
Gke AI troubleshooting tpu vm repair stuck
Diagnoses GKE TPU VMs stuck in Compute Engine REPAIRING state and routes recovery
- Scripts
- None
- Platforms
- Claude Code, Codex, Cursor
- GitHub stars
- 24.4K
- Last commit
- 2 days ago
- License
- Apache-2.0
- Listing updated
- 1 day ago
Similar skills
About Gke AI troubleshooting tpu vm repair stuck
Gke AI troubleshooting tpu vm repair stuck is an open-source agent skill for cloud and infrastructure. It is available on Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose.
This skill walks an agent through finding GKE TPU VM instances that are stuck in the Compute Engine REPAIRING state. It checks the managed instance group status, reviews Compute Engine host error and maintenance events in audit logs, and compares node readiness and slice impact across single-host, multi-host, dynamic slicing and TPU All Capacity mode node pools.
The workflow also checks whether node auto-repair is enabled, reviews the GKE repair operation history, and looks for faulty host reports in All Capacity mode. The skill is read-only: it tells the agent never to drain, delete or re-create nodes, and it gives any fix to the user to apply. Where a different problem applies, it hands off to sibling troubleshooting skills.
The agent needs gcloud, kubectl and access to the GKE cluster, with the Kubernetes Engine Viewer, Compute Viewer and Logs Viewer roles. The skill uses gcloud and kubectl commands run in the terminal, and it links to Google Cloud documentation for each step.
When your agent uses it
Diagnose GKE TPU VM instances stuck in Compute Engine repair (REPAIRING state) after host errors, maintenance events, or faulty host reports in single-host, multi-host, dynamic slicing, and TPU All Capacity mode node pools. Use when TPU instances in a slice node pool remain in REPAIRING, host error events appear in audit logs, or TPU slices are unschedulable due to a VM in repair. Don't use for nodes reporting Ready: Unknown when no VM is in REPAIRING (use gke-ai-troubleshooting-node-unresponsive-timeout), other NotReady nodes when no VM is in REPAIRING (use gke-node-notready), Pending Pods when all nodes are Ready (use gke-workload-troubleshooting), JobSet preemption events (use gke-ai-troubleshooting-jobset-interruption), or Slice custom resource lifecycle failures (use gke-ai-troubleshooting-tpu-dynamic-slices-monitoring).
From the skill's SKILL.md. Your agent reads this description to decide when to load the skill.
What it can access
From a static check of its 2 files (26 KB) on GitHub on 17 November 2026. Nothing was run. Read the files before you install a skill: it can tell your agent to run anything its tools allow.
Check another skill before you install it.
Install
Copy the skill folder to ~/.claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck to use it in all your projects:
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/google/skills.git "$tmp" &&
mkdir -p ~/.claude/skills &&
rm -rf ~/.claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
cp -R "$tmp/skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuck" ~/.claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
rm -rf "$tmp"For one project only, put the folder in .claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck inside it instead.
Copy the skill folder to ~/.agents/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck to use it in every repository:
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/google/skills.git "$tmp" &&
mkdir -p ~/.agents/skills &&
rm -rf ~/.agents/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
cp -R "$tmp/skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuck" ~/.agents/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
rm -rf "$tmp"For one repository only, put the folder in .agents/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck inside it instead.
Copy the skill folder to ~/.cursor/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck to use it in every project:
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/google/skills.git "$tmp" &&
mkdir -p ~/.cursor/skills &&
rm -rf ~/.cursor/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
cp -R "$tmp/skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuck" ~/.cursor/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
rm -rf "$tmp"For one project only, put the folder in .cursor/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck inside it instead.
Run in a terminal (add --scope workspace to install it for the current workspace only):
gemini skills install https://github.com/google/skills.git --path skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuckOr copy the folder to ~/.gemini/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck yourself.
For GitHub Copilot in VS Code, copy the skill folder to ~/.copilot/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck:
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/google/skills.git "$tmp" &&
mkdir -p ~/.copilot/skills &&
rm -rf ~/.copilot/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
cp -R "$tmp/skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuck" ~/.copilot/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck &&
rm -rf "$tmp"For one project only, put the folder in .github/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck inside it instead.
Get the folder:
git clone --depth 1 https://github.com/google/skills.gitThen copy skills/skills/cloud/gke-ai-troubleshooting-tpu-vm-repair-stuck into your agent's skills folder as gke-ai-troubleshooting-tpu-vm-repair-stuck. Agents that support SKILL.md read one folder per skill.
What it does
- VM repair status check
- Lists managed instances and flags any VM reporting REPAIRING or another non-running action.
- Audit log event review
- Reviews host error, maintenance and restart events in Compute Engine audit logs around an incident.
- Node pool impact analysis
- Explains how a repair affects single-host, multi-host, dynamic slicing and All Capacity mode pools.
- Auto-repair verification
- Checks whether auto-repair is enabled and reviews recent GKE repair operations and their failures.
- Faulty host report lookup
- Looks for host-as-faulty operations in All Capacity mode and sets expectations for multi-day repairs.
- Escalation and handoffs
- Routes unrelated node or workload problems to sibling skills and prepares details for Cloud support.
Good for
- Find out why a TPU VM in my slice node pool has been in REPAIRING for hours
- Check audit logs for host errors that happened on a TPU node
- Determine whether auto-repair is enabled on my TPU node pool
- Review a failed GKE repair operation and decide whether to contact support
- Understand whether a faulty host report in All Capacity mode explains a stuck VM
Gke AI troubleshooting tpu vm repair stuck FAQ
What do I need before using this skill?
You need gcloud and kubectl set up with cluster credentials, the Google Kubernetes Engine, Compute Engine and Cloud Logging APIs enabled, and the Kubernetes Engine Viewer, Compute Viewer and Logs Viewer IAM roles.
Does this skill change my cluster or VMs?
No. It is read-only and tells the agent not to drain, delete or re-create nodes or change the cluster. Any fix is given to you to apply yourself.
When should I not use this skill?
Use other skills when nodes report Ready Unknown with no VM in REPAIRING, other NotReady nodes with no VM in REPAIRING, Pending Pods on healthy nodes, JobSet preemption, or Slice custom resource failures.
How long can a faulty host repair in All Capacity mode take?
The skill cites the TPU documentation, which says the host repair process can take between 3 and 14 days, or longer.
How do I install Gke AI troubleshooting tpu vm repair stuck in Claude Code?
Copy the skill folder from GitHub to ~/.claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck to use it in all your projects, or to .claude/skills/gke-ai-troubleshooting-tpu-vm-repair-stuck in one project.
Does Gke AI troubleshooting tpu vm repair stuck run scripts?
No. It is instructions only, with no scripts. This comes from a static check of the files on GitHub; nothing was run. Read the files before you install it.
Which AI agents can use Gke AI troubleshooting tpu vm repair stuck?
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose. Each reads skills in the SKILL.md format from its own folder; the install steps on this page list them.
What are the best alternatives to Gke AI troubleshooting tpu vm repair stuck?
The closest alternatives to Gke AI troubleshooting tpu vm repair stuck are Gke AI troubleshooting tpu reservation stockout, Gke AI troubleshooting node unresponsive timeout, Gke AI troubleshooting tpu slice mig stuck updating, Gke AI troubleshooting handle disruption GPU tpu and Gke AI troubleshooting tpu partition overlap.
Where this comes from
- Description: from the skill's own SKILL.md.
More from google/skills

Route platform choices to Google's own skills by fetching and filtering its skill catalog
- Open source
- 24.4K GitHub stars
- Cloud and infrastructure

Guardrails for gcloud commands: validate syntax with help, scope projects, dry-run first
- Open source
- 24.4K GitHub stars
- Cloud and infrastructure

Authenticate to Google Cloud: ADC, impersonation, service accounts and IAM roles
- Open source
- 24.4K GitHub stars
- Cloud and infrastructure

First-time Google Cloud setup: sign in, pick a project, link billing, with consent gates
- Open source
- 24.4K GitHub stars
- Cloud and infrastructure

Look up Google developer docs through the Developer Knowledge MCP server or REST fallback
- Open source
- 24.4K GitHub stars
- Developer tools

Add, remove or clear audience members with the Google Data Manager API
- Open source
- 24.4K GitHub stars
- Marketing and SEO



