Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL
All skills

gke-compute-classes

Write and debug GKE ComputeClass YAML for Spot fallback, GPU/TPU targeting, and machine-family selection.

  • Part of google/skills
  • 7.9k installs
  • DevOps & Self-hosting · Containers and clusters

When to use it

You have a workload spec but need a ComputeClass that picks Spot with on-demand fallback or targets specific accelerators, and pods are stuck Pending.

For example

You say

Use gke-compute-classes to write a ComputeClass for my stateless batch job in us-central1: prefer Spot N4, fall back to on-demand C3, and target L4 GPUs for one tier.

You get
  • ComputeClass YAML template
  • Pod toleration notes
  • CUD alignment comment

How it works

  • Caps a primary machine family with a CapacityQuota so excess demand spills to fallback families instead of pending.
  • Explains the two meanings of "balanced": node spread via locationPolicy versus pod spread via topologySpreadConstraints.
  • Flags schema traps like bootDiskSize versus bootDiskSizeGb and the three-rule limit on shared priorityScore.

Before installing

This skill includes a script your assistant may run on your computer.