Collects hardware inventory and GPU attestation evidence from server BMCs over Redfish.
NVIDIA Fleet Intelligence Out-of-Band Collector
The Fleet Intelligence Out-of-Band Collector gathers hardware inventory and GPU attestation evidence from server BMCs over Redfish. It runs independently of the host operating system and uploads data to Fleet Intelligence for infrastructure inventory and GPU attestation workflows.
Features
- Collect inventory for systems, processors, memory, PCIe devices, storage, firmware, and other supported Redfish resources.
- Retrieve GPU attestation evidence using backend-issued nonces.
- Schedule collection across configured BMC targets with shared per-BMC concurrency and rate limits.
- Preserve successfully collected inventory when an individual Redfish domain fails.
Requirements and deployment
Use a compatible Linux container runtime or the Kubernetes Helm deployment. Provide network access to the target Redfish BMCs and Fleet Intelligence, each target's BMC address and MAC, a BMC credentials file, and a Fleet Intelligence bootstrap token for enrollment. Store credentials in Kubernetes Secrets for Helm installations.
The Helm deployment includes separate NICo hardware health and OpenTelemetry gateway workloads for BMC metrics and logs.
Publisher and license
Published by NVIDIA. The Fleet Intelligence Out-of-Band Collector source is licensed under the Apache License 2.0. Bundled components and dependencies retain their respective licenses; consult the included notices.