feat: add fabric-manager-monitor for GPU infrastructure health checks#891

Open

dmvevents wants to merge 17 commits into

NVIDIA:mainfrom

dmvevents:feature/fabric-manager-monitor

Commits on May 8, 2026

feat: add fabric-manager-monitor for GPU infrastructure health checks (#883)
dmvevents
committed
fix: address CodeRabbit review findings
dmvevents
committed
fix: address CodeRabbit round 2 review findings
dmvevents
committed
docs: add ADR-030 fabric-manager-monitor scope to docs/designs/
dmvevents
committed
refactor: remove DCGM-overlapping checks from fabric-manager-monitor
dmvevents
committed
feat: add per-GPU fabric state metrics with gpu_index
dmvevents
committed
feat: split FM_UNRESPONSIVE into 3 granular failure states
dmvevents
committed
refactor: rename fabric-manager-monitor to system-service-monitor
dmvevents
committed
docs: add GKE container-optimized-OS monitoring approach to README
dmvevents
committed
demo: rename system-service-monitor -> system-services-monitor and drop DCGM duplicates
dmvevents
committed
fix: address CodeRabbit major findings (logger fallback, TOCTOU, alert severity)
dmvevents
committed

Commits on May 14, 2026

refactor: remove cuda_validation from health-monitor (will land in preflight-checks/ in follow-up PR)
dmvevents
committed