Kubernetes
Container orchestration
The blog has become a teaching channel, with the real releases arriving as Gateway API and deprecation notices.
◆Recent moves
- 7d ago
How to Pretty-Print Your Kubernetes YAML as KYAML and Why You'd Want To
SIG CLI's KYAML — a strict subset of YAML defined in KEP 5295 — gets a practical walkthrough. It fits the project's pattern of narrowing what users are allowed to express rather than adding expressiveness: no new parser, no new format, just fewer ways to write a manifest that parses differently than it reads.
View source ↗ - 15d ago
Gateway API v1.6: TCPRoute and UDPRoute Graduate to Standard
⚡ SPARKGateway API v1.6 takes raw TCP and UDP routing to GA in the v1 API, extending the project's networking standard past the HTTP and TLS layer it had established. The simultaneous split of experimental resources into a separate API group is the same boundary-drawing instinct visible in KYAML and the v1.37 deprecations.
View source ↗ - 18d ago
Kubernetes v1.37 Sneak Peek
The v1.37 preview lists what breaks: kubectl run loses --filename, and static Pods can no longer reference Secrets or ConfigMaps — a bug that let them read API resources they were never meant to touch. Both are corrections of accidental capability rather than feature removals.
View source ↗ - 20d ago
How the controller-runtime Cache Actually Works, and Why Your Controller Does Not Crash the API Server
A long explainer on how controller-runtime's cache serves Get and List calls from an informer rather than the API server. Useful for controller authors, but it documents existing behaviour and changes nothing in the platform.
View source ↗ - 1mo ago
Building a Custom Metrics Exporter for Kubernetes
A build-it-yourself tutorial for a Prometheus metrics exporter feeding the HorizontalPodAutoscaler. Educational content on existing primitives, with no product change behind it.
View source ↗ - 1mo ago
Operating AI/ML Workloads on Kubernetes: A Headlamp Plugin for Kubeflow
A Headlamp plugin surfaces Kubeflow's CRDs — notebooks, training jobs, pipelines — inside a general-purpose Kubernetes UI, so operators debugging a stuck ML workload stay in one tool instead of dropping to kubectl. Concrete evidence that AI/ML workloads are being treated as ordinary cluster tenants.
View source ↗