A hands-on walkthrough of using HolmesGPT with Azure AI Foundry and Azure OpenAI to diagnose a deliberately broken Kubernetes pod. The setup uses a local kind cluster with a pod stuck in Pending due to a nodeSelector mismatch. HolmesGPT autonomously investigates the cluster state, reads events, checks nodes, identifies the root cause, and suggests concrete fixes — either removing the nodeSelector or labeling the node. The post highlights how AI-assisted troubleshooting can reduce the manual kubectl command-chaining typically needed to diagnose Kubernetes scheduling issues.
Table of contents
PrerequisitesRun the DemoWhat HolmesGPT FindsFix OptionsApplying the FixWhy Bother?Going FurtherFinal ThoughtsResources1.1K Impressions