Create an A4X Slurm cluster

This document describes how to deploy a Slurm cluster that uses A4X accelerator-optimized machine types by using Cluster Toolkit.

The A4X machine series runs on an exascale platform based on NVIDIA's rack-scale architecture using NVIDIA GB200 Grace Blackwell Superchips and is optimized for compute and memory-intensive, network-bound ML training and HPC workloads.

To learn more about the A4X machine series, see the A4X Max and A4X machine series section in the Compute Engine documentation.

A4X limitations

The following limitations apply to A4X VMs in your Slurm cluster:

Caution: The Compute Engine Service Level Agreement (SLA) doesn't apply to the A4X machine series.

• You don't receive sustained use discounts or flexible committed use discounts for instances that use this machine type.
• You can only create instances in certain regions and zones.
• You can't use Persistent Disk (regional or zonal). You can only use Google Cloud Hyperdisk.
• This machine type is only available on the NVIDIA Grace platform.
• Machine type changes aren't supported for A4X. To switch to or from this machine type, you must create a new instance.
• You can't run Windows operating systems on this machine type. For a list of supported Linux operating systems, review the supported operating systems for GPU instances.
• For A4X instances, when you use ethtool -S to monitor GPU networking, the physical port counters that end in _phy don't update. This is expected behavior for instances that use the MRDMA Virtual Function (VF) architecture. For more information, see MRDMA functions and network monitoring tools.
• A4X instances don't support the following:
  • Sole-tenancy
  • Spot VMs
  • Flex-start VMs
• You can't attach Hyperdisk ML disks created before February 4, 2026 to A4X machine types.

Before you begin

Before creating a Slurm cluster, if you haven't already done so, complete the following steps:

1. Choose a consumption option: your choice of consumption option determines how you get and use GPU resources.

   To learn more, see Choose a consumption option.

2. Obtain capacity: the process to obtain capacity differs for each consumption option.

   To learn about the process to obtain capacity for your chosen consumption option, see Capacity overview.

   Note: When you request A4X Max or A4X capacity, you obtain it in the all capacity mode. This mode is the only supported reservation operational mode for A4X Max or A4X machine types. For more information about all capacity mode, see Reservation operational mode.
3. Ensure that you have enough Filestore capacity quota: you need to have enough Filestore quota in your target region before deploying. The required minimum capacity depends on the machine types in your cluster:
   • A4X: requires a minimum of 10 TiB (10,240 GiB) of HIGH_SCALE_SSD (zonal) capacity.

   To check quota or request a quota increase, see the following:

   • To check quota, see View API-specific quota.
   • If you don't have enough quota, request a quota increase.
4. Install Cluster Toolkit: to provision Slurm clusters, you must use Cluster Toolkit version v1.62.0 or later.

   To install Cluster Toolkit, see Set up Cluster Toolkit.

In the Google Cloud console, activate Cloud Shell.

Activate Cloud Shell

At the bottom of the Google Cloud console, a Cloud Shell session starts and displays a command-line prompt. Cloud Shell is a shell environment with the Google Cloud CLI already installed and with values already set for your current project. It can take a few seconds for the session to initialize.

Required roles

To complete this tutorial, you need IAM roles granted to your Compute Engine default service account and your user account.

Get required roles for your Compute Engine default service account

To get the permissions that you need to deploy a Slurm cluster, ask your administrator to grant you the following IAM roles:

• Storage Object Viewer (roles/storage.objectViewer) on your project
• Compute Instance Admin (v1) (roles/compute.instanceAdmin.v1) on your project
• Service Account User (roles/iam.serviceAccountUser) on the service account itself

For more information about granting roles, see Manage access to projects, folders, and organizations.

You might also be able to get the required permissions through custom roles or other predefined roles.

Get required roles for your user account

To get the permissions that you need to create a Cloud Storage bucket, ask your administrator to grant you the Storage Admin (roles/storage.admin) IAM role on your project.

Set up a storage bucket

Cluster blueprints use Terraform modules to provision Cloud infrastructure. A best practice when working with Terraform is to store the state remotely in a version enabled file. On Google Cloud, you can create a Cloud Storage bucket that has versioning enabled.

To create this bucket and enable versioning from the CLI, run the following commands:

gcloud storage buckets create gs://BUCKET_NAME \
    --project=PROJECT_ID \
    --default-storage-class=STANDARD --location=BUCKET_REGION \
    --uniform-bucket-level-access
gcloud storage buckets update gs://BUCKET_NAME --versioning

Replace the following:

• BUCKET_NAME: a name for your Cloud Storage bucket that meets the bucket naming requirements.
• PROJECT_ID: your project ID.
• BUCKET_REGION: any available location.

Open the Cluster Toolkit directory

To use Slurm with Google Cloud, you must install Cluster Toolkit. After you install Cluster Toolkit, verify that you are in the Cluster Toolkit directory by running the following command:

cd cluster-toolkit

This cluster deployment requires Cluster Toolkit v1.62.0 or later. To check your version, you can run the following command:

./gcluster --version

Create a deployment file

Create a deployment file that you can use to specify the Cloud Storage bucket, set names for your network and subnetwork, and set deployment variables such as project ID, region, and zone.

To create a deployment file, follow the steps for your required machine type and consumption option.

To create your deployment file, use a text editor to create a YAML file named a4xhigh-slurm-deployment.yaml and add the following content.

Tip: For a more detailed example, copy and paste the contents of examples/machine-learning/a4x-highgpu-4g/a4xhigh-slurm-deployment.yaml into a file in your workspace and edit it.

terraform_backend_defaults:
  type: gcs
  configuration:
    bucket: BUCKET_NAME
vars:
  deployment_name: DEPLOYMENT_NAME
  project_id: PROJECT_ID
  region: REGION
  zone: ZONE
  a4x_cluster_size: NUMBER_OF_VMS
  a4x_reservation_name: RESERVATION_NAME

Replace the following:

• BUCKET_NAME: the name of your Cloud Storage bucket, which you created in the previous section.
• DEPLOYMENT_NAME: a name for your deployment. If creating multiple clusters, ensure that you select a unique name for each one.
• PROJECT_ID: your project ID.
• REGION: the region that has the reserved machines.
• ZONE: the zone where you want to provision the cluster. If you're using a reservation-based consumption option, the region and zone information was provided by your account team when the capacity was delivered.
• NUMBER_OF_VMS: the number of A4X VMs in your cluster. You can specify any number of VMs. However, A4X VMs are physically interconnected by a multi-node NVLink system in groups of 18 VMs (72 GPUs) to form an NVLink domain.

  For optimal network performance, we recommend that you specify a value that is a multiple of 18 VMs (for example, 18, 36, or 54). When you create an A4X cluster, the A4X blueprint automatically creates and applies a compact placement policy with a GPU topology of 1x72 for each group of 18 VMs. For more information about A4X topology, see A4X fundamentals.

• RESERVATION_NAME: the name of your reservation.

Provision a Slurm cluster

Cluster Toolkit provisions the cluster based on the deployment file that you created in the previous step and the default cluster blueprint. For more information about the software that is installed by the blueprint, including NVIDIA drivers and CUDA, learn more about Slurm custom images.

To provision the cluster, use the ./gcluster deploy command for your machine type from the Cluster Toolkit directory. This process can take about 20 to 30 minutes.

./gcluster deploy -d a4xhigh-slurm-deployment.yaml examples/machine-learning/a4x-highgpu-4g/a4xhigh-slurm-blueprint.yaml

Note: A4X uses the ARM64 architecture, so any software installed on these VMs must be compatible with that architecture.

Connect to the Slurm cluster

To access your cluster, you must sign in to the Slurm login node. To sign in, you can use either Google Cloud console or Google Cloud CLI.

Console

1. Go to the Compute Engine > VM instances page.

   Go to the VM instances page

2. Locate the login node. The node has a name with the pattern DEPLOYMENT_NAME +login-001.

3. From the Connect column of the login node, click SSH.

gcloud

To connect to the login node, complete the following steps:

1. Identify the login node by using the gcloud compute instances list command.

   gcloud compute instances list \
     --zones=ZONE \
     --filter="name ~ login" --format "value(name)"
   

   If the output lists multiple Slurm clusters, you can identify your login node by the DEPLOYMENT_NAME that you specified.

2. Use the gcloud compute ssh command to connect to the login node.

   gcloud compute ssh LOGIN_NODE \
     --zone=ZONE --tunnel-through-iap
   

   Replace the following:

   • ZONE: the zone where the VMs for your cluster are located.
   • LOGIN_NODE: the name of the login node, which you identified in the previous step.

Test network performance on the Slurm cluster

We recommend that you validate the functionality of provisioned clusters. To do so, use NCCL tests, which are NVIDIA Collective Communications Library (NCCL) tests that are optimized for the Google environment.

Redeploy the Slurm cluster

If you need to increase the number of compute nodes or add new partitions to your cluster, you might need to update configurations for your Slurm cluster by redeploying. You can speed up a redeployment by using an existing image from a previous deployment. To avoid creating new images during a redeploy, specify the --only flag.

To redeploy the cluster using an existing image do the following:

1. Connect to the cluster. For instructions, see Connect to the Slurm cluster.

2. Use the ./gcluster deploy command for your required machine type:

./gcluster deploy -d a4xhigh-slurm-deployment.yaml examples/machine-learning/a4x-highgpu-4g/a4xhigh-slurm-blueprint.yaml --only cluster-env,cluster -w

This command is only for redeployments where an image already exists, it only redeploys the cluster and its infrastructure.

Destroy the Slurm cluster

By default, the A4X, A4, and A3 Ultra blueprints turn on deletion protection on the Filestore instance. To delete the Filestore instance when you destroy the Slurm cluster, you must turn off deletion protection before you use the destroy command.

For instructions, see set or remove deletion protection on an existing instance.

1. If you are connected to the cluster, then disconnect from it.

2. Before you use the destroy command, navigate to the root of the Cluster Toolkit directory. By default, DEPLOYMENT_FOLDER is located at the root of the Cluster Toolkit directory.

3. To destroy the cluster, use the ./gcluster destroy command:

   ./gcluster destroy DEPLOYMENT_FOLDER

   Replace the following:

   • DEPLOYMENT_FOLDER: the name of the deployment folder. It's typically the same as DEPLOYMENT_NAME.

   When the destruction completes, the output is similar to the following:

     Destroy complete! Resources: xx destroyed.
   

   To learn how to cleanly destroy infrastructure and for advanced manual deployment instructions, see the deployment folder located at the root of the Cluster Toolkit directory: DEPLOYMENT_FOLDER/instructions.txt

What's next

• Verify reservation consumption
• View the topology of a compute instance
• Learn how to manage host events:
  • Manage host events across VMs
  • Manage host events across reservations
• Monitor VMs in your Slurm cluster
• Test and optimize with NCCL/gIB
• Report faulty host