Command & flag index

Every command and flag the guides teach, in one place: what it does, and when and why you'd reach for it. Search it like a reference, or switch to cards to drill.

How to use it. Type a command, a flag, or a symptom ("noisy", "slow rail", "drain", "MIG", "802") to filter everything at once. Table view is for looking things up. Cards view is for drilling: click a card to flip it, and mark the ones you know so they drop out of the deck (your marks are saved in this browser).

Quiz view tests you on whatever the filters select: pick Domains 1 + 4 and quiz only those, or one area, or one objective. Questions are multiple choice in three styles: what a command does, which command matches a description, and which command fits a situation. Misses are remembered in this browser so you can drill just those.

Each tool's heading says why the tool exists and links to the guide section that teaches it. Exact flags can vary by version: when in doubt, the tool's own --help wins.

Learn the tools in the order you'd use them on a broken node: is it alive (BMC: ipmitool, Redfish) → is the box healthy (nvsm) → are the GPUs healthy (nvidia-smi, dcgmi) → is NVLink right (Fabric Manager, nvidia-smi nvlink) → is the fabric right (ibstat, ibdiagnet, mlxlink) → does it perform (nccl-tests, HPL).
Exam coverage by objective

Each entry carries the NCP-AII blueprint objectives it serves, colored by domain: 1.x Bring-Up 31%4.x Test & Verification 33%3.x Control Plane 19%5.x Troubleshooting 12%2.x BlueField, MIG 5%beyond the blueprint. Click an objective to filter by it. Objectives with 0 entries are knowledge objectives (sequences, topologies, physical installation) covered in the guides rather than by commands.

Watch the look-alikes: nvidia-smi (GPUs) vs nvsm (whole DGX) vs dcgmi (GPU diagnostics); ibstat (my port) vs sminfo (the SM) vs ibdiagnet (whole fabric); flint (one device's firmware) vs mlxfwmanager (all devices, update) vs mlxconfig (settings, not firmware); -n in nccl-tests (iterations) vs -n in srun (tasks) vs -n in cmsh imageupdate (nodes).