Implementation:Kubeflow Kubeflow Notebook CRD Creation

Knowledge Sources	Kubeflow Notebooks README
Domains	MLOps, Experimentation, Kubernetes
Last Updated	2026-02-13 00:00 GMT

Overview

Concrete tool for provisioning interactive notebook environments provided by the Kubeflow Notebooks component.

Description

The Notebook CRD (Custom Resource Definition) is the Kubernetes-native mechanism through which Kubeflow manages interactive computing environments. When a user creates a Notebook resource, the Notebook Controller provisions a StatefulSet running the selected IDE image (Jupyter, RStudio, or VS Code Server) within the user's namespace. The resulting pod has access to namespace-scoped resources including persistent volumes, secrets, and GPU allocations.

Notebook CRD creation can be performed either through the Kubeflow Central Dashboard UI (which provides a guided form) or directly via kubectl apply for programmatic or GitOps-driven workflows. The CRD encapsulates the full pod specification including container image, resource requests and limits, volume mounts, and environment variables.

External Reference

Usage

Use Notebook CRD creation when:

A data scientist needs a managed Jupyter, RStudio, or VS Code environment on the Kubeflow platform.
GPU-accelerated interactive work is required and must be scheduled through Kubernetes.
The team requires namespace-isolated environments with persistent storage for ongoing experimentation.
Environment provisioning must be automated or version-controlled as part of infrastructure-as-code practices.

Code Reference

Source Location

Repository: kubeflow/notebooks
File: config/crd/bases/kubeflow.org_notebooks.yaml (CRD schema)

Signature

apiVersion: kubeflow.org/v1
kind: Notebook
metadata:
  name: <notebook-name>
  namespace: <user-namespace>
spec:
  template:
    spec:
      containers:
        - name: <notebook-name>
          image: <jupyter|rstudio|codeserver-image>
          resources:
            requests:
              cpu: "<cpu-request>"
              memory: "<memory-request>"
              nvidia.com/gpu: "<gpu-count>"
            limits:
              cpu: "<cpu-limit>"
              memory: "<memory-limit>"
              nvidia.com/gpu: "<gpu-count>"
          volumeMounts:
            - name: workspace
              mountPath: /home/jovyan
      volumes:
        - name: workspace
          persistentVolumeClaim:
            claimName: <pvc-name>

Import

# Apply Notebook CRD via kubectl
kubectl apply -f notebook.yaml

# Or use the Kubeflow Central Dashboard UI to create a Notebook Server
# Navigate to: Central Dashboard > Notebooks > New Notebook

I/O Contract

Inputs

Name	Type	Required	Description
metadata.name	string	Yes	Name of the Notebook resource
metadata.namespace	string	Yes	Kubernetes namespace (typically the user Profile namespace)
spec.template.spec.containers[].image	string	Yes	Container image for the IDE (e.g., Jupyter, RStudio, Code Server)
spec.template.spec.containers[].resources	object	Yes	CPU, memory, and GPU resource requests and limits
spec.template.spec.volumes	list	No	Persistent volume claims for workspace and data storage
spec.template.spec.containers[].env	list	No	Environment variables for the notebook container

Outputs

Name	Type	Description
Running Notebook Pod	Kubernetes Pod	A StatefulSet-managed pod running the selected IDE
Notebook URL	string	Accessible URL routed through Istio gateway for the notebook UI
Persistent Workspace	PVC-backed storage	Durable workspace that survives pod restarts

Usage Examples

Basic Usage

apiVersion: kubeflow.org/v1
kind: Notebook
metadata:
  name: my-jupyter-notebook
  namespace: my-namespace
spec:
  template:
    spec:
      containers:
        - name: my-jupyter-notebook
          image: kubeflownotebookswg/jupyter-scipy:v1.9.0
          resources:
            requests:
              cpu: "1"
              memory: "2Gi"
            limits:
              cpu: "2"
              memory: "4Gi"
          volumeMounts:
            - name: workspace
              mountPath: /home/jovyan
      volumes:
        - name: workspace
          persistentVolumeClaim:
            claimName: my-notebook-pvc

GPU-Enabled Notebook

apiVersion: kubeflow.org/v1
kind: Notebook
metadata:
  name: gpu-notebook
  namespace: ml-team
spec:
  template:
    spec:
      containers:
        - name: gpu-notebook
          image: kubeflownotebookswg/jupyter-pytorch-cuda:v1.9.0
          resources:
            requests:
              cpu: "4"
              memory: "16Gi"
              nvidia.com/gpu: "1"
            limits:
              cpu: "8"
              memory: "32Gi"
              nvidia.com/gpu: "1"
          volumeMounts:
            - name: workspace
              mountPath: /home/jovyan
            - name: datasets
              mountPath: /data
      volumes:
        - name: workspace
          persistentVolumeClaim:
            claimName: gpu-notebook-workspace
        - name: datasets
          persistentVolumeClaim:
            claimName: shared-datasets

Related Pages

Implements Principle

Principle:Kubeflow_Kubeflow_Experiment_And_Prototype

Requires Environment

Environment:Kubeflow_Kubeflow_Kubernetes_Cluster_Environment

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment