Description
GKE ComputeClasses is an agent skill for configuring, optimizing and troubleshooting ComputeClasses on Google Kubernetes Engine. A ComputeClass describes the kinds of nodes a workload should run on, so you can ask for Spot VMs with an on-demand fallback to control cost, target specific accelerators such as L4, H100 or v5p TPUs for GPU and TPU workloads, pick a machine family such as c3, c4 or n4 for performance, or place workloads in a particular zone next to zonal resources. The skill's rules set an engagement order of giving a generalized answer first and refining it only if needed. It also tells the agent to verify subtle or undocumented behavior directly in the open-source cluster autoscaler code, using a local clone or the raw files on GitHub, and to report the upstream commit and date when a behavior changed. A code index reference maps packages and symbols. It is intended for platform engineers who want predictable cost and accelerator placement on GKE. It is published by Google on skills.sh under the google skills repository, with about 6.1K installs. The skills.sh listing shows Gen Agent Trust Hub and Socket passing, with Snyk showing Warn.…
Highlights
- Spot VMs with on-demand fallback
- Accelerator targeting for GPUs and TPUs
- Machine family and zone placement
- Verifies behavior in the open-source autoscaler code
- Generalized answers first, then refinement
Ratings & Reviews
No ratings yet
Be the first to rate gke-compute-classes.
People also visit
airunway-aks-setup
End-to-end AI Runway setup on AKS from bare cluster to running model deployment
gke-cluster-autoscaler
Trigger on mention of GKE cluster autoscaler, node autoscaling
google-cloud-solution-guided-gke-ai-migration
Guides the migration of existing AI workloads
gke-compute-classes FAQ
What is a ComputeClass?
A description of the kinds of nodes a workload should run on, which GKE uses to provision them.
What can I target with it?
Spot VMs with on-demand fallback for cost, GPUs and TPUs such as L4, H100 and v5p, machine families such as c3, c4 and n4, and zones.
Where is the source for behavior questions?
The skill points to the open-source cluster autoscaler repository and asks the agent to check the code for subtle behavior.
How do I install it?
Run npx skills add https://github.com/google/skills --skill gke-compute-classes.
What are the best gke-compute-classes alternatives?
Popular alternatives to gke-compute-classes include airunway-aks-setup, gke-backup-dr and gke-cluster-autoscaler. See all gke-compute-classes alternatives