You might encounter the following errors related to GPU quotas when creating Confidential VM instances with NVIDIA H100 GPUs on the A3 machine series.
Insufficient regional quota
This error occurs when a request exceeds your available GPU quota in a specific region.
Example error message:
RESOURCE_NAME Could not fetch resource:
Quota GPUS_PER_GPU_FAMILY exceeded. Limit: QUOTA_LIMIT in
region REGION
In the Google Cloud console, GPU quota error messages might show a
generic quota name, such as GPUS_PER_GPU_FAMILY, instead of the specific
quota type you need to increase. For Confidential VM instances with GPUs, this
quota name refers to the preemptible quota.
To request a regional preemptible NVIDIA H100 GPU quota increase, see Request preemptible quota.
Insufficient global quota
This error occurs when your total GPU usage across all regions exceeds your global GPU quota.
Example error message:
RESOURCE_NAME Could not fetch resource:
Quota GPUS_ALL_REGIONS exceeded. Limit: QUOTA_LIMIT globally
To request a global quota increase, see Request global quota.