December Adventure: That’s a wrap! – Stochastic Stenography

Welp.

Don’t know about you, but this December flew by for me. I should have anticipated that, come the week of Christmas, I would have approximately negative time with which to continue #DecemberAdventure. Still, it took me by surprise. So let me try to wrap up where things stand.

Cluster-wide reverse proxy

Even if I haven’t implemented this yet (due to reasons I’ll get to), I think at this point I understand pretty well how to do it. It more or less boils down to:

Set my local DNS (i.e., my Firewalla router) to point some version of local.my.domain to my reverse proxy
Create local CNAME entries that point to various local services
Add a *.local.my.domain sans domain in my traefik configuration to get the SSL to work for these endpoints
Set up a kubernetes Service object for each local service I want traefik to be able to access

That’s about it! I’m still a little iffy on step 2, mainly because I don’t know what will prevent any device on the network from skipping the top-level (local) domain pointing to the reverse proxy and going straight for the CNAME entry that is actually being requested. But I suppose, at this point, I just need to deploy it and see what happens.

Cat Magic deck

I haven’t really made any changes to my Cat deck—i.e., it’s still mostly a template—but I did receive a nice Amazon gift card for Christmas, so I used it to purchase some of the cards I’m quite sure I’ll need. Hopefully that will help me winnow this thing down to a standard 60-card deck.

Pi cluster next steps

In doing all this research for the reverse proxy, it became increasingly clear that I need to overhaul the entire cluster—hence the reason why I stopped just short of implementing anything regarding the reverse proxy.

Right now, the cluster has no “local” storage (meaning, no storage connected directly to any of the Pi SBCs); everything instead goes over the network. This works, certainly, but it’s not ideal, particularly when the power goes out or I need to selectively shut things down: I have to be careful in what order I shut down 1) the Pi cluster, 2) the network file storage, and 3) the network hubs connecting everything. It would be substantially easier if I had storage connected directly to the Pi machines that the k3s cluster could use.

So I used some Christmas money to purchase a couple of SSDs. Meaning that I’ve also spent a lot of time looking into Longhorn and Garage distributed block storage solutions for kubernetes clusters. I have everything ready to go there, but in so doing I need to prepare in case I’m not able to transfer everything over from the current network storage location.

Meaning I need to prepare to more or less start from scratch. Which will, frankly, simplify things. It’ll also give me opportunities to update the k3s clients, the base OSes (from Ubuntu 22 to 24), and the various helm charts (not to mention fixing the namespace issues I enumerated in an earlier post).

Admittedly it does feel like every personal project ends up with some version of “re-do everything I’ve already done so this one tiny addition will work,” but hey, isn’t that what learning is all about?

2025

I’m working on a 2024 Retrospective post. It’s gonna be a doozy. Hopefully I’ll have it up in the next couple days.

Citation

BibTeX citation:

@online{quinn2025,
  author = {Quinn, Shannon},
  title = {December {Adventure:} {That’s} a Wrap!},
  date = {2025-01-01},
  url = {https://magsol.github.io/2025-01-01-december-adventure-thats-a-wrap},
  langid = {en}
}

For attribution, please cite this work as:

Quinn, Shannon. 2025. “December Adventure: That’s a Wrap!” January 1, 2025. https://magsol.github.io/2025-01-01-december-adventure-thats-a-wrap.