Hotel Documents, Set-Top Boxes and a New Homepage

Friday ran straight on into Saturday. Most of the work happened after midnight, between about 00:34 and 05:30, and it was three jobs back to back: finishing the hotel network documents, testing the TV boxes, and swapping the homepage. The afternoon and evening were quieter.

Finishing the topology documents

Just after midnight I finished the 2 October post and got it live, then went back to the hotel topology pages. I wanted them to match the type and layout on my portfolio site, so I changed the fonts, made the body text smaller, and made the big title a lot smaller after it wrapped onto two lines on a phone. I also built a dark version using the same palette as the portfolio's dark mode.

Next I retitled the pages, swapped in the location codes we already use in NetBox, and added rack positions. Only 40 of the 92 mapped devices had a NetBox entry, and 26 of those had a rack position. For the main rack I went with the layout I recorded on site rather than NetBox, because NetBox had two switches in the wrong slots.

Then I made the client version, in light and dark. It leaves out anything resolved, anything marked as informational, evidence file names, the on-site checklist, owner details and the proposed new VLAN. The numbering of the observations and open questions was redone, so the final set reads O-01 to O-28 and Q-01 to Q-16 with no gaps that hint at removed items. I added a glossary tab as well, 67 terms in one A to Z list, since most people reading it won't know what IGMP snooping is.

The first pass at keeping the proposed VLAN out of the client copy went wrong. A multicast section still carried the proposal, and my first fix removed ten observations that were fine to keep. I rewrote the check so it blocks the proposal wording, then restored the ten. Both client files are now clean, and the footer carries the capture date and who prepared it.

Do the set-top boxes get their own stream

There are seven set-top boxes in the test setup, and six of them were already awake. One was in standby, and it was the only one with a TV plugged in, so I left it alone. I woke the other six, got each past the welcome screen and put each on a different TV channel.

All six were receiving video, at between about 1 and 7.6 Mbit/s depending on the channel. Then I logged into the second TV switch read-only and looked at the multicast table. Each stream sat on only the port of the box that had asked for it, and nowhere else, so IGMP snooping is doing its job. The standby box's port had no stream, as expected.

Getting into that switch took two goes. I'd tried to pull the password out of the password manager the wrong way and every attempt was refused. The documented route, a references file run through the wrapper script, worked first time. I'd forgotten my own procedure.

A new www homepage

I wanted www.n5hq.me to follow the ericli design. I'd been sent a friend's site, which is a good fit, so I forked it and turned it into a homepage that is just "N5HQ", a theme toggle, an email link and the blue terrain along the bottom. A second, separate hosting project keeps it independent of the portfolio, and I checked that the old links still go where they used to: /about, /blog, /portfolio and /dev/v3 land on the portfolio subdomain, and /portal goes to the portal.

Moving the domains over was the one part I did by hand. I removed them from the old project and added them to the new one in the dashboard, and the site was serving the new page at 05:01. After that I replaced my own logo with the friend's original animated one, cut the headline to a quarter of its size, and added a "Coming soon." line under it that opens a full-screen video. The last push went out at 05:31.

Afternoon and evening

At about 13:20 I checked whether any of this was in NetBox. It wasn't, so I loaded the topology. Everything it touched carries a tag, the old value of every changed field is saved so I can roll back, and nothing was deleted. I checked the result through the API: 43 devices and 46 cables were created, 89 infrastructure addresses were added, and 9 VLANs went into their own group, without the proposed one. One rack in the separate building can't be tied to a building, because the evidence points two ways, so it has no location and is tagged as unconfirmed.

I also worked out the next steps for getting the TV system onto the hotel network: get the hotel's contractor to agree the plan first, then prove it on the island (the IPTV kit on its own isolated network), then pilot one building.

Around 17:15 I did cable maintenance on the home server. I'd meant to keep the nodes online, but the main switches went off at some point and the lab lost power in stages. The nodes rebooted about 18:16, and by 18:35 everything was verified up: quorum 5 of 5, all nine HA services started, all 21 containers running, and the UniFi switches and access points rejoined. One service had started before the seedbox mount was ready, and I fixed that.

In the evening I was thinking about remote access to the hotel's core switches. A Pi like the one in my lab would do it, as long as it's a new one and not pi02, which is the cluster's tie-break vote and my new DNS box.

Still open