A service-to-service call graph represents programs as nodes and network calls as directed edges. It exposes system boundaries before function-level reading begins. section 9.0 section 9.1
Build the graph from evidence distributed across the repository and deployment configuration:
- List long-running programs, containers, databases, queues, workers, and external services as nodes. section 9.0 section 9.1
- Draw edges for HTTP calls, SQL queries, queued jobs, webhooks, and other wires between programs. section 9.1
- Read Docker Compose, Procfiles, Kubernetes manifests, environment variables, SDK initialization, and Terraform state to find hidden dependencies. section 9.0
- Label edges as internal or public and synchronous or asynchronous to expose latency, uptime, delivery, and silent-failure risks. section 9.1
A booking request through the graph
A user device reaches a CDN and load balancer, then a web/API server. That server reads a cache, writes a database, sends asynchronous work through a queue and worker, and may call payment, email, or object-storage services. A request trace shows which nodes the booking actually touches.
