Getting Started with GPU Workloads on PodWarden

Developer Relations

A step-by-step guide to deploying your first GPU-accelerated workload: node labelling, resource requests, and validating GPU visibility inside the pod.

Getting Started with GPU Workloads on PodWarden

This tutorial walks you through deploying a simple GPU workload on a PodWarden-managed cluster.

Prerequisites

  • A cluster with at least one GPU node (NVIDIA or AMD)
  • The NVIDIA device plugin or AMD GPU operator installed via the catalog

Step 1 — Label the GPU Node

Add an accelerator label to the target node from the cluster admin panel or via kubectl.

Step 2 — Deploy a Test Workload

Create a Pod spec that requests one GPU unit under resources.limits. Set nodeSelector to match the accelerator label so the scheduler places the pod correctly.

Step 3 — Verify GPU Visibility

Check the pod logs for the device info table. A successful run confirms the GPU driver is visible inside the container.