Deployment · Diagram Design

Checkout service, production

Edge caching absorbs the reads, three ingress-fronted API pods handle the writes, and a single Postgres primary streams WAL to a standby in the data zone.

Checkout service, production Deployment diagram placing the checkout service across an edge CDN zone, a production Kubernetes zone running ingress and API pods, and a data zone with a primary Postgres instance replicating to a standby. EDGE PROD / EU-WEST-1 DATA HTTPS:443 HTTP:8080 TLS:5432 WAL STREAM CDN Cloudflare edge-cache v1.2 POD ingress nginx-ingress v1.11 POD x3 app checkout-api v2.4.1 sidecar-otel v0.9 MANAGED rds-primary postgres 16.2 MANAGED rds-standby postgres 16.2 LEGEND Focal / primary Infra node Store / standby Managed / CDN Cross-zone Replication In-zone call

THE HEADLINE

One primary, one point of failure

Every write lands on `rds-primary`. The standby streams WAL continuously but isn't hot — a primary failure is the one scenario this topology hasn't automated away yet.

Three replicas behind one ingress

  • `checkout-api` runs 3 pod replicas
  • `sidecar-otel` ships traces from the same pod
  • nginx-ingress fans requests across all three

Edge absorbs static reads

Cloudflare's edge cache serves cached responses directly; only cache misses and writes reach the ingress zone.