GET /api/orders¶

%% API request path through a Kubernetes cluster.
%% Strip the scenario block and this is ordinary Mermaid.
flowchart LR
  client[External Client]
  lb[(Cloud Load Balancer)]

  subgraph cluster[Kubernetes Cluster]
    ing[Ingress Controller]

    subgraph cp[control plane]
      api[kube-apiserver]
      etcd[(etcd)]
    end

    subgraph ns[namespace: prod]
      svc[ClusterIP Service]
      pod1[Pod A]
      pod2[Pod B]
    end
  end

  client --> lb
  lb --> ing
  ing --> svc
  svc --> pod1
  svc --> pod2
  api --> etcd

view podA "Inside Pod A" from "../pod-a.dgm"

interact {
  %% Pod A is a door: clicking it opens the diagram of what happens inside.
  click pod1 -> view podA { label: "Zoom into Pod A" }

  %% The control plane is not on the request path, so it stays out of the
  %% way until someone asks for it.
  click cluster -> reveal cp

  %% Pod B is the endpoint that was not chosen — jump to the step that says why.
  click pod2 -> step balance
}

scenario "GET /api/orders" { speed: 1.0, loop: true }

  step edge "Request reaches the load balancer" {
    flow client -> lb { label: "GET /api/orders", dur: 700ms }
    highlight lb { style: active }
  }

  step terminate "TLS terminates at the ingress controller" {
    flow lb -> ing { label: "HTTP/1.1", dur: 600ms }
    highlight ing { style: active }
  }

  step route "Ingress rule matches host and path" {
    note ing "host: api.example.com\npath: /api/*"
    flow ing -> svc { dur: 500ms }
  }

  step balance "kube-proxy picks a healthy endpoint" {
    flow svc -> pod1 { label: "10.1.2.3:8080", dur: 600ms }
    dim pod2
  }

  step work "Pod A handles the request" {
    highlight pod1 { style: busy, dur: 900ms }
    note pod1 "200 OK in 14ms"
  }

  step respond "Response travels back to the client" {
    flow pod1 -> svc -> ing -> lb -> client {
      label: "200 OK",
      dur: 1400ms,
      style: response
    }
  }