Skip to main content

Create a Slurm-on-Kubernetes cluster

Deploy a Slurm HPC cluster on top of an existing Kubernetes cluster.

Create a Slurm-on-Kubernetes cluster

This guide walks through deploying a Slurm cluster on top of an existing Kubernetes cluster. Slurm-on-Kubernetes lets you run traditional HPC batch jobs using Kubernetes compute pools as the backing infrastructure.

Deploy a Slurm HPC cluster on top of an existing Kubernetes cluster.

TimeAbout 15 minutes
You will needA Kubernetes cluster in ready status with at least one compute pool
OutcomeA Slurm cluster running on Kubernetes, ready for HPC batch jobs

Create from the Slurm cluster list​

Open Clusters​

Click Clusters in the left sidebar, then click the Slurm tab.

Start the wizard​

Click Prepare Cluster.

Select the deployment target​

In the Configure step, select Kubernetes Cluster as the deployment target. A list of ready Kubernetes clusters appears.

Choose the target cluster​

Click the target cluster, then click Configure Slurm Cluster.

Create from the Kubernetes detail page​

Open the Kubernetes cluster​

Click the target Kubernetes cluster name to open its detail page.

Open the Slurm Clusters tab​

Click the Slurm Clusters tab.

Create a Slurm cluster​

Click Create Slurm Cluster.

Configure compute pools and partitions​

From either entry point, the wizard opens with the Compute & Partitions step.

Compute pools​

Two pools are pre-configured: Slurm Controller (control plane) and Compute Workers.

For each pool, configure:

Select a profile​

Select an instance type:

  • AWS parent cluster: click to open the instance type browser and select any EC2 type (for example, t3.medium or c5n.4xlarge)
  • Non-AWS parent cluster: choose from Small (4 vCPU, 8 GiB), Medium (8 vCPU, 16 GiB), or Large (16 vCPU, 32 GiB)

Configure GPU and scaling​

  • GPU: toggle to enable GPU compute on the pool
  • Max Nodes: the upper autoscaling bound (default: 1 for controller, 10 for compute)

Click Add Compute Pool to add additional worker pools. At least one controller pool and one compute pool are required.

Partitions​

A default partition named partition-1 is pre-configured. For each partition:

Set the partition name​

Set the Partition Name.

Choose the compute group​

Choose which Compute Group it routes to.

Set default status​

Toggle Default status. Only one partition can be default.

Advanced options​

Click Advanced Options to configure TLS, NodePort exposure, job profiling, and the Kubernetes scheduler bridge.

Submit the cluster​

Click Create Slurm Cluster. The wizard shows a three-stage progress stepper:

  1. Registering cluster: creates the cluster record and provisions a Keycloak client
  2. Creating compute pools: provisions each pool on the parent Kubernetes cluster
  3. Creating Slurm cluster: finalizes the Slurm deployment

Wait for all stages to complete. The cluster appears in the Slurm list with ready status.

Success looks like this: the Slurm-on-Kubernetes cluster is provisioned, the controller and workers are running as pods, and job submissions succeed.

Submit a test job​

Create a test script​

Navigate to Jobs > Scripts and create a simple test script.

Submit the script​

Click Actions > Submit Script.

Select the cluster​

In the Cluster dropdown, select the new Slurm-on-Kubernetes cluster.

Choose a partition and submit​

Select a partition and click Submit.

Verify​

Verify the job completes on the Submissions page.

Ask AI
Ask a question about Vantage Compute...