How to Retrieve All Pod IPs Using Kubernetes Headless Services and DNS

What is a Headless Service?

When a Service is created, Kubernetes can assign it a DNS name. Setting spec.clusterIP to "None" tells Kubernetes not to allocate an IP address, creating a Headless Service that relies on DNS to expose the individual Pod IPs.

Service Types

ClusterIP – default, Service reachable only inside the cluster.

NodePort – exposes the Service on a static port on each node.

LoadBalancer – uses an external cloud load balancer.

By using a Headless Service, the cluster avoids load‑balancing to a single IP and instead creates DNS A/AAAA records for each selected Pod.

DNS Resolution with Headless Services

Kubernetes generates a DNS name for the Service, and other components can resolve it to the set of Pod IPs. The DNS naming follows the pattern <service-name>.<namespace>.svc.cluster.local.

Practical Example

Assume three MongoDB Pods form a replica set. Each Pod is labeled app=mongodb. A Headless Service selects these Pods and provides a DNS name that the application can query.

apiVersion: v1
kind: Service
metadata:
  name: mongodb-headless-service
  namespace: infrastructure
spec:
  selector:
    app: mongodb
  clusterIP: None
  ports:
  - port: 27017
    targetPort: 27017

From a .NET application, the DNS query can be performed with Dns.GetHostAddresses from the System.Net namespace, returning an array of IP addresses that can be used to build a MongoDB connection string.

var addresses = Dns.GetHostAddresses("mongodb-headless-service.infrastructure.svc.cluster.local");
var connectionString = $"mongodb://{string.Join(",", addresses)}";

This approach lets the connection string be generated dynamically, so scaling the database cluster does not require manual updates.

Note: MongoDB already supports DNS‑based service discovery via a Headless Service, eliminating the need to manually construct the connection string.