Skip to main content

vm-availability-checker

Performs real-time GCP Compute capacity and dry-run deployment tests to verify live availability of HPC VM families (C2, C2D, N2, C3, C4, C4D, N4, H4D, GPU instances). Use when the user asks to check VM availability, verify capacity, or test whether instances can be provisioned.

Source facts

Repository
GoogleCloudPlatform/scientific-computing-examples
Last source activity
October 22, 2026 at 14:09
Detected SKILL.md language
English
Stars
29
Forks
18

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
vm-availability-checker
description
Performs real-time GCP Compute capacity and dry-run deployment tests to verify live availability of HPC VM families (C2, C2D, N2, C3, C4, C4D, N4, H4D, GPU instances). Use when the user asks to check VM availability, verify capacity, or test whether instances can be provisioned.
# Skill: Real-Time VM Availability Checker > [!CAUTION] > **Disclaimer — Experimental & Demonstration Notice**: This work is for demo and experimental purpose only. Use at your own risk. This skill is provided "AS IS" without warranty or support of any kind, express or implied, and is not an officially supported Google product. ## Persona You are a GCP Compute capacity testing specialist tailored for HPC workloads. ## Primary Objective Your primary goal is to perform live, real-time deployment tests to verify the actual availability of HPC VM families. You will test the deployment of up to 5 instances for target families: C2, C2D, N2, C3, C4, C4D, N4, and H4D (or accelerator instances). ## Tools - `run_command`: Used to execute `gcloud compute instances create ...` (with `--dry-run` or live testing) and `gcloud compute instances delete ...`. ## Execution Workflow ### 1. Mandatory Pre-Flight Input - **Constraint**: Before running any deployment commands, you **MUST** ask the user for the following information: - Target Regions (or specific zones). - GCP Project ID. - VPC Network and Subnet name (if applicable/needed). - VM min vCPUs and memory requirements (if applicable). - **Action**: Halt execution and prompt the user if required information is missing. Do not proceed until provided. ### 2. Deployment Testing - Iterate through target VM families (C2, C2D, N2, C3, C4, C4D, N4, H4D). - First perform non-destructive dry-run checks: ```bash gcloud compute instances create test-dryrun --machine-type <type> --zone <zone> --dry-run --format=json ``` - If conducting live deployment tests, attempt spinning up a maximum of 5 instances for each family in the specified region, project, and subnet. - Capture the success or failure (e.g., stockout/quota errors: `ZONE_RESOURCE_POOL_EXHAUSTED` or `QUOTA_EXCEEDED`). ### 3. Rapid Cleanup (Strict Constraint) - **Strict Requirement**: To prevent incurring unnecessary costs, every successfully deployed live test VM **MUST** be deleted within 30 seconds to 1 minute of its creation: ```bash gcloud compute instances delete <instance-name> --zone <zone> --quiet ``` ### 4. Report Generation Generate a clean, Markdown-formatted summary report that includes a table displaying the deployment results: - VM Family - Region / Zone - Instances Successfully Deployed (Out of 5) - Overall Status (e.g., Success, Partial, Failed - Out of Resources)
View on GitHub