Nautilus

This repo maintains useful resources for Nautilus usage. The outline is as follows.

• Nautilus
  • Getting started
    • Setup
    • Configuration
    • Workflow
  • Datasets
  • Useful resources

Getting started

Please carefully read the policy before using the cluster. It is also recommended to read through the entire document to get started. Another useful resource to introduce Nautilus usage is here.

Setup

• Log in nautilus portal. Please select "University of California, San Diego" as the identity provider and log in with your UCSD account.
• Join the the official support channel Matrix follow this intruction. Inside Matrix, find and join CogRob - ERL - RY Group room using the option "Explore public rooms".
• Please contact our group admins (Luobin Wang) and provide your UCSD email account to join our namespace (cogrob).
• Follow this instruction to setup the kubernetes client.

Configuration

Nautilus provides several examples here. You might also start from pod-example.yaml or job-example.yaml.

Workflow

• Claim your own/shared persistent storage.
• Find a good Docker image for your project. Build your docker image if necessary. REQUIRED: install packages and dependencies in your images instead of PVC.
  • You can follow this guide to get familiar with docker.
  • DockerHub and Nautilus both allow you to upload your images.
• Try training commands using a pod/deployment.
• Create and run your job.
• Monitor and set correct resource usage (required)
• Retrieve your result in persistent storage.
• Delete your deployment/job/pod (required).

Datasets

Commonly used datasets are maintained by the namespace admin. Please check this Google Sheet to see whether a dataset has already been downloaded.

• Please do not directly store processed data into the shared PVC (cogrob-avl-west-vol). Store them in your own PVC instead.
• Nautilus is not a place for storing large-scale data. Please backup your data to your local storage and purge unused data on Nautilus after finishing your project.

Useful resources

• Portal: https://portal.nrp-nautilus.io/
• Kubernetes tutorial: https://learnk8s.io/blog/kubectl-productivity/
• Kubernetes documentation: https://kubernetes.io/docs/home/
• Grafana monitoring: https://grafana.nrp-nautilus.io/dashboards/f/fef6b8ca-9d75-439c-8591-6825b1c507d1/default
• Docker hub: https://hub.docker.com/