Ray: How to launch a Ray Cluster on Hawk?

This guide shows you how to launch a Ray cluster on HLRS' Hawk system.

Table of Contents

• Ray: How to launch a Ray Cluster on Hawk?
  • Table of Contents
  • Prerequisites
  • Getting Started
  • Usage
  • Notes

Prerequisites

Before running the application, make sure you have the following prerequisites installed in a conda environment:

• Python 3.9: This specific python version is used for all uses, you can select it using while creating the conda environment. For more information on, look at the documentation for Conda on HLRS HPC systems.
• Conda Installation: Ensure that Conda is installed on your local system. For more information, look at the documentation for Conda on HLRS HPC systems.
• Ray: You can install Ray inside
• Conda Pack: Conda pack is used to package the Conda environment into a single tarball. This is used to transfer the environment to Vulcan.

Getting Started

1. Clone this repository to your local machine:

   git clone <repository_url>
   

2. Go into the directory and create an environment using Conda and environment.yaml. Note: Be sure to add the necessary packages in environment.yaml:

   ./deployment_scripts/create-env.sh <your-env>
   

3. Send all files using deploy-env.sh:

   ./deployment_scripts/deploy-env.sh <your-env> <destination_host>:<destination_directory>
   

4. Send all the code to the appropriate directory on Vulcan using scp:

   scp <your_script>.py <destination_host>:<destination_directory>
   

5. SSH into Vulcan and start a job interatively using:

   qsub -I -N DaskJob -l select=4:node_type=clx-21 -l walltime=02:00:00
   

6. Go into the directory with all code:

   cd <destination_directory>
   

7. Initialize the Dask cluster:

   source deploy-dask.sh "$(pwd)"
   

   Note: At the moment, the deployment is verbose, and there is no implementation to silence the logs. Note: Make sure all permissions are set using chmod +x for all scripts.

Usage

To run the application interactively, execute the following command after all the cluster's nodes are up and running:

python

Or to run a full script:

python <your-script>.py

Note: If you don't see your environment in the python interpretor, then manually activate it using:

conda activate <your-env>

Do this before using the python interpretor.

Notes

Note: Dask Cluster is set to verbose, add the following to your code while connecting to the Dask cluster:

client = Client(..., silence_logs='error')

Note: Replace all filenames within <> with the actual values applicable to your project.