Skip to content

Coder Environment Documentation

Warning

If your workspace was created before 11/01/2024, please do not update your workspace version before backing up your data! Otherwise, your data might be lost.

TL;DR

  1. Coder provides a quick-start development environment similar to JupyterHub, running directly on our cluster. No Kubernetes knowledge required!
  2. FPGA Requests: Easily request and use Xilinx Alveo U55C FPGAs for specialized workloads.
  3. Preconfigured with Xilinx Vivado/Vitis License Server, including P4 license support, so you can compile and deploy a variety of FPGA workloads right out of the box.
  4. Git Integration: All workspaces come with SSH keys for seamless GitLab connections.
  5. Workspace Timeout: For large images (e.g., PyTorch), if your workspace times out after 5 minutes, wait an additional 5 minutes before restarting. If issues persist, contact Nautilus Support.
  6. Important: Deleting any workspace also deletes its associated Persistent Volume Claim (PVC), so be sure to backup any critical data. ⚠️

Overview

We offer a Coder environment hosted at coder.nrp-nautilus.io on our cluster, providing an easy-to-use, JupyterHub-like experience without needing Kubernetes expertise. You can create an account, sign in with your institutional credentials via OpenID Connect, and start using Coder immediately. Users are allowed up to 5 active workspaces and can easily request Xilinx Alveo U55C FPGAs for enhanced computing power.

Each workspace includes a persistent home folder, initially limited to 5GB, with the option to request additional storage. Workspaces are launched from a selection of templates tailored to different development needs.

Coder CLI

For easier management, you can install the Coder CLI on your local machine here. The CLI enables you to manage your account, handle workspaces, and SSH into any workspace directly from your terminal.

Workspace Templates

1. General Template

This template provides a setup similar to JupyterHub, with configuration options for: - Region - CPU Cores - Memory - GPU Type - FPGA Requests (highlighted options in bold)

Each image includes features like noVNC Desktop GUI, JupyterLab, an in-browser Terminal 💻, VSCode integration, and Cursor integration.

2. Selkies

Preloaded with Selkies-GStreamer Streaming Service for fast, GPU-accelerated remote desktop access. This setup includes VSCode Integration, PyCharm, Wine for Windows applications, and other useful tools.

3. CUDA/PyTorch/TensorFlow

Preloaded with an extensive selection of machine learning libraries, plus GPU support for ML/DL tasks.

4. U55C FPGA Vitis Workflow

Ideal for FPGA development, preloaded with multiple versions of Vivado and Vitis, along with access to the Xilinx License Server (includes VitisNetP4 license for P4 workloads). Allows for one or multiple FPGA requests.

5. ESnet FPGA SmartNICs

Includes all tools from the U55C FPGA Vitis Workflow, plus ESnet SmartNIC tools for configuring Xilinx Alveo U55C FPGAs as P4-programmable SmartNICs with 100Gbps per port ⚡.

6. Other Environments

Additional images are available for various development needs, such as Node.js, RStudio, and Golang.

Workspace Management

After setting up a workspace, you can always adjust configurations later through the settings menu.

Important: Deleting any workspace also deletes its associated Persistent Volume Claim (PVC), so be sure to backup any critical data. ⚠️

Timeout Note for Large Images

If a large image (e.g., PyTorch) times out after 5 minutes, it may still be starting in the background. Wait an additional 5 minutes and try restarting. If issues continue, reach out to Nautilus Support.

Git Integration

All workspaces come preconfigured with an SSH key, available in your user settings, simplifying connections to GitLab. This makes backing up code a straightforward task.

For further assistance, feel free to contact the Nautilus Support team.