K8S+Helm 安装 Jupyterhub

前提条件：准备好的k8s环境和helm环境

准备配置文件

生成随机数

openssl rand -hex 32
# 8fc2826e9ce6930ec26c9fd541c0620b448a947357edbdf9647516af16bbc798

配置config.yaml如下，token是上一步生成

proxy:
  secretToken: "8fc2826e9ce6930ec26c9fd541c0620b448a947357edbdf9647516af16bbc798"

安装JupyterHub

1. 安装helm部署仓库,注意需要全局代理

helm repo add jupyterhub https://jupyterhub.github.io/helm-chart/
helm repo update

安装完成后会有如下显示

Hang tight while we grab the latest from your chart repositories...
...Successfully got an update from the "sqlfiddle" chart repository
...Successfully got an update from the "jupyterhub" chart repository
...Successfully got an update from the "emberstack" chart repository
Update Complete. ⎈ Happy Helming!⎈

2. 接下来使用helm安装JupyterHub，并使用我们刚刚配置的config.yaml

创建helm-install-hupyterhub.sh文件，并复制如下进去：

# Suggested values: advanced users of Kubernetes and Helm should feel
# free to use different values.
RELEASE=jhub
NAMESPACE=jhub

helm upgrade --install $RELEASE jupyterhub/jupyterhub \
  --namespace $NAMESPACE  \
  --version=0.9.0 \
  --values config.yaml

创建k8s的jhub namespace

kubectl create ns jhub

注意：遇到问题：Release “jupyterhub” does not exist. Installing it now.，

# 在master节点执行
kubectl get pod --namespace jhub
# 查看pod节点异常情况
kubectl describe po hook-image-puller-h9mwm --namespace jhub

错误原因是 镜像无法拉取

打开代理环境，获取镜像

docker pull gcr.io/google_containers/kube-scheduler-amd64:v1.13.12
docker pull gcr.io/google_containers/pause:3.1

当部署成功时

kubectl -n jhub get po
# continuous-image-puller-x5fwx     1/1     Running   0          4m45s
# hub-75d769476f-tdz8l              1/1     Running   0          4m45s
# proxy-d769dd5db-c7pqf             1/1     Running   0          4m45s
# user-scheduler-64b57d78d5-8dv7s   1/1     Running   0          4m45s
# user-scheduler-64b57d78d5-zk2cb   1/1     Running   0          4m45s
kubectl get service --namespace jhub
# hub            ClusterIP      10.99.233.188            8081/TCP                     6m12s
# proxy-api      ClusterIP      10.100.28.140            8001/TCP                     6m12s
# proxy-public   LoadBalancer   10.104.115.249   localhost     443:32556/TCP,80:30533/TCP   6m12s

修改proxy-public svc 将 type: LoadBalancer 改成 type: NodePort

kubectl edit service proxy-public -n jhub

查看service

kubectl get service --namespace jhub
# hub            ClusterIP   10.99.233.188            8081/TCP                     16m
# proxy-api      ClusterIP   10.100.28.140            8001/TCP                     16m
# proxy-public   NodePort    10.104.115.249           443:32556/TCP,80:30533/TCP   16m

此时我们在浏览器中通过 80 映射的端口去访问jupyterhub

jupyterhub的默认用户：bi:bi

测试运行