Replies: 2 comments 1 reply
-
Right now the windows-startup-script module is still marked experimental and is focused on the use case of supporting Windows execute points in HTCondor pools. It doesn't have the same level of industrialization as the Linux startup-script solution. In particular, the Linux solution will create the scripts as Cloud Storage objects and download/execute them iteratively. The Windows solution is just a long string to supply in metadata. This example, however, should build the NVIDIA GPU drivers on a Windows VM: blueprint_name: example-win
vars:
project_id: your-project
deployment_name: example-win
region: us-central1
zone: us-central1-a
deployment_groups:
- group: primary
modules:
- id: network
source: modules/network/vpc
settings:
enable_iap_rdp_ingress: true
enable_iap_winrm_ingress: true
- id: win_startup
source: community/modules/scripts/windows-startup-script
settings:
install_nvidia_driver: true
- id: vm
source: modules/compute/vm-instance
use:
- network
settings:
machine_type: n1-standard-16
guest_accelerator:
- type: nvidia-tesla-t4
count: 1
instance_image:
family: windows-2016
project: windows-cloud
metadata:
windows-startup-script-ps1: $(join("", win_startup.windows_startup_ps1)) I recommend the Cloud docs for further details of connecting to Windows VMs. |
Beta Was this translation helpful? Give feedback.
-
The serial port logs available in the Cloud Console for each VM can also be a good simple way of monitoring progress on windows VMs. An example line you should see:
|
Beta Was this translation helpful? Give feedback.
-
The windows-startup-script module's README doesn't describe how to use the output (windows_startup_ps1) to specify startup scripts. Specifically, I am not sure if the list should specify paths to local scripts that will then be included in the image, or if they can be URLs, or if both are supported. A sample would be very helpful.
Beta Was this translation helpful? Give feedback.
All reactions