Millet Porridge

Are basic elements really indispensable? How to express and use an object's internal state can follow different ideas. Readers can treat this as several ways of writing the same character — just clarify the logical concepts it contains.

How to organize programs with mathematical thinking — cleverly using one function to construct another

University had a similar elective, but I didn't study it well before and wrote the essay carelessly — failed outright. Recently with free time I'm going through that MIT course. This article mainly shares a few design ideas from it; readers with relevant background are advised before reading, otherwise you may not know what's being said.

kubectl is the official K8s command-line tool for conveniently operating K8s clusters. This article introduces some unconventional kubectl usages — not that you must try them, just that when needed you can remember kubectl may have similar features, instead of writing a new tool on the client-api.

Recently I integrated a K8s system whose Ingress support is incomplete with many features cut, but it supports full-featured Istio. So I researched and tried Istio's features to see whether they meet our needs. I didn't go deep into Istio's internals — just implementing specific features, understanding and using it my own way; corrections welcome.

An earlier blog introduced the scheme for fine-grained pod control — we can targetedly adjust pod counts by adding custom resources. But that's a basic primitive; what about providing it to users as a platform feature? This post revolves around this fixed-pod-count method, introducing how we leverage the feature.

I'd been clamoring to add a flame-graph profiling tool to our PaaS platform. I originally wanted to add it in mid-March but was delayed by other urgent matters. These two weeks I told my colleague I would definitely add this feature — and this week it went live as wished.

In earlier articles I introduced how to make random requests with wrk and how to use lua scripts. This blog mainly introduces how to finely control concurrent requests with wrk during stress testing.

Recently one machine in our self-built Kubernetes cluster was intruded and used for mining; we later found the cause — fortunately it was only used for mining... Network security is a serious matter; I hope readers gain some inspiration to harden their own clusters.

After a program migrated to our PaaS platform it would sporadically return 502. This article records the 502 investigation process.