Homelab and big upgrades to blog posts
With the new year, I've tried to make it a habit to write on this site more and not leave it in the dust. Now that I've removed the original iteration's concept of "projects" and any external integrations, the only things left are these blog posts.
So, I've made it a point to keep working on them at some sort of pace. That pace has already fallen off, and it's only April! I don't feel guilty about it, though, because I've been working on one "bigger project" for this blog that has required more time to get going. It's nearing completion, so it should hopefully be coming out soon after this one!
I also have been researching and looking into another project, which is what this post will go into.
I've started looking into building a homelab to help brush up on my kubernetes skills, and build out some fancier setups all locally so there's less risk. This has taken up a good portion of my free time so, less time building out any blog ideas.
So what's a homelab and what do you plan on doing with it?
A homelab is a cluster of computers, in my case, a handful of Raspberry Pis, hooked up together like a tiny data center. These tiny computers will then be leveraged to work with Kubernetes to run multiple different programs and test out larger-scale system designs all locally and fairly cheaply. There's a good amount of upfront cost, and I have to pay to run all these little machines, but it seemed like a better bet to get going than to set up the same thing online, then feel guilty about watching my money go away if I leave it running.
The sort of systems I'm looking into are self-hosting LLM models, which has been, let's say… an experience, haha. I'm also looking at other types of more complex systems like n8n and migrating some of my ideas to my local "server cluster" to leverage the infrastructure and costs.
But the main goal is to be able to play around with kubernetes without breaking the bank, and playing with some core constraints.
Raspberry Pis?
I've always been a fan of these systems but have always tried to do too much with them, mostly trying to run them as desktops. You can do that, but the experience isn't great. I've always heard of a "homelab" type project but was never in a place to commit to buying more than one new Raspberry Pi. I've been looking to make this change, as I have some extra funds that can go to a project like this. I also believe spending some money on something that will allow me to learn is an investment of sorts, so it's money well spent.
The main selling point of using Raspberry Pis for a cluster is when they idle and they don't use too much power. Having multiple units ramps things up, but it should help the bottom line since I pay for all the electricity they will end up using.
With multiple units, I'll be able to split up the load and be able to work with some level of purposeful complexity using Kubernetes. I'm even considering pulling out some older Raspberry Pis to put into the system for even more constraints.
So, what will they actually be running?
Kubernetes is the main target for learning, but I believe I'll play around with a few other systems and setups around LLMs/Ollama for local model running, as well as tools like n8n and my own custom-rolled scripts.
I think the main thing I'll be avoiding is running anything on the public internet due to scale and risk concerns. That will probably impact what I can do with stuff like n8n, but I'll take the hit there for the sake of simplicity.
There might be a blog post I end up writing in the future that details what I actually end up doing. So stay tuned!
Keep learning, keep building and keep asking questions!