

1、配置 local-path-provisioner

inventory/mycluster/group_vars/k8s-cluster/addons.yml

local_path_provisioner_enabled: true
local_path_provisioner_claim_root: /data/k8s/local-path-provisioner/
local_path_provisioner_image_tag: "v0.0.17"

2、安装 local-path-provisioner

ansible-playbook -i inventory/mycluster/hosts.yaml --tags "apps" --become --become-user

集群配置

设置 dns上游

1、配置 dns 使用主机 resolve.conf 并设置 dns 上游服务器地址

inventory/mycluster/group_vars/all/all.yml

upstream_dns_servers:
- <ip>:<port>

inventory/k8s-cluster/k8s-cluster.yml

2、更新 dns

ansible-playbook -i inventory/mycluster/hosts.yaml --tags "apps,resolvconf" --

离线安装

在某些私有部署的场景下，无法连接互联网进行下载安装，这需要提前将安装所需的各种依赖资源下载到本地，方法如下：

https://github.com/kubernetes-sigs/kubespray/blob/master/docs/offline-environment.mdgithub.com