kube-dns doesn't expose service scoped dns names for pod IPs

[pmalek]

Problem statement

As described in https://kubernetes.io/docs/concepts/services-networking/dns-pod-service/#a-aaaa-records-1:

Any Pods exposed by a Service have the following DNS resolution available:
pod-ipv4-address.service-name.my-namespace.svc.cluster-domain.example

This is not the case for kube-dns. We're using (generating) those dns names from EndpointSlices for inter cluster communication between services. It does work with coredns.

This is relevant for a number of users since GKE by default uses kube-dns.

Usage of service scoped dns names is preferred because it allows for wildcard certificates like *.service-name.my-namespace.svc.cluster-domain.example to be used.

Proposed solution

Serve service scoped dns names.

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten

[programmer04]

/remove-lifecycle rotten

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[pmalek]

/remove-lifecycle stale

[Michcioperz]

Seems like a bug in the documentation, as according to the DNS specification mentioned in the section linked in the issue, section 2.4.1, it's only provided for headless services (clusterIP: None) and with pod hostnames rather than IP addresses as the subdomain.

[davidxia]

As @Michcioperz brought up, I wonder if kube-dns is supposed to provide these A records or if it's a bug in the docs. Perhaps a maintainer can comment if kube-dns will ever provide this functionality?

In the meantime, I made a PR to update the docs.

[programmer04]

I'm in favor of changing the behavior to fulfill what OP described, because

• CoreDNS that is popular does it
• GKE Autopilot uses Cloud DNS which also behaves the same as CoreDNS

thus we're currently in Hyrum's Law to expect this behavior from the K8s cluster as a whole (no matter the underlying DNS and official specs, etc.)

[Michcioperz]

Can you provide a repro Deployment+Service where CoreDNS and Cloud DNS do what OP described? I'm pretty confident that, like I said, Cloud DNS only does it for headless services.

[programmer04]

I made tests and support for the availability of DNS resolution like e.g. 10-96-2-4.example-service.default.svc for Pods exposed by a Service depends on a DNS provider and it looks like that

	Cloud DNS	Core DNS	kube-dns
ClusterIP	❌	✔️	❌
Headless	✔️	✔️	❌

tested with the below configuration

apiVersion: apps/v1
kind: Deployment
metadata:
  name: example-deployment
  labels:
    app: example-app
spec:
  replicas: 1
  selector:
    matchLabels:
      app: example-app
  template:
    metadata:
      labels:
        app: example-app
    spec:
      containers:
        - name: example-container
          image: nginx:latest
          ports:
            - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: example-service
spec:
  clusterIP: None # To comment/un-comment
  selector:
    app: example-app
  ports:
    - protocol: TCP
      port: 80
      targetPort: 80