Practical self-hosting · beginner
02 · A machine is not an inventory
Separate remembered hardware from observed capabilities.
What you can take away
- Build a dated inventory that identifies uncertainty.
- On a machine you administer, inspect the OS through its settings or supported release command, record the installed container-engine version, and, only if NVIDIA tooling is installed, run `nvidia-smi -L`
Your trail through this idea
Checking saved progress…
An old conversation says a server has a particular GPU. A container sees no GPU today. Both statements can be true. Ian’s NVIDIA and local-model troubleshooting history makes the difference between a remembered plan and an observed system especially important.
Prerequisite: complete the preceding lesson or bring its worksheet.
Understand the system
An inventory has a time, a scope, and evidence. A hardware label describes a device; a driver version describes software; an application capability describes what one process can actually use. Keep these in separate columns. Never promote a shopping plan, pasted example, or assistant suggestion into a current deployment fact.
NVIDIA’s management-interface reference documents inspection commands. On an NVIDIA-equipped practice host, nvidia-smi -L lists visible devices. A successful result does not prove that a particular container runtime or model can use them. An unavailable command also does not prove that the machine lacks hardware.
Work through the case
Use an original inventory with rows for operating system, storage pool, container engine, GPU, model server, identity provider, and backup destination. For each, record observation date, observation method, owner, and confidence. Give uncertain rows a next check instead of filling them from memory. Keep serial numbers, hostnames, addresses, and inventory outputs private.
Make something useful
Try this
Original exercise
On a machine you administer, inspect the OS through its settings or supported release command, record the installed container-engine version, and, only if NVIDIA tooling is installed, run nvidia-smi -L. Copy only sanitized capability summaries to your service card. Compare one row with a historical note and explain any mismatch. The lab is read-only; stop if you are unsure which host your terminal controls.
Check your reasoning
A quick check · for your own learning
What does a successful host GPU listing establish?
A self-study prompt. You can retry; this is not an exam.
Keep the result and follow the sources
Your deliverable is a dated capability sheet, not a universal description of Ian’s current hardware. A useful unknown is more valuable than an impressive invented specification. Next, identify which account is allowed to use those capabilities.
Sources and scope
- Primary documentation or archival source, reviewed 14 November 2026. Additional direct links appear beside the claims they support.
- Editorial basis: Ian's historical self-hosting questions and current repository inspection informed the case selection. Historical requests do not establish current deployment. Private chat records, credentials and host details are excluded.
- Examples and exercises are original educational synthesis. External sources support the stated technical or historical context, not every hypothetical scenario.