Examkin

Question

Difficulty: MediumProvisioning Compute Engine and Kubernetes Engine Clusters

A gaming company is deploying a fleet of GPU-accelerated Compute Engine virtual machines in the us-central1 region using Terraform to host an upcoming real-time gaming event. Pre-deployment testing in a low-scale development environment completed successfully. However, when the automated CI/CD deployment pipeline executes the production Terraform apply job to provision 150 NVIDIA T4 GPUs, the deployment fails instantly before any virtual machines are instantiated. Which action should the Cloud Architect take to resolve this issue?

  1. Submit a request to increase the regional NVIDIA T4 GPU quota in the us-central1 region before re-running the deployment pipeline.Answer
  2. B
    Store the Terraform state file in an unversioned Cloud Storage bucket to eliminate concurrency conflicts during instance creation.
  3. C
    Grant the CI/CD pipeline service account the Service Account Admin role instead of the Service Account User role.
  4. D
    Migrate the workload to a Google Kubernetes Engine (GKE) cluster to automatically bypass regional Compute Engine GPU quota restrictions.

Answer

Submit a request to increase the regional NVIDIA T4 GPU quota in the us-central1 region before re-running the deployment pipeline.
Requesting a regional GPU quota increase in the target region resolves the immediate provisioning failure because Compute Engine GPU allocations are constrained by default project quota limits. Verifying and requesting required quota prior to large-scale infrastructure deployments prevents automated provisioning pipelines from failing.

Step-by-Step Solution

1
Identify the cause of immediate resource provisioning failure during large-scale GPU deployment.
GCP projects enforce default regional quota limits on specialized hardware resources like GPUs, causing API allocation calls to fail immediately if requested capacity exceeds project limits.
Quota limits prevent unexpected resource consumption and require explicit administrative requests before provisioning large compute workloads.
2
Evaluate the proposed resolution steps against GCP operational best practices.
Requesting a regional GPU quota increase via the Cloud Console or Service Usage API ensures sufficient allocation exists before launching the deployment.
Pre-requesting quota increases avoids pipeline failures and ensures infrastructure deployment completes cleanly.

Key Concept

Compute Engine Resource Quota Management
Rate this question

Topics

Provisioning Compute Engine and Kubernetes Engine Clusters

More questions from this topic