Monitoring HTCondor Jobs and GPU Memory with Prometheus and Grafana

Chronological Source Flow

Back

AI Fusion Summary

Part 5 of a six‑part series details monitoring HTCondor jobs and GPU memory using Prometheus and Grafana. An RTX 3060 shows 4.18 GiB free VRAM while an unclaimed slot advertises zero GPUs, preventing a GPU‑requesting job from matching. The guide outlines troubleshooting steps: checking the scheduler, advertised resources, and the physical card to diagnose idle or mismatched jobs.