K3d Bootstrap Cluster

Appearance

ExternalDNS ​

Precondition ​

Cluster has to be deployed with the httpbin sample. Otherwise this demo wouldn't work.

Components ​

ComponentSource
etcdPlain Kubernetes manifest (etcd.yaml) using registry.k8s.io/etcd:3.5.16-0
CoreDNSHelm chart coredns/coredns
ExternalDNSOfficial Helm chart external-dns/external-dns (registry.k8s.io/external-dns/external-dns)

Installation ​

bash
cd examples/external-dns

# With HAProxy Ingress Controller (auto-detected if not set)
HAPROXY_FLAG=Yes bash setup.sh

# With Kong Gateway (Gateway API / HTTPRoute)
KONG_FLAG=Yes bash setup.sh

If neither flag is set, the ingress mode is auto-detected from the cluster.

ExternalDNS is configured with the CoreDNS provider. It watches Ingress resources (HAProxy mode) or HTTPRoutes (Kong mode) and writes DNS records into etcd, which CoreDNS serves.

How it works ​

  1. etcd is deployed as a single-pod Deployment — CoreDNS and ExternalDNS use it as the DNS record store.
  2. CoreDNS is configured with the etcd plugin for the example.com zone and exposed via NodePort 30053.
  3. ExternalDNS watches Ingress/HTTPRoute resources and registers A records in etcd using the Ingress controller's LoadBalancer IP.
  4. HAProxy Ingress Controller must have publishService.enabled: true (already set in manifests/haproxy-helm.yaml) so that the Ingress .status.loadBalancer is populated — ExternalDNS reads that IP as the DNS target.

CoreDNS configuration ​

After installation the generated ConfigMap looks like this (etcd endpoint IP varies):

yaml
apiVersion: v1
kind: ConfigMap
data:
  Corefile: |-
    .:53 {
        etcd example.com {
            stubzones
            path /skydns
            endpoint http://10.43.x.x:2379
        }
        debug
        errors
        health {
            lameduck 5s
        }
        ready
        reload
        loadbalance
    }

Test DNS resolution ​

CoreDNS is exposed on NodePort 30053, which k3d maps to a random host port. Look up the port with:

bash
docker ps --format '{{.Ports}}' | grep 30053
# e.g. 0.0.0.0:49625->30053/udp

Then query directly from your host:

bash
dig @localhost httpbin.example.com -p <PORT>

Expected answer:

;; ANSWER SECTION:
httpbin.example.com.  30  IN  A  172.25.0.4

The IP is the ClusterIP of the HAProxy LoadBalancer service (or Kong Gateway service in Kong mode).

You can also run an in-cluster test:

bash
kubectl run -it --rm --restart=Never --image=infoblox/dnstools:latest dnstools
# Inside the container:
# nslookup httpbin.example.com <coredns-service-ip>