The system we build: "Orbit"

Orbit is a small but realistic order-processing platform. It is deliberately shaped like the systems you run at work: a synchronous HTTP path, an asynchronous queue path, a relational database, a cache, and background workers that have to scale on queue depth rather than CPU.

                                   ┌──────────────────────── Kubernetes cluster ───────────────────────────┐
                                   │                                                                         │
  client ──HTTPS──► Gateway API ───┼──► catalog (Go, 3+ pods) ──► Redis (cache)                              │
  (k6 / curl)      (Envoy Gateway) │        │                                                                │
                   + cert-manager  │        └──────────────► Postgres (CloudNativePG, 3 instances)           │
                                   │                              ▲                                          │
                                   ├──► orders (Go, 3+ pods) ─────┘                                          │
                                   │        │  XADD orders.created                                           │
                                   │        ▼                                                                │
                                   │     Redis Stream ◄── XREADGROUP ── worker (Go, 0..N pods, KEDA-scaled)  │
                                   │                                                                         │
                                   │  every pod ──OTLP──► OTel Collector (agent DS) ──► OTel gateway ──┐     │
                                   │                                                                    │     │
                                   │      Prometheus ◄─metrics──────────────────────────────────────────┤     │
                                   │      Loki       ◄─logs─────────────────────────────────────────────┤     │
                                   │      Tempo      ◄─traces───────────────────────────────────────────┘     │
                                   │      Grafana (dashboards, SLOs)  Alertmanager ──► Slack / PagerDuty      │
                                   │                                                                         │
                                   │  Argo CD (GitOps)  Argo Rollouts (canary)  KEDA  Kyverno  Velero        │
                                   └─────────────────────────────────────────────────────────────────────────┘

Component	What it does	What it teaches
`catalog`	`GET /products`, `GET /products/{id}`; read-heavy, Redis cache-aside	stateless scaling, HPA on CPU, cache patterns
`orders`	`POST /orders` writes Postgres, publishes an event to a Redis Stream	transactions, idempotency, canary releases, RPS-based scaling
`worker`	consumes `orders.created`, "charges payment", marks order paid	queue-driven autoscaling (KEDA, scale-to-zero), retries, poison messages
Postgres	system of record (CloudNativePG operator)	StatefulSets, operators, backups, failover
Redis	cache + stream	StatefulSet basics, persistence trade-offs
Observability	OTel Collector, Prometheus, Loki, Tempo, Grafana	the three signals, correlation, SLOs
Platform	Envoy Gateway, cert-manager, Argo CD, Argo Rollouts, KEDA, Kyverno, Velero	the "platform layer" every real cluster grows

Final repository layout (you build it up chapter by chapter):

orbit/
├── go.mod
├── cmd/
│   ├── catalog/main.go
│   ├── orders/main.go
│   └── worker/main.go
├── internal/
│   ├── platform/         # shared: config, logging, telemetry, http server, health
│   ├── catalog/
│   ├── orders/
│   └── worker/
├── migrations/           # SQL migrations (goose)
├── Dockerfile            # one multi-stage Dockerfile, SERVICE build-arg
├── .dockerignore
├── compose.yaml          # local dev stack
├── load/orders.js        # k6 load test
├── lab/                  # Lima VM: create-vm.sh, bootstrap.sh, extract-blocks.py (§0)
├── deploy/
│   ├── base/             # Kustomize base: one folder per service
│   └── overlays/{dev,staging,prod}
├── platform/             # cluster add-ons as Argo CD Applications (Helm values)
└── .github/workflows/ci.yaml