AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

aks-gpu-inference

Diagnose Day-2 AKS GPU and KAITO incidents using profile-aware, read-only evidence. WHEN: 'Insufficient nvidia.com/gpu', GPU pod Pending, model-load OOM, DCGM/VRAM, KAITO Workspace not ready, or GPU autoscaling. DO NOT USE FOR: setup (airunway-aks-setup), non-GPU incidents (aks-troubleshooting), standalone VM quota (azure-quotas), or generic cost (cost-analysis or cost-optimization from the optional azure-cost plugin).

Original source
microsoft
License
MIT
Pinned revision
3f5b461b2314e01c400564051e92469114c5b68f

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

Ask your assistant

Use Agentic Memory skills to read microsoft--aks-gpu-inference at revision 3f5b461b2314e01c400564051e92469114c5b68f, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 5

references/kaito-workspaces.md

KAITO Workspace Day-2 readiness

This is diagnosis only. Route add-on enablement, Workspace creation, model selection, and provider selection to airunway-aks-setup.

Read-only evidence

kubectl get workspace <workspace> -o yaml
kubectl describe workspace <workspace>
az aks show --subscription <subscription> -g <resource-group> -n <cluster> \
  --query '{id:id,location:location,provisioningState:provisioningState}' -o json
az vm list-usage --location <region> -o json
az vm list-skus --location <region> --resource-type virtualMachines \
  --all -o json

Microsoft Learn says machine readiness can take up to 10 minutes and Workspace readiness up to 20 minutes depending on model size. After the observed duration exceeds the applicable documented window, use only status and conditions reported by the Workspace, including documented ResourceReady, InferenceReady, and WorkspaceSucceeded when present. Do not invent NodeClaim, controller-log, or private condition chains.

Correlate a not-ready status with:

  1. exact GPU-family quota usage and limit in the AKS region; and
  2. exact Workspace GPU instance type availability or restrictions in that region.

If neither is decisive, preserve the status, timestamps, region, instance type, and missing evidence, then hand off to the cluster/quota owner or Azure Support.

Deleting a Workspace does not remove GPU node pools provisioned for it; Microsoft Learn requires those pools to be found and deleted manually. Present Workspace deletion and each pool deletion as separate charged-resource mutations, explain the cleanup dependency, and wait for explicit authorization. Do not stop/start the cluster as a diagnostic action.

Source: AI toolchain operator.