Adding a VM to a GPU cluster

Written by

Updated at November 8, 2024

In GPU clusters, you can only create VMs running on the AMD EPYC™ with NVIDIA® Ampere® A100 platform with 8 GPUs. You will need to prepare a disk image with drivers by following this guide and use it when creating your VM.

Note

Currently, GPU clusters are only available in theru-central1-a availability zone. You can only add a VM to a GPU cluster from the same availability zone.

CLI

Terraform

export YC_GPU_CLUSTER=$(yc compute gpu-cluster list --format=json | jq -r .[].id)
export YC_ZONE="ru-central1-a"
export SUBNET_NAME="my-subnet-name"
export SUBNET_ID=$(yc vpc subnet get --name=$SUBNET_NAME --format=json | jq -r .id)
yc compute instance create --name node-gpu-test \
  --create-boot-disk size=64G,image-id=<ID_of_image_with_drivers>,type=network-ssd \
  --ssh-key=$HOME/.ssh/id_rsa.pub \
  --gpus 8 --cores 224 --memory=952G \
  --zone $YC_ZONE \
  --network-interface subnet-id=$SUBNET_ID,nat-ip-version=ipv4 \
  --platform gpu-standard-v3 \
  --gpu-cluster-id=$YC_GPU_CLUSTER

If you don't have Terraform, install it and configure the Yandex Cloud provider.

In the Terraform configuration file, describe the parameters of the resource to create:

provider "yandex" {
  zone = "ru-central1-a"
}

resource "yandex_compute_disk" "boot-disk" {
  name     = "<disk_name>"
  type     = "<disk_type>"
  zone     = "ru-central1-a"
  size     = "<disk_size>"
  image_id = "<ID_of_image_with_drivers>"
}

resource "yandex_compute_instance" "default" {
  name           = "vm-gpu"
  platform_id    = "gpu-standard-v3"
  zone           = "ru-central1-a"
  gpu_cluster_id = "<GPU_cluster_ID>"

  resources {
    cores  = "224"
    memory = "952"
    gpus   = "8"
  }

  boot_disk {
    disk_id = yandex_compute_disk.boot-disk.id
  }

  network_interface {
    subnet_id = "${yandex_vpc_subnet.subnet-1.id}"
    nat       = true
  }

  metadata = {
    user-data = "#cloud-config\nusers:\n  - name: <username>\n    groups: sudo\n    shell: /bin/bash\n    sudo: 'ALL=(ALL) NOPASSWD:ALL'\n    ssh_authorized_keys:\n      - ${file("<path_to_public_SSH_key>")}"
  }
}

resource "yandex_vpc_network" "network-1" {
  name = "network1"
}

resource "yandex_vpc_subnet" "subnet-1" {
  name           = "subnet1"
  zone           = "<availability_zone>"
  v4_cidr_blocks = ["192.168.10.0/24"]
  network_id     = "${yandex_vpc_network.network-1.id}"
}

Where:

yandex_compute_disk: Boot disk description, where image_id is the ID of the image with the drivers.
gpu_cluster_id: GPU cluster ID. This is a required parameter.
yandex_vpc_network: Cloud network description.
yandex_vpc_subnet: Description of the subnet to create your VM in.

Note

If you already have suitable resources, such as a cloud network and subnet, you do not need to describe them again. Use their names and IDs in the appropriate parameters.

For more information about the yandex_compute_instance resource parameters, see the Terraform provider documentation.

Under metadata, specify your username and path to the public SSH key. For more information, see VM metadata.
Create resources:

In the terminal, change to the folder where you edited the configuration file.
Make sure the configuration file is correct using the command:
```
terraform validate
```
If the configuration is correct, the following message is returned:
```
Success! The configuration is valid.
```
Run the command:
```
terraform plan
```
The terminal will display a list of resources with parameters. No changes are made at this step. If the configuration contains errors, Terraform will point them out.
Apply the configuration changes:
```
terraform apply
```
Confirm the changes: type yes in the terminal and press Enter.

This will create a VM in the specified GPU cluster. You can check the new VM and its configuration using the management console or this CLI command:

yc compute instance get <VM_name>

Adding a VM to a GPU cluster

Was the article helpful?