Skip to content
Klutch.sh Docs

How to Deploy JupyterLab

Introduction

JupyterLab is a next-generation web-based user interface for Project Jupyter. Deploying JupyterLab on Klutch.sh provides scalable, secure infrastructure for interactive computing, with support for persistent storage and automated CI/CD.

This guide covers deploying JupyterLab on Klutch.sh using a Dockerfile, configuring persistent storage, and best practices for production deployments.

Prerequisites

  • A Klutch.sh account (sign up here)
  • A GitHub repository for your JupyterLab deployment (or fork of the JupyterLab repo)
  • Basic knowledge of Docker and data science concepts

1. Prepare your JupyterLab repository

  • Fork or clone the JupyterLab repository, or create a wrapper repo for your customizations.
  • Store large assets (such as notebooks, data, or configuration files) outside the Git repo; use persistent volumes or object storage and mount or fetch them at runtime.

Refer to the Klutch.sh Quick Start Guide for repository setup and GitHub integration.

2. Sample non-Docker deployment (Klutch.sh build)

You can deploy JupyterLab from your repo without a Dockerfile using Klutch.sh’s build system:

  1. Push your repo to GitHub. Include a start script (for example: start.sh) that installs dependencies and runs JupyterLab.
  2. In Klutch.sh, create a new project and app, and connect your repository.
  3. Set the start command to the script or JupyterLab’s start command (example: jupyter lab --ip=0.0.0.0 --port=8888 --no-browser).
  4. Attach persistent volumes for notebooks, data, or configuration files (see Volumes Guide).
  5. Set the app port to 8888 (or your configured port).
  6. Click “Create” to build and deploy.

Notes:

  • Configure runtime secrets (API keys, tokens) as environment variables in Klutch.sh.
  • For advanced use, customize the start script to load assets from mounted volumes or object storage.

A Dockerfile ensures reproducible builds and full control over dependencies. Example:

FROM jupyter/base-notebook:latest


# Optional: Add custom configuration or extensions
# COPY ./notebooks /home/jovyan/work


EXPOSE 8888

For custom assets, mount persistent volumes or fetch from object storage at startup.

4. Persistent storage & volumes

JupyterLab requires persistent storage for notebooks, data, and configuration files:

  • Create a persistent volume in Klutch.sh and mount it to /home/jovyan/work or your chosen path.
  • Configure JupyterLab (or your startup script) to read/write from the mounted path.

Example mount mapping in Klutch.sh app settings:

/home/jovyan/work  <- my-jupyterlab-storage

If using object storage (S3-compatible), store credentials in environment variables and fetch assets at startup.

5. Environment variables and secrets

  • Store API keys, tokens, and other secrets in Klutch.sh environment variables (never in the repo).
  • Use the Klutch.sh UI to mark secrets and prevent them from being logged.

6. Scaling, monitoring, and best practices

  • Use health checks and readiness probes if supported by JupyterLab.
  • Monitor CPU, memory, and latency; scale instances as needed.
  • Pin dependency versions and use multi-stage Docker builds for smaller images.
  • Use CI to build and publish images, or let Klutch.sh build from your repo and tag releases.
  • Restrict public access to endpoints; require authentication or place behind an API gateway.

Resources

Deploying JupyterLab on Klutch.sh gives you a reproducible, scalable path to serve interactive computing environments. For advanced setups, you can add a startup script to fetch assets from S3, a multi-stage Dockerfile for smaller images, or CI/CD integration to automate builds and deployments.