GitHub Repository

GPU Observability with workload attribution. One OTLP agent per node ties hardware metrics (NVIDIA, AMD, Intel Gaudi) to the K8s pod or Slurm job burning the GPU.

47 starsPython

L9gpu – GPU telemetry that ties each GPU to the K8s pod or Slurm job

Name: L9gpu – GPU telemetry that ties each GPU to the K8s pod or Slurm job
Availability: InStock
Author: nishantmodak

by nishantmodak·Jul 6, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidSolve My ProblemSlick

Finally ties GPU metrics to actual workloads when DCGM only gives you uuids.

Strengths

•Workload attribution solves the actual question DCGM leaves unanswered: which job is burning this H100.
•Vendor-neutral OTLP output means no lock-in to a specific observability backend.
•Multi-vendor GPU support across NVIDIA, AMD, and Intel Gaudi in one agent.

Weaknesses

•GPU observability is a well-funded, crowded space with established competitors.
•Core insight is obvious once stated—attribution via K8s/Slurm metadata isn't architecturally novel.