Skip to main content

​
Summary

This page walks you through creating and using an nscale Slurm cluster from the nscale Console UI—from provisioning a cluster and configuring partitions and storage, to accessing the login environment and running jobs with standard Slurm commands. Use this guide if you:
  • Need a managed Slurm environment without installing or operating Slurm yourself
  • Want to provision compute via the Console UI and then submit workloads using familiar Slurm CLI commands (sbatch, srun, squeue, sacct)

​
Availability

This feature is currently only available for reserved cloud services.

​
Requirements

  • SSH access to the login environment requires a registered SSH public key in the Console
  • Note: Slurm CLI tools are preinstalled in the login environment, so you don’t need to install sbatch/srun locally to submit jobs (you’ll run them after connecting)

​
Step-by-Step

  1. Open the Slurm cluster creation flow
    • In the Console left navigation, go to Services → Slurm Cluster → Create New Cluster
  2. Create a cluster
    • Enter a Cluster Name
    • Select the Project the cluster belongs to
    • Choose the Region where the cluster will be deployed
    • Click Next: Partitions
Slurm Workflow 01 1
  1. Configure partitions
    • Set a Partition Name
    • Select a Node Type (instance flavor) from the available options
    • Specify the Node Count (number of nodes in that partition)
    • Add additional partitions if needed
    • Click Next: Storage
  2. Configure storage
    • Choose the Storage Type (currently VAST is supported)
    • Set the Storage Size in GB
    • Optional: set a Mount Path (default is /mnt/storage)
    • Add additional storage resources if required
    • Click Next: SSH Key
Slurm Workflow 02
  1. Select an SSH key
    • Choose the SSH public key you uploaded in the Console to enable access to the cluster’s login environment
  2. Configure additional options
    • as an option you can enable dedicated infrastructure nodes allowing a cluster creation with a minimum of 1 node instead of 3
    • choose if you want to have the Nscale HPC stack preinstalled (optional)
  3. Review your cluster, create cluster and wait for provisioning
Slurm Workflow 03 1
  1. Access the login environment
    • From the cluster details in the Console, use the provided SSH access details to connect to the login environment
    • You land in a shell with Slurm CLI tools available (sbatch, srun, squeue, sacct)
  2. Submit and monitor jobs (from the login environment)
    • Use standard Slurm commands to run and track jobs.
    • Jobs appear in the queue, start running when resources are available, and complete successfully.

​
Tips, Best Practices, and Code Examples

  • Use Slurm CLI from the login environment
    • The intended workflow is: provision via Console → connect to login environment → submit jobs via Slurm CLI.
  • Code examples (from the login environment)
Basic connectivity and scheduling sanity checks: 
# Run a simple command through Slurm
srun hostname
Monitor the queue: 
# View queued/running jobs
squeue
Verify GPU visibility (example output format depends on your cluster): 
# Example: show nodes and GPU resources
sinfo -o "%N %G"
View accounting details for a completed job: 
# Replace JOBID with your job ID
sacct -j JOBID