Troubleshooting#

When something fails in an NVIDIA® AI Enterprise Infrastructure deployment, start here to find the fix quickly.

If a single product is failing (for example, GPU Operator pods, Container Toolkit runtime errors, vGPU™ install or licensing reachability failures, or Network Operator issues), use the component resource index to open that product’s troubleshooting or known-issues documentation.

Component Resource Index#

Use this table when a failure is isolated to one Infrastructure component. Open Troubleshooting for how to diagnose and fix the issue, and Known issues for documented product limitations.

Table 133 Infrastructure component troubleshooting resources#

Component

Troubleshooting

Known issues

NVIDIA Data Center GPU Driver

Data Center drivers documentation.

Data Center drivers documentation (release notes for your version).

NVIDIA DOCA™ Driver for Networking

BlueField® software troubleshooting.

DOCA documentation.

NVIDIA Fabric Manager

Fabric Manager user guide.

Data Center drivers documentation (separate package for the Virtual GPU Manager; ships with the Data Center GPU Driver).

NVIDIA DOCA Microservices

BlueField software troubleshooting.

DOCA services.

NVIDIA vGPU for Compute

vGPU for Compute user guide License server reachability.

Limitations (per-hypervisor known product limitations).

NVIDIA Container Toolkit

Container Toolkit troubleshooting.

Container Toolkit release notes.

NVIDIA Run:ai

NVIDIA Run:ai documentation.

NVIDIA Run:ai release notes.

NVIDIA DPU Operator (DPF).

DPF documentation (open Operational Readiness > Troubleshooting).

DPF documentation (open Release notes for known issues).

NVIDIA GPU Operator

GPU Operator troubleshooting.

GPU Operator release notes.

NVIDIA Network Operator

Network Operator common issues.

Networking documentation.

NVIDIA NIM™ Operator

NIM Operator documentation.

NIM Operator release notes.

NVIDIA Base Command Manager (BCM).

Base Command Manager documentation.

BCM 11.x release notes, BCM 10.x release notes.