Skip to content

Latest commit

 

History

History
53 lines (35 loc) · 1.83 KB

slurm-cluster.md

File metadata and controls

53 lines (35 loc) · 1.83 KB

Slurm Deployment Guide

Instructions for deploying a GPU cluster with Slurm

Requirements

  • Control system to run the install process
  • One server to act as the Slurm controller/login node
  • One or more servers to act as the Slurm compute nodes

Installation Steps

  1. Install a supported operating system on all nodes.

    Install a supported operating system on all servers via a 3rd-party solution (i.e. MAAS, Foreman) or utilize the provided OS install container.

  2. Set up your provisioning machine.

    This will install Ansible and other software on the provisioning machine which will be used to deploy all other software to the cluster. For more information on Ansible and why we use it, consult the Ansible Guide.

    # Install software prerequisites and copy default configuration
./scripts/setup.sh

  3. Create and edit the Ansible inventory.

    Ansible uses an inventory which outlines the servers in your cluster. The setup script from the previous step will copy an example inventory configuration to the config directory.

    Edit the inventory:

    # Edit inventory
# Add Slurm controller/login host to `slurm-master` group
# Add Slurm worker/compute hosts to the `slurm-node` groups
vi config/inventory

# (optional) Modify `config/group_vars/*.yml` to set configuration parameters

  4. Install Slurm.

    # NOTE: If SSH requires a password, add: `-k`
# NOTE: If sudo on remote machine requires a password, add: `-K`
# NOTE: If SSH user is different than current user, add: `-u ubuntu`
ansible-playbook -l slurm-cluster playbooks/slurm-cluster.yml

Using Slurm

Now that Slurm is installed, try a "Hello World" example using MPI.