Application Management

Kubernetes Serverless Without the Vendor Lock-In (Here's How)

Traffic is never constant. Maybe your app gets hammered during business hours and barely touched at night. Maybe it’s steady all day but spikes unpredictably. Maybe there are 15 minutes a day when nobody’s using it at all. The point is, a fixed number of replicas is always wrong. You’re either wasting resources or under-provisioned.

What you actually want is an app that scales with demand. More replicas when traffic goes up. Fewer when it drops. And in the extreme case, zero replicas when there’s no traffic at all. Now, scaling to zero is easy. Just set the replica count to zero and you’re done. The hard part is coming back up without losing any requests. If someone sends a request and nothing is running, that request needs to be held, not dropped.

That’s what we’re building today. Not with Knative. Not with AWS Lambda. Just standard Kubernetes with a few smart components wired together. We’ll start with a single static replica, add Prometheus-based autoscaling, and then push it all the way to true scale-to-zero with zero lost requests.

AI Meets Kubernetes: Simplifying Developer and Ops Collaboration

Platform engineers face a tough challenge: developers know what they need, but they don’t understand complex infrastructure. Platform engineers understand infrastructure, but they struggle to anticipate every developer requirement. The result? Internal Developer Platforms that miss the mark and platforms that require endless iterations to make them useful.

But what if there was a third collaborator that could bridge this gap? AI has vast technical knowledge but doesn’t know your company’s specific rules and constraints. However, when you combine developer intent, platform engineering expertise, and AI’s technical knowledge, something powerful emerges.

I’m about to show you a working system that demonstrates this three-way collaboration. We’ll watch developers express their needs in natural language, see AI translate those into deployments using platform engineering building blocks, and then I’ll reveal the architecture that makes this possible.

Exploring KCL: Configuration and Data Structure Language; CUE and Pkl Replacement?

I’m in pain… and it’s self-inflicted… and I like it.

I tend to go through an endless number of tools, services, and format in search for better ways to do my job. I’m never satisfied. I always think that there is something better out there. So I go through pain of learning a new tool or a language only to jump into a new one shortly afterwards. Spending endless hours going through new stuff does not make sense, but I can’t help myself. Hopefully, I might save you from doing it yourself. That’s my goal. Go through the pain of trying out everything so that you don’t have to.

Is Pkl the Ultimate Data Format? Unveiling the Challenger to YAML, JSON, and CUE

Pkl is CUE killer!

Pkl is JSON and YAML killer!

Those are two sentences I heard lately. The former could be the case, while the latter is just silly. Let me explain.