Interactive challenge

vLLM Inference Challenge

Deploy a GPU-backed OpenAI-compatible endpoint and prove scheduling, health, TTFT, queueing, and rollback readiness.

Prerequisites

GPU node poolGPU scheduling basicsModel artifact access

Guided step

Prepare the platform boundary

Create or identify the namespace, labels, and ownership metadata that make the workload reviewable.

Commands

kubectl create namespace llm-serving --dry-run=client -o yaml
kubectl label namespace llm-serving workload-class=llm owner=platform-ai --overwrite
kubectl get namespace llm-serving --show-labels

Expected signals

  • Namespace ownership is visible.
  • Workload class is encoded as a label.
  • The lab has a clear place to run validation commands.

Checks

Paste the `kubectl get namespace llm-serving --show-labels` output.

Confirm that the namespace has an owner label or documented owner.

Hints and solution

No hints opened for this step yet.