日本語はわからないけど、
やっぱりかっこいい。
Experiments, troubleshooting notes, tests, and things worth remembering.
Part 3 closes the troubleshooting series with an evidence ledger: what each test proved, which suspects were cleared or rejected, what remains open, and the exact restart point for the next session.
A practical Layer 1–7 network troubleshooting runbook built from the same incident as Part 1: capture link state first, prove VLAN and route locality, test both TCP directions, and change one variable at a time.
A bottom-up troubleshooting diary: an 80 Mbps symptom, a Fast Ethernet PoE injector, a marginal coupler panel, one-way TCP collapse, a bad VLAN test, and the local 720 Mbps result that finally cleared the AP.
Turning stack assumptions into executable health invariants, authenticated management checks, stateful DOWN/RECOVERED alerting, and an external dead-man monitor.
How I layered filename policy, queue enforcement, authoritative imports, safe canaries, and read-only media validation without pretending any one control makes downloads safe.
A practical way to structure a Docker media automation stack: clear service ownership, hardlink-friendly storage, and a deliberately small VPN failure domain.
