Cloud Run GPU codelab fails due to insufficient regional memory quota

[Yashfeen11]

Issue

I am following the Google Cloud Run ADK + BigQuery + MCP codelab and reached the step where Gemma 4 is deployed to Cloud Run with an NVIDIA RTX PRO 6000 GPU.

The deployment fails before the Cloud Run service is created.

Environment

• Project: code-lab-2-cohort3
• Model: google/gemma-4-31B-it
• GPU: nvidia-rtx-pro-6000
• CPU: 20
• Memory: 80Gi
• Max instances: 1
• Region: us-central1

Deployment command/configuration

The codelab specifies:

export CLOUD_RUN_CPU_NUM=20
export CLOUD_RUN_MEMORY_GB=80
export CLOUD_RUN_MAX_INSTANCES=1
export CLOUD_RUN_CONCURRENCY=16

Error

Cloud Run returns:

Deployment failed

ERROR: (gcloud.beta.run.deploy)
spec.template.metadata.annotations[autoscaling.knative.dev/maxScale]:
Max instances must be set to 0 or fewer to set the requested total memory.

Consider running your workload in a region with greater capacity,
or requesting an increase in quota for this region.

Quota violated:
MemAllocPerProjectRegion requested: 85899345920
allowed: 42949672960

The requested memory is 80 GiB, while the project has only approximately 40 GiB of MemAllocPerProjectRegion quota.

I initially tried asia-southeast1, which had the same 40 GiB limit. I then tried us-central1, where the same quota error occurs.

I also checked the Cloud Console quota page. The MemAllocPerProjectRegion quota is adjustable, but the console says:

"Based on your service usage history, you are not eligible for a quota increase at this time."

Expected behavior

The codelab should allow the provided lab project to deploy the documented RTX PRO 6000 configuration requiring 80 GiB of memory, or provide guidance for what to do when the lab project's quota is only 40 GiB.

[vladkol]

@Yashfeen11, your Cloud project must have a lower quota that required by Cloud Run's RTX 6000 machines.