Slow Servers' Gemlog: Second Racking

Posted: 2026-01-26

Yesterday, I racked three more servers. Prometheus, the first host, is close to full (mostly with SporeStack infrastructure VPSs.) I also wanted to have a spare VPS host and a spare router in the rack.

It was my third time to the datacenter. The second was the initial racking, and the first was a tour. I had never had parking issues before. The datacenter is in downtown Spokane. There's paid meter parking right in front of the building, which the other two times had plenty of space.

Unfortunately, yesterday was a Sunday, when street parking was free. I did not forsee how little parking there would be! Fortunately, I did find a spot, a little ways away. I barely managed to take my laptop bag, bag of ethernet cables, keyboards, switch, etc, and the three servers all in one go. I was very glad to be using short depth servers!

This is particularly exciting as Spokane has quite the hobo problem. Most are harmless, but many seem pretty sketchy. I didn't run into any this time, though.

I'll probably try to avoid Sunday for racking new gear, or at least racking more than one server. At least I didn't have to use any more quarters.

Anyway, the servers are in. They include:

I should be able to fit 10 VPS hosts in the 1/3 cabinet before growing out of it. That does not include the spare.

Router + switches

I had been thinking that Thor and Freya could both share a switch with the ethernet uplink to Ardent Networks. If Freya failed, I could still get into Thor and bring things back online from here.

I'd have one switch for the uplink and one switch for the hosts in this scenario.

It doesn't address redundancy of the uplink, which may or may not be an issue -- something I'll discuss with Ardent.

Dilemma

The dilemma was that unplugging the uplink from Freya would result in some downtime. I could probably do it in a minute or two (or less?), but since I didn't schedule a maintenance, this didn't sit well with me. On one hand, 1-2 minutes of downtime now could save two hours of downtime (due to driving time) if Freya were to go down.

But, though Freya is a single point of failure, she seems stable, and this is probably something better to schedule.

I also wonder about having redundant switches for the VPS hosts. Something that might happen in the future, we will see.

Speaking of the future...

A future issue, identified

I realized that I don't have enough space on the PDU for 10 hosts. Currently enough for 6, maybe 7, depending on what I do with switch(es.) So not an issue for a while, but hopefully something I'll have to address in the future.

Now this may be something that can be resolved without any downtime. I'll talk to Neutron about it and see what we can do.

Pictures

The three new hosts have pictures! Just navigate to the hosts page, then each server's prospective page. Still nothing for Prometheus or Freya.

Unfortunately, I forgot my smartphone (which maybe I shouldn't have at all) and didn't have any way to take a picture of the cabinet after this second racking.

Random tidbit

Fun fact: Did you know that Spokane has a Freya street and a Thor street? I've never seen that anywhere else.