Skip to main content

Configure Slurm partitions

Add, edit, and delete partitions on an existing Slurm cluster from the cluster detail page.

Configure Slurm partitions

Add, edit, and delete partitions on an existing Slurm cluster.

TimeAbout 5 minutes per operation
You will needA Slurm cluster in ready status
OutcomePartitions configured for your job queues

This guide walks through adding, editing, and deleting partitions on an existing Slurm cluster. Partitions are job queues that map to a pool of compute nodes with a specific instance type and autoscaling range.

Add a partition

Open the cluster list

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

Open the cluster detail page

Click the cluster name to open the detail page.

Open the Partitions tab

Click the Partitions tab.

Add a partition

Click Add Partition.

Configure the partition

Fill in the partition configuration:

  • Partition Name: a short identifier (for example, gpu or high-mem)
  • Compute Node: click Select Compute Node to browse instance types by vCPU, GPU, and price (AWS) or select a profile (non-AWS)
  • Maximum node count: the upper bound for autoscaling

Save

Click Add Partition to save.

Success looks like this: the partition is created or updated, and sinfo shows the new partition available for job submissions.

The new partition appears in the Partitions tab. Nodes are provisioned on demand when jobs target this partition.

Edit a partition

Open the edit form

On the Partitions tab, click the edit icon on the partition row.

Modify settings

Change the instance type or maximum node count as needed.

Save

Click Save.

tip

Start with a low maximum node count and increase as you understand your workload patterns. Each additional node incurs compute costs when jobs are queued.

Set the default partition

Toggle the Default flag on a partition row. Jobs submitted without an explicit --partition flag route to the default partition. Only one partition can be default at a time.

Delete a partition

Delete

Click the delete icon on the partition row.

Confirm

Confirm the deletion.

warning

Deleting a partition cancels any jobs currently queued or running on that partition. Drain the partition first if you have active workloads.

Autoscaling behaviour

Slurm autoscales each partition independently. When jobs queue, nodes are provisioned up to the maximum count. When the queue drains, idle nodes are terminated after a cooldown period.

For partition field definitions and Slurm configuration details, see the Partitions reference.

Ask AI
Ask a question about Vantage Compute...