TA-Nvidia

TA-Nvidia is a free, source-available Splunk add-on that brings enterprise NVIDIA GPU observability to any Splunk instance running on hosts with NVIDIA drivers. It repeatedly captures the COMPLETE nvidia-smi state; temperature, utilization, fan speed, power draw and limits, framebuffer memory, graphics/SM/clocks, ECC errors, PCIe throughput, performance state, compute processes, vGPU, and more and indexes it two ways with zero manual configuration: 1. Full JSON events (sourcetype nvidia:smi) in index=main for deep-dive fielded search, and 2. Time-series metrics (nvidia.* metric names) in index=metrics via a metric-schema pipeline, enabling fast mstats trends and visualizations. Two pre-built dashboards ship with the app GPU Overview (per-GPU cards for temp, utilization, memory, power, clock, fan) and GPU Metrics (long-window trended charts via mstats). Why this matters: GPU-equipped servers are expensive and failure-prone. Overheating, ECC errors, clock throttling, memory exhaustion, and power-limiting silently destroy GPU workload performance (AI/ML training and inference, VDI, HPC, video/transcode). Splunk has no free, turnkey NVIDIA telemetry app, TA-Nvidia closes that gap in about one minute, on any distribution (Tested on datacenter and workstation cards, current driver: 580.x, RTX PRO 6000 Blackwell). Licensing: free for personal, evaluation, and non-commercial use. Commercial/production use requires a license from the author: michael@bentleypc.com. Source is available on GitHub: https://github.com/bentleymi/TA-Nvidia