Kubernetes
Раздел про Kubernetes для backend-разработчика: pod/deployment/service, сеть и Ingress, Spring Boot в кластере (probes, JVM, graceful shutdown), Helm и GitOps, отладка.
Why it matters for UCP. Kubernetes is the runtime environment for UCP services: probes, JVM resources, graceful shutdown, configuration, and rollout behavior are the developer's responsibility, not just the platform's. Without this mental model, "the cluster is acting up" incidents can't be diagnosed. Part of the learning program.
The section is split into five articles: the model (fundamentals), networking, the application in the cluster (Spring Boot), the path to production (deployment), and operations. The examples use the same order/limits services as the rest of the site.
Articles in this section
- Fundamentals: pod, deployment, service — desired state, pod mortality, the request path, the vocabulary of objects.
- Networking and traffic — Service types, DNS names, Ingress and Gateway API, endpoints + readiness, NetworkPolicy, where a service mesh fits.
- Spring Boot in Kubernetes — liveness/readiness/startup via Actuator, requests/limits and JVM memory, graceful shutdown, ConfigMap/Secrets, HPA.
- Deployment and configuration — manifests vs kustomize vs Helm, the mechanics of a rolling update, rollbacks, GitOps, migrations during zero-downtime rollout.
- Operations and debugging — the kubectl toolkit and an incident map: CrashLoopBackOff, OOMKilled, Pending, flapping readiness.
Related
- Graceful Shutdown Style Guide — the clean termination that a rolling update relies on.
- Observability Style Guide — logs and metrics, without which k8s incidents are blind.
- Spring Actuator — the health groups that probes are built on.
- Resilience patterns — what Kubernetes won't do for the service.