Technology advocacy · architecture principles

Golang · Kubernetes · Systems

Systems explained through code & live visual instances.

I write down the constraint I hit, and the one line I want to remember next time. Go and Kubernetes are the work. The lab is there so I can click the system, not only read about it.

01 // Technology Teaching Tracks

Browse all guides →

02 // Interactive Architecture Workbench

Full screen lab →

Select an architecture instance below and click the state controls inside the diagram to simulate Pod crashes, Goroutine fan-out, Zigbee mesh hops, or 60FPS game engine loops:

Interactive Architecture Instance · Kubernetes

Kubernetes Control Plane, Self-Healing & Service Routing

STATE: STEADY Desired = 3 · Ready = 3 · ClusterIP 10.96.0.10 load-balancing across all 3 healthy Go API pods.
CONTROL PLANE (MASTER NODE) · DESIRED STATE ENGINE WORKER NODES · DATA PLANE & POD WORKLOADS (NAMESPACE: production) Watch / Reconcile YAML Controller Manager ReplicaSet Loop: Desired = 3 Status: 3/3 Matched kube-apiserver REST + Watch Gate (:6443) AuthN / AuthZ / Admission etcd Store Raft Consensus KV /registry/pods/... kube-scheduler Filter & Score Nodes CPU/Mem Requests OK Ingress (:443) api.divineforge TLS Termination kubelet + kube-proxy CRI Containerd + IPVS Probes: /healthz :8080 Service: ClusterIP go-api-svc · 10.96.0.10:80 Endpoints: 3 Ready Pods ConfigMap + PVC env-config · db-secret PVC: 20Gi (ReadWriteOnce) pod/go-api-7f8c-1 10.244.1.12 · Node-01 Image: go-api:v1.4.0 ● Running (Ready) pod/go-api-7f8c-2 10.244.1.18 · Node-01 Image: go-api:v1.4.0 ● Running (Ready) pod/go-api-7f8c-3 10.244.2.09 · Node-02 Image: go-api:v1.4.0 ● Running (Ready) pod/go-api-4 Node-02 Slot Standby / Surge ○ Idle Slot Control Loop Networking / Service Healthy Pod (Ready) Fault / Evicted Pod State / etcd / PVC

Desired-State Reconciliation

  • • Never deploy bare Pods; a Deployment manages a ReplicaSet.
  • • When a Pod crashes, Controller Manager detects actual < desired and schedules a replacement automatically.

Service Discovery & Readiness

  • • Pods are ephemeral and their IPs change on every reschedule.
  • • Service (ClusterIP) + readinessProbe ensures traffic only routes to healthy containers.

12-Factor Config & Persistence

  • • Decouple config via ConfigMap and credentials via Secret mounts.
  • • Stateful workloads bind a PersistentVolumeClaim (PVC) so data outlives pod restarts.
Companion Essay: Kubernetes Architecture From First Principles Read full technical article →

03 // Latest Writing & System Builds

All writing →

gZen is the other room. I go there when I am done taking a system apart. It is a slow walk on a low-poly shore, a sanctuary, and a card when I stop. STACKY is the one on the path. Nothing to finish, and nothing to streak. The forge stays on this site. The walk is gzen.io.

04 // Explore the Forge

Search DivineForge