操作系统信息

本地：M1 macpro 14寸

云端：腾讯云轻量云主机2c4g

安装流程

个人强烈建议使用allinone安装，链接：https://kubesphere.io/zh/docs/v3.3/quick-start/all-in-one-on-linux/

手动部署k8s

在我自己用sealos部署的k8s上手动部署kubsphere碰到问题

没有默认的storageClass - 手动创建个local解决
error: http2: client connection lost - 多次重试解决
k8s版本太高 - sealos支持的k8s版本比kubesphere的版本高很多

Allinone部署k8s+ks

如果有问题，可以exec到ks-installer的pod里，看下results/ks-config里边的详细日志

问题解决 1. Failed to ansible-playbook ks-config.yaml 把ks-installer的pod删了再来一遍即可

至此，ks和k8s就在腾讯云主机上部署完了

暴露公网ip

此时从~/.kube/config中获取kubeconfig是无法在mac上直接访问的，会报${公网ip}invaild，如下解决

// 删除当前apiserver证书 
1. rm /etc/kubernetes/pki/apiserver.* 

// 这里包含上面报错时的所有ip，再加上一个域名lb.kubesphere.local，不加会导致kubelet等组件无法连接apiserver 
2. kubeadm init phase certs apiserver --apiserver-cert-extra-sans 10.233.3.1,10.0.12.27,127.0.0.1,101.35.48.16,lb.kubesphere.local 

3. kubeadm certs renew all 
4. 重启一下云主机或者各管理组件

本地开发ks

先上官方文档 - https://github.com/kubesphere/community/blob/master/developer-guide/development/quickstart.md

安装telepresence

用来从本地访问集群svc - 官方文档

brew install datawire/blackbird/telepresence-arm64
telepresence helm install
telepresence connect ~

Connected to context kubernetes-admin@cluster.local (https://xxx:6443)

curl -ik https://kubernetes.default

这里通了就完事了

clone kubesphere

git clone - 仓库
把quickstart中的kubesphere.yaml放入ks-apiserver目录下

本地启动报错

W0421 23:50:27.682958   16878 reflector.go:424] pkg/client/informers/externalversions/factory.go:129: failed to list *v1alpha3.DevOpsProject: the server could not find the requested resource (get devopsprojects.devops.kubesphere.io)
E0421 23:50:27.682985   16878 reflector.go:140] pkg/client/informers/externalversions/factory.go:129: Failed to watch *v1alpha3.DevOpsProject: failed to list *v1alpha3.DevOpsProject: the server could not find the requested resource (get devopsprojects.devops.kubesphere.io)
W0421 23:50:27.896590   16878 reflector.go:424] pkg/client/informers/externalversions/factory.go:129: failed to list *v2beta1.GlobalRuleGroup: the server could not find the requested resource (get globalrulegroups.alerting.kubesphere.io)
E0421 23:50:27.896619   16878 reflector.go:140] pkg/client/informers/externalversions/factory.go:129: Failed to watch *v2beta1.GlobalRuleGroup: failed to list *v2beta1.GlobalRuleGroup: the server could not find the requested resource (get globalrulegroups.alerting.kubesphere.io)
W0421 23:50:27.964632   16878 reflector.go:424] pkg/client/informers/externalversions/factory.go:129: failed to list *v2beta1.ClusterRuleGroup: the server could not find the requested resource (get clusterrulegroups.alerting.kubesphere.io)
E0421 23:50:27.964663   16878 reflector.go:140] pkg/client/informers/externalversions/factory.go:129: Failed to watch *v2beta1.ClusterRuleGroup: failed to list *v2beta1.ClusterRuleGroup: the server could not find the requested resource (get clusterrulegroups.alerting.kubesphere.io)
W0421 23:50:31.658739   16878 reflector.go:424] pkg/client/informers/externalversions/factory.go:129: failed to list *v1alpha1.Webhook: the server could not find the requested resource (get webhooks.auditing.kubesphere.io)
E0421 23:50:31.658768   16878 reflector.go:140] pkg/client/informers/externalversions/factory.go:129: Failed to watch *v1alpha1.Webhook: failed to list *v1alpha1.Webhook: the server could not find the requested resource (get webhooks.auditing.kubesphere.io)

这一段排查了很久，因为集群里的ks-apiserver:v3.3.2运行正常，为什么本地会报错？

结论：因为官方QuickStart里给的kubesphere.yaml打开了一些allinone没有安装的功能

解决办法如下

其它注意事项

由于我们k8s只有一个master节点，所以要把master节点的taint给删掉，否则ks-installer等组件都会一直pending！