Skip to content
Get packs

GKE TPU Stockout Troubleshooter

Diagnoses GKE TPU node pool creation failures caused by reservation stockouts from unhealthy nodes

Source Download skill
Repository
google/skills
Stars
21,782
Licence
Apache-2.0
Files
2 files

What it does

When a Google Kubernetes Engine (GKE) cluster fails to create or scale up a TPU node pool with errors like GCE_STOCKOUT or ZONE_RESOURCE_POOL_EXHAUSTED, the agent follows a structured checklist to find out whether broken, NotReady TPU nodes with auto-repair turned off are still occupying slots in the reserved capacity. It then explains the cause and recommends fixes, which the user applies themselves. The agent is told to run only read-only diagnostic commands and never to drain, delete or re-create nodes.

Great for

  • Platform or ML engineers running TPU workloads on GKE Standard clusters with a Compute Engine reservation
  • Teams whose TPU node pool scale-ups fail even though their reservation should have capacity

Skip it if

  • Autopilot clusters, because they always auto-repair nodes so this cause does not apply
  • General pod scheduling or container crash problems that do not involve a TPU reservation, since the skill's description excludes them

Try asking

  • “My TPU node pool creation failed with GCE_STOCKOUT even though I have a reservation. Can you find out why?”
  • “Our GKE TPU scale-up is stuck with ZONE_RESOURCE_POOL_EXHAUSTED. Check whether old NotReady nodes are holding reserved capacity.”
  • “Help me work out how many free slots remain in our TPU reservation before we create a new node pool.”
Step-by-step guide: how to use GKE TPU Stockout TroubleshooterInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 2 original files, the Apache-2.0 licence and an attribution note.

    gke-ai-troubleshooting-tpu-reservation-stockout.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip gke-ai-troubleshooting-tpu-reservation-stockout.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip gke-ai-troubleshooting-tpu-reservation-stockout.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give the agent the cluster name, project, zone and reservation name up front so it can run the checks without guessing.
  • Have the agent read the linked Google Cloud documentation sections it cites, and review any fix it proposes before running it yourself, since remediation steps are marked High Risk.
  • Use a read-only IAM account for diagnosis, as the skill asks for viewer roles and only requires Cluster Admin for remediation you choose to perform.

What's inside

  • references/attribution.md
  • SKILL.md

Redistributed unmodified from google/skills under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.