K8S + docker + jupyterhub 计算集群配置

1. Docker,k8s,nvidia-docker2 --> daemon.json

{
    "insecure-registries" : ["10.1.21.39:48080"],
    "registry-mirrors":["https://vds6zmad.mirror.aliyuncs.com"],
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

2. 配置helm

   https://blog.csdn.net/Thomson_tian/article/details/104488986

3. 每台节点登录Harbor

docker login + username

3. 创建一个StorageClass ,StorageClassName = hdu-storage

4. 修改juputerhub/value.xml

   storageClassName = 
   storageClass = 
   
   secretToken = 
    # secretToken = 'openssl rand -hex 32'
   

5. Helm 安装jupyterhub

   jupyterhub 资源文件 10.1.21.30:/home/hdu/hqProject/jupyterhub

   安装

   helm install /home/hdu/hqProject/jupyterhub --version=0.9.0 --name=jhub --namespace=jhub

   删除

   helm del --purge jhub && kubectl delete namespace jhub

   查看安装好的信息

   kubectl --namespace=jhub get svc proxy-public
   
   kubectl --namespace=jhub get pod
   

6. 配置ingress

   • 拷贝30 /home/hdu/hqProject/charts-master/stable/nginx-ingress 这个目录

   • 拷贝/home/hdu/hqProject/nfsconfig/ingress.yaml

   • 执行以下命令

      helm install /home/hdu/hqProject/charts-master/stable/nginx-ingress --namespace=jhub
      
      kubectl create -f   /home/hdu/hqProject/nfsconfig/ingress.yaml 
      # 不要忘记在配置文件中 指定namespace: jhub
     

     查看最后配置情况kubectl --namespace=jhub get svc proxy-public

参考

https://zhuanlan.zhihu.com/p/50407362