Service Discovery — Finding Services Dynamically

Dynamically locate healthy service instances without hardcoded IPs — services register themselves, clients or load balancers query the registry.

When to use

• Any microservice environment where instances scale dynamically
• Client-side discovery when you need custom load balancing logic

Tradeoffs

• Registry is a critical dependency — must be HA or all service discovery fails
• DNS TTL caching can cause stale routing to failed instances

Go

func registerService(client *consul.Client, addr string, port int) (string, error) {
    id := fmt.Sprintf("payment-%s-%d", addr, port)
    reg := &consul.AgentServiceRegistration{
        ID:      id,
        Name:    "payment-service",
        Address: addr,
        Port:    port,
        Check: &consul.AgentServiceCheck{
            HTTP:     fmt.Sprintf("http://%s:%d/healthz", addr, port),
            Interval: "10s",
            Timeout:  "2s",
        },
    }
    return id, client.Agent().ServiceRegister(reg)
}

func deregisterService(client *consul.Client, id string) error {
    return client.Agent().ServiceDeregister(id)
}

Python

import consul

def register_service(c: consul.Consul, addr: str, port: int) -> str:
    service_id = f"payment-{addr}-{port}"
    c.agent.service.register(
        name="payment-service",
        service_id=service_id,
        address=addr,
        port=port,
        check=consul.Check.http(
            f"http://{addr}:{port}/healthz",
            interval="10s",
            timeout="2s",
        ),
    )
    return service_id

def deregister_service(c: consul.Consul, service_id: str) -> None:
    c.agent.service.deregister(service_id)

Gotcha: Kubernetes DNS-based service discovery works out of the box for in-cluster traffic. Don't add Consul or etcd unless you need cross-cluster or hybrid-cloud discovery.