slurm-single-node-config

Configuration files for single-node SLURM cluster

STEP 1: System upgrade and setup

Write an executable run_first_script_from_github.sh script like

#!/bin/bash

export NODE_LABEL=15cpu-60ram
export GITHUB_TAG=v0.0.21

curl -s "https://raw.githubusercontent.com/fractal-analytics-platform/slurm-single-node-config/refs/tags/$GITHUB_TAG/first_config_script.sh" -o first_config_script.sh
bash first_config_script.sh

and run it with sudo:

$ sudo ./run_first_script_from_github.sh

STEP 2: GPU drivers

sudo apt install ubuntu-drivers-common
# Confirm
sudo ubuntu-drivers install --gpgpu
# Take note of which version it installs (e.g. 535, 580, ..)

# (reboot)

# Replace 580 with the versions noted above
sudo apt install nvidia-utils-580-server

# Test nvidia-smi
nvidia-smi
# Expected output starts e.g. with
# | NVIDIA-SMI 580.95.05              Driver Version: 580.95.05      CUDA Version: 13.0     |

# Test that apt updates are not broken
sudo apt update
sudo apt upgrade # should not fail

Actual PyTorch-based test:

python3.12 -m venv /tmp/venv
source /tmp/venv/bin/activate
python3.12 -m pip install torch numpy
python3.12 -c 'import torch; assert torch.cuda.is_available(); print(torch.cuda.get_device_name(0)); assert torch.cuda.get_device_name(0) == "Tesla T4"; print("ALL OK!");'
deactivate
rm -r /tmp/venv

STEP 3: Lock the VM

Lock the VM from the openstack dashboard.

STEP 4: SSH keys

Create the appropriate file

sudo -u fractal-worker mkdir /home/fractal-worker/.ssh
sudo -u fractal-worker touch /home/fractal-worker/.ssh/authorized_keys

and then populate it with appropriate keys:

• (required) vm-specific fractal-worker key from the main Fractal instance
• (optional) Relevant team members

STEP 5: Check SLURM

First, run sinfo. If it fails, the likely solution is

sudo systemctl restart slurmctld.service