I’ve been working on a project to learn kubernetes.
Since I’ve been unemployed I’ve noticed that a lot of Go Software Engineer listings involve Kubernetes for some reason, I guess there are more people adding operators or other modifications than I would expect, or maybe I’m misreading the market, but I’ve always wanted to understand this auspicious tech trend better anyway, so I’m working on a project called Cadence Reader.
I do happen to understand that K8s is about automated container deployment. I can understand how a company like any of the malicious 7 might need to scale any individual app up or down as the boomers or zoomers open and close apps on their phones at different hours of the day. I even understand that you’d want some kind of centralized, overcomplicated management mechanism for all the load balancing and certificate issuance and permission managment that must require.
I’ve gotten away with systemd my entire career, to the point where I’m now both familiar and comfortable with it, and dare I say it: I don’t think the companies I’ve worked for have had a real need for anything bigger, but at least at my last job there was enough compute deman or at least enough distribution and or separation of compute tasks that it was all run on a big cloud platform inside a managed kubernetes cluster. It’s very expensive, and that’s where my apprehension really starts: I don’t know why anyone goes out of their way to learn this technology except that nobody wants to hire you unless you already know their entire stack. There’s really not that much demand for these particular skills inside organizations and so it’s a diminishingly small group of people who ever get access to learn them in the actual enviroments where they’re deployed.
Anywho, I contrived a use case so that I would have to make a seprate landing page, web app for api service, and importer: A blog syndication site which makes a reader tap their screen or press the spacebar to load the next word, sentence, or paragraph of a blog post. By holding the content back and making them request each fragment I get accurate metrics about how far they read and how fast and even, perhaps, implied data about which segments of the text were most interesting if a large enough group of people happened to slow down in one particular spot. Measuring these things will be difficult, but at least it means I’ll have an easy way as my own end user to synthesize real world change in compute load: I read an article word-by-word and tap the spacebar as fast as I can and watch in k9s as compute load rises on the drip service and a new pod gets spun up to handle every other request, or something like that. That’s the hope. It’s possible Go is too efficient and I’ll still have to generate synthetic load, but it feels more interesting than writing another to-do app and putting it inside k3s or minikube, which is just transparent larping. I prefer my larping to be more opaque.
That’s all this blog post is for. I just had to add a post with this title as the mechanism to authorize the syndication platform to import all the posts. This way I can always delete just this one post to remove my blog from that site. Whether or not any given blog author has access to upload a DNS record or flat file for authorization (my original authentication ideas, still alternates I may add) I don’t know. I do know if you have a blog and you want to add it you should at the very least be able to put up a post with a specific title. This lowers the barrier to just necessitating an RSS feed, a weirdly common thing even though I expect so many younger internet dwellers are unlikely to know about it now.