README.md

June 10, 2026 ยท View on GitHub

Description

This module creates a GPU accelerated virtual machine that can be accessed using Chrome Remote Desktop.

Note: This is an experimental module. This module has only been tested in limited capacity with the Cluster Toolkit. The module interface may have undergo breaking changes in the future.

Example

The following example will create a single GPU accelerated remote desktop.

  - id: remote-desktop
    source: community/modules/remote-desktop/chrome-remote-desktop
    use: [network1]
    settings:
      install_nvidia_driver: true

Setting up the Remote Desktop

  1. Once the remote desktop has been deployed, navigate to https://remotedesktop.google.com/headless.
  2. Click through Begin, Next, & Authorize.
  3. Copy the code snippet for Debian Linux.
  4. SSH into the remote desktop machine. It will be listed under VM Instances in the Google Cloud web console.
  5. Run the copied command and follow instructions to set up a PIN.
  6. You should now see your machine listed on the Chrome Remote Desktop page under Remote devices.
  7. Click on your machine and enter PIN if prompted.

License

Copyright 2026 Google LLC

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

 http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Requirements

NameVersion
terraform>= 1.12.2

Providers

No providers.

Modules

NameSourceVersion
client_startup_script../../../../modules/scripts/startup-scriptn/a
instances../../../../modules/compute/vm-instancen/a

Resources

No resources.

Inputs

NameDescriptionTypeDefaultRequired
add_deployment_name_before_prefixIf true, the names of VMs and disks will always be prefixed with deployment_name to enable uniqueness across deployments.
See name_prefix for further details on resource naming behavior.
boolfalseno
auto_delete_boot_diskControls if boot disk should be auto-deleted when instance is deleted.booltrueno
bandwidth_tierTier 1 bandwidth increases the maximum egress bandwidth for VMs.
Using the tier_1_enabled setting will enable both gVNIC and TIER_1 higher bandwidth networking.
Using the gvnic_enabled setting will only enable gVNIC and will not enable TIER_1.
Note that TIER_1 only works with specific machine families & shapes and must be using an image th
at supports gVNIC. See official docs for more details.
string"not_enabled"no
deployment_nameCluster Toolkit deployment name. Cloud resource names will include this value.stringn/ayes
disk_size_gbSize of disk for instances.number200no
disk_typeDisk type for instances.string"pd-balanced"no
enable_osloginEnable or Disable OS Login with "ENABLE" or "DISABLE". Set to "INHERIT" to inherit project OS Login setting.string"ENABLE"no
enable_public_ipsIf set to true, instances will have public IPs on the internet.booltrueno
guest_acceleratorList of the type and count of accelerator cards attached to the instance. Requires virtual workstation accelerator if Nvidia Grid Drivers are required
list(object({
type = string,
count = number
}))
[
{
"count": 1,
"type": "nvidia-tesla-t4-vws"
}
]
no
install_nvidia_driverInstalls the nvidia driver (true/false). For details, see https://cloud.google.com/compute/docs/gpus/install-drivers-gpubooln/ayes
instance_countNumber of instancesnumber1no
instance_imageImage used to build chrome remote desktop node. The default image is
name="debian-12-bookworm-v20250610" and project="debian-cloud".
NOTE: uses fixed version of image to avoid NVIDIA driver compatibility issues.

An alternative image is from name="ubuntu-2204-jammy-v20240126" and project="ubuntu-os-cloud".

Expected Fields:
name: The name of the image. Mutually exclusive with family.
family: The image family to use. Mutually exclusive with name.
project: The project where the image is hosted.
map(string)
{
"name": "debian-12-bookworm-v20250610",
"project": "debian-cloud"
}
no
labelsLabels to add to the instances. Key-value pairs.map(string){}no
machine_typeMachine type to use for the instance creation. Must be N1 family if GPU is used.string"n1-standard-8"no
metadataMetadata, provided as a mapmap(string){}no
name_prefixAn optional name for all VM and disk resources.
If not supplied, deployment_name will be used.
When name_prefix is supplied, and add_deployment_name_before_prefix is set,
then resources are named by "<deployment_name>-<name_prefix>-<#>".
stringnullno
network_interfacesA list of network interfaces. The options match that of the terraform
network_interface block of google_compute_instance. For descriptions of the
subfields or more information see the documentation:
https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/compute_instance#nested_network_interface
_NOTE:_ If network_interfaces are set, network_self_link and
subnetwork_self_link will be ignored, even if they are provided through
the use field. bandwidth_tier and enable_public_ips also do not apply
to network interfaces defined in this variable.
Subfields:
network (string, required if subnetwork is not supplied)
subnetwork (string, required if network is not supplied)
subnetwork_project (string, optional)
network_ip (string, optional)
nic_type (string, optional, choose from ["GVNIC", "VIRTIO_NET", "RDMA", "IRDMA", "MRDMA"])
stack_type (string, optional, choose from ["IPV4_ONLY", "IPV4_IPV6"])
queue_count (number, optional)
access_config (object, optional)
ipv6_access_config (object, optional)
alias_ip_range (list(object), optional)
list(object({
network = string,
subnetwork = string,
subnetwork_project = string,
network_ip = string,
nic_type = string,
stack_type = string,
queue_count = number,
access_config = list(object({
nat_ip = string,
public_ptr_domain_name = string,
network_tier = string
})),
ipv6_access_config = list(object({
public_ptr_domain_name = string,
network_tier = string
})),
alias_ip_range = list(object({
ip_cidr_range = string,
subnetwork_range_name = string
}))
}))
[]no
network_self_linkThe self link of the network to attach the VM.string"default"no
network_storageAn array of network attached storage mounts to be configured.
list(object({
server_ip = string,
remote_mount = string,
local_mount = string,
fs_type = string,
mount_options = string,
client_install_runner = map(string)
mount_runner = map(string)
}))
[]no
on_host_maintenanceDescribes maintenance behavior for the instance. If left blank this will default to MIGRATE except for when placement_policy, spot provisioning, or GPUs require it to be TERMINATEstring"TERMINATE"no
project_idProject in which Google Cloud resources will be createdstringn/ayes
regionDefault region for creating resourcesstringn/ayes
service_accountService account to attach to the instance. See https://www.terraform.io/docs/providers/google/r/compute_instance_template.html#service_account.
object({
email = string,
scopes = set(string)
})
{
"email": null,
"scopes": [
"https://www.googleapis.com/auth/cloud-platform"
]
}
no
spotProvision VMs using discounted Spot pricing, allowing for preemptionboolfalseno
startup_scriptStartup script used on the instancestringnullno
subnetwork_self_linkThe self link of the subnetwork to attach the VM.stringnullno
tagsNetwork tags, provided as a listlist(string)[]no
threads_per_coreSets the number of threads per physical corenumber2no
zoneDefault zone for creating resourcesstringn/ayes

Outputs

NameDescription
instance_nameName of the first instance created, if any.
startup_scriptscript to load and run all runners, as a string value.