Deployment model · one service, end to end

Draft: not yet agreed. Nothing here is settled until you answer the open points.

Example: service tin on platform nitroso · Bun API + worker · Postgres + queue

What lands in the clusters

Each landscape gets the same app objects on every one of its clusters, plus one set of records on Primordial. Only names, config and secret values change per landscape.

  • App cluster, namespace nitroso: 2 Deployments, a Service, an HTTPRoute, a ConfigMap, an ExternalSecret (→ Secret), a db-init Job. DECIDED · R20
  • Primordial: the dependency records, the Problem record, Grafana alerts and dashboards, and the Argo CD Applications and Kargo stages that deploy it all. DECIDED · ARCH §8
  • No ServiceMonitor: apps push telemetry to Alloy. DECIDED · R20; observability §2

Switch the landscape: names and values change; the set of objects stays the same.

Landscape

pichu = dev. First stop; promoted automatically when a Freight appears. Database type in the PoC: Neon. DECIDED · ARCH §8 PoC TESTED · PoC

  • designed and decided
  • built in a branch or the PoC
  • open or unclear

{l}-{mark}an app cluster of {l} ({tier}). Every cluster of the landscape gets the same objects. PAPERS SILENT · real cluster names

namespace nitroso the platform namespace

namespace base-layer namespaces sulfoxide, before any service

namespace (cluster-wide) no namespace

Primordialthe one control cluster, shared by all landscapes

namespace nitroso records; or nitroso-{l} (open, K6)

namespace argocd deploy objects

namespace kargo: nitroso promotion objects

Tap an object to see its source repo and file.

All pages