kubernetes中的有状态jupyter笔记本

Question

尝试在Kubernetes中部署有状态的jupyter笔记本，但无法保存笔记本中编写的代码，每当笔记本pod关闭时，所有代码都会被删除。我尝试使用持久音量，但无法达到预期效果。

更新

将挂载路径更改为"/home/jovyan“，因为jyputer将ipynb保存在此位置。但现在获得PermissionError: Errno 13权限被拒绝：'/home/jovyan/.local‘，同时部署pod。

kind: Ingress
metadata:
  name: jupyter-ingress
spec:
  backend:
    serviceName: jupyter-notebook-service
    servicePort: 8888

---

kind: Service
apiVersion: v1
metadata:
  name: jupyter-notebook-service
spec:  
  clusterIP: None
  selector:
    app: jupyter-notebook
  ports:
  - protocol: TCP    
    port: 8888
    targetPort: 8888

---

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: jupyter-notebook
  labels:
    app: jupyter-notebook
spec:
  replicas: 1
  serviceName: "jupyter-notebook-service"
  selector:
    matchLabels:
      app: jupyter-notebook
  template:
    metadata:
      labels:
        app: jupyter-notebook
    spec:
      serviceAccountName: dsx-spark
      volumes:
        - name: jupyter-pv-storage
          persistentVolumeClaim:
            claimName: jupyter-pv-claim 
      containers:
      - name: minimal-notebook
        image: jupyter/pyspark-notebook:latest
        ports:
        - containerPort: 8888
        command: ["start-notebook.sh"]
        args: ["--NotebookApp.token=''"]
        volumeMounts:
          - mountPath: "/home/jovyan"
            name: jupyter-pv-storage

---

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: jupyter-pv-claim
spec:
  storageClassName: jupyter-pv-storage
  accessModes:
  - ReadWriteOnce 
  resources:
    requests:
      storage: 1Gi

---
---

apiVersion: v1
kind: PersistentVolume
metadata:
  name: jupyter-pv-volume
  labels:
    type: local
spec:
  storageClassName: jupyter-pv-storage
  capacity:
    storage: 1Gi
  accessModes:
    - ReadWriteOnce 
  hostPath:
    path: "/home/jovyan"

---

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: jupyternotebook-pv-storage
  annotations:
    storageclass.kubernetes.io/is-default-class: "true"
  labels:
provisioner: kubernetes.io/vsphere-volume
parameters:
    diskformat: zeroedthick
    ```

Answer 1

Pod with jupyter notebook是非根用户，因此无法装载，因此我们在创建Pod之前使用initContainers更改Persistent Volume Claim的用户/权限。

kind: StatefulSet
metadata:
  name: jupyter-notebook
  labels:
    app: jupyter-notebook
spec:
  replicas: 1
  serviceName: "jupyter-notebook-service"
  selector:
    matchLabels:
      app: jupyter-notebook
  template:
    metadata:
      labels:
        app: jupyter-notebook
    spec:

      serviceAccountName: dsx-spark
      volumes:
        - name: ci-jupyter-storage-def
          persistentVolumeClaim:
            claimName: my-jupyter-pv-claim 
      containers:
      - name: minimal-notebook
        image: jupyter/pyspark-notebook:latest
        ports:
        - containerPort: 8888
        command: ["start-notebook.sh"]
        args: ["--NotebookApp.token=''"]
        volumeMounts:
          - mountPath: "/home/jovyan/work"
            name: ci-jupyter-storage-def

      initContainers:
      - name: jupyter-data-permission-fix
        image: busybox
        command: ["/bin/chmod","-R","777", "/data"]
        volumeMounts:
        - name: ci-jupyter-storage-def
          mountPath: /data```

Answer 2

正如我在评论中提到的，你需要确保：

1. 给定Pod的存储必须由PersistentVolume资源调配程序根据请求的存储类进行资源调配，或由管理员预先调配。volumeClaimTemplates将使用PersistentVolumes提供stable storage。删除实例或PersistentVolumes时，与实例的PersistentVolume声明关联的StatefulSet不会被删除。
2. 容器是以有权访问该卷的用户身份运行的。可以通过将权限更改为777或您已经注意到的使用适当的initContainers.

来完成