kube-dns

DNS in Kubernetes

Kubernetes offers a DNS cluster addon, which most of the supported environments enable by default. We use SkyDNS as the DNS server, with some custom logic to slave it to the kubernetes API server.

What things get DNS names?

The only objects to which we are assigning DNS names are Services. Every Kubernetes Service is assigned a virtual IP address which is stable as long as the Service exists (as compared to Pod IPs which can change over time due to crashes or scheduling changes). This maps well to DNS, which has a long history of clients that, on purpose or on accident, do not respect DNS TTLs (see previous remark about Pod IPs changing).

Where does resolution work?

Kubernetes Service DNS names can be resolved using standard methods (e.g. gethostbyname) inside any pod, except pods which have the hostNetwork field set to true.

Supported DNS schema

The following sections detail the supported record types and layout that is supported. Any other layout or names or queries that happen to work are considered implementation details and are subject to change without warning.

Services

A records

"Normal" (not headless) Services are assigned a DNS A record for a name of the form my-svc.my-namespace.svc.cluster.local. This resolves to the cluster IP of the Service.

"Headless" (without a cluster IP) Services are also assigned a DNS A record for a name of the form my-svc.my-namespace.svc.cluster.local. Unlike normal Services, this resolves to the set of IPs of the pods selected by the Service. Clients are expected to consume the set or else use standard round-robin selection from the set.

SRV records

SRV Records are created for named ports that are part of normal or Headless Services. For each named port, the SRV record would have the form _my-port-name._my-port-protocol.my-svc.my-namespace.svc.cluster.local. For a regular service, this resolves to the port number and the CNAME: my-svc.my-namespace.svc.cluster.local. For a headless service, this resolves to multiple answers, one for each pod that is backing the service, and contains the port number and a CNAME of the pod of the form auto-generated-name.my-svc.my-namespace.svc.cluster.local.

Backwards compatibility

Previous versions of kube-dns made names of the for my-svc.my-namespace.cluster.local (the 'svc' level was added later). This is no longer supported.

Pods

A Records

When enabled, pods are assigned a DNS A record in the form of pod-ip-address.my-namespace.pod.cluster.local.

For example, a pod with ip 1.2.3.4 in the namespace default with a dns name of cluster.local would have an entry: 1-2-3-4.default.pod.cluster.local.

####A Records and hostname based on Pod's hostname and subdomain fields Currently when a pod is created, its hostname is the Pod's metadata.name value.

With v1.2, users can specify a Pod annotation, pod.beta.kubernetes.io/hostname, to specify what the Pod's hostname should be. If the annotation is specified, the annotation value takes precendence over the Pod's name, to be the hostname of the pod. For example, given a Pod with annotation pod.beta.kubernetes.io/hostname: my-pod-name, the Pod will have its hostname set to "my-pod-name".

With v1.3, the PodSpec has a hostname field, which can be used to specify the Pod's hostname. This field value takes precedence over the pod.beta.kubernetes.io/hostname annotation value.

v1.2 introduces a beta feature where the user can specify a Pod annotation, pod.beta.kubernetes.io/subdomain, to specify what the Pod's subdomain should be. If the annotation is specified, the fully qualified Pod hostname will be "...svc.". For example, given a Pod with the hostname annotation set to "foo", and the subdomain annotation set to "bar", in namespace "my-namespace", the pod will set its own FQDN as "foo.bar.my-namespace.svc.cluster.local"

With v1.3, the PodSpec has a subdomain field, which can be used to specify the Pod's subdomain. This field value takes precedence over the pod.beta.kubernetes.io/subdomain annotation value.

Example:

apiVersion: v1
kind: Pod
metadata:
  name: busybox
  namespace: default
spec:
  hostname: busybox-1
  subdomain: default
  containers:
  - image: busybox
    command:
      - sleep
      - "3600"
    name: busybox

If there exists a headless service in the same namespace as the pod and with the same name as the subdomain, the cluster's KubeDNS Server will also return an A record for the Pod's fully qualified hostname. Given a Pod with the hostname set to "foo" and the subdomain set to "bar", and a headless Service named "bar" in the same namespace, the pod will see it's own FQDN as "foo.bar.my-namespace.svc.cluster.local". DNS will serve an A record at that name, pointing to the Pod's IP.

With v1.2, the Endpoints object also has a new annotation endpoints.beta.kubernetes.io/hostnames-map. Its value is the json representation of map[string(IP)][endpoints.HostRecord], for example: '{"10.245.1.6":{HostName: "my-webserver"}}'. If the Endpoints are for a headless service, then A records will be created with the format ...svc. For the example json, if endpoints are for a headless service named "bar", and one of the endpoints has IP "10.245.1.6", then a A record will be created with the name "my-webserver.bar.my-namespace.svc.cluster.local" and the A record lookup would return "10.245.1.6". This endpoints annotation generally does not need to be specified by end-users, but can used by the internal service controller to deliver the aforementioned feature.

With v1.3, The Endpoints object can specify the hostname for any endpoint, along with its IP. The hostname field takes precedence over the hostname value that might have been specified via the endpoints.beta.kubernetes.io/hostnames-map annotation.

With v1.3, the following annotations are deprecated: pod.beta.kubernetes.io/hostname, pod.beta.kubernetes.io/subdomain, endpoints.beta.kubernetes.io/hostnames-map

How do I find the DNS server?

The DNS server itself runs as a Kubernetes Service. This gives it a stable IP address. When you run the SkyDNS service, you want to assign a static IP to use for the Service. For example, if you assign the DNS Service IP as 10.0.0.10, you can configure your kubelet to pass that on to each container as a DNS server.

Of course, giving services a name is just half of the problem - DNS names need a domain also. This implementation uses a configurable local domain, which can also be passed to containers by kubelet as a DNS search suffix.

How do I configure it?

The easiest way to use DNS is to use a supported kubernetes cluster setup, which should have the required logic to read some config variables and plumb them all the way down to kubelet.

Supported environments offer the following config flags, which are used at cluster turn-up to create the SkyDNS pods and configure the kubelets. For example, see cluster/gce/config-default.sh.

ENABLE_CLUSTER_DNS="${KUBE_ENABLE_CLUSTER_DNS:-true}"
DNS_SERVER_IP="10.0.0.10"
DNS_DOMAIN="cluster.local"
DNS_REPLICAS=1

This enables DNS with a DNS Service IP of 10.0.0.10 and a local domain of cluster.local, served by a single copy of SkyDNS.

If you are not using a supported cluster setup, you will have to replicate some of this yourself. First, each kubelet needs to run with the following flags set:

--cluster-dns=<DNS service ip>
--cluster-domain=<default local domain>

Second, you need to start the DNS server ReplicationController and Service. See the example files (ReplicationController and Service), but keep in mind that these are templated for Salt. You will need to replace the {{ <param> }} blocks with your own values for the config variables mentioned above. Other than the templating, these are normal kubernetes objects, and can be instantiated with kubectl create.

How do I test if it is working?

First deploy DNS as described above.

1 Create a simple Pod to use as a test environment.

Create a file named busybox.yaml with the following contents:

apiVersion: v1
kind: Pod
metadata:
  name: busybox
  namespace: default
spec:
  containers:
  - image: busybox
    command:
      - sleep
      - "3600"
    imagePullPolicy: IfNotPresent
    name: busybox
  restartPolicy: Always

Then create a pod using this file:

kubectl create -f busybox.yaml

2 Wait for this pod to go into the running state.

You can get its status with:

kubectl get pods busybox

You should see:

NAME      READY     STATUS    RESTARTS   AGE
busybox   1/1       Running   0          <some-time>

3 Validate DNS works

Once that pod is running, you can exec nslookup in that environment:

kubectl exec busybox -- nslookup kubernetes.default

You should see something like:

Server:    10.0.0.10
Address 1: 10.0.0.10

Name:      kubernetes.default
Address 1: 10.0.0.1

If you see that, DNS is working correctly.

How does it work?

SkyDNS depends on etcd for what to serve, but it doesn't really need all of what etcd offers (at least not in the way we use it). For simplicity, we run etcd and SkyDNS together in a pod, and we do not try to link etcd instances across replicas. A helper container called kube2sky also runs in the pod and acts a bridge between Kubernetes and SkyDNS. It finds the Kubernetes master through the kubernetes service (via environment variables), pulls service info from the master, and writes that to etcd for SkyDNS to find.

Inheriting DNS from the node

When running a pod, kubelet will prepend the cluster DNS server and search paths to the node's own DNS settings. If the node is able to resolve DNS names specific to the larger environment, pods should be able to, also. See "Known issues" below for a caveat.

If you don't want this, or if you want a different DNS config for pods, you can use the kubelet's --resolv-conf flag. Setting it to "" means that pods will not inherit DNS. Setting it to a valid file path means that kubelet will use this file instead of /etc/resolv.conf for DNS inheritance.

Known issues

Kubernetes installs do not configure the nodes' resolv.conf files to use the cluster DNS by default, because that process is inherently distro-specific. This should probably be implemented eventually.

Linux's libc is impossibly stuck (see this bug from 2005) with limits of just 3 DNS nameserver records and 6 DNS search records. Kubernetes needs to consume 1 nameserver record and 3 search records. This means that if a local installation already uses 3 nameservers or uses more than 3 searches, some of those settings will be lost. As a partial workaround, the node can run dnsmasq which will provide more nameserver entries, but not more search entries. You can also use kubelet's --resolv-conf flag.

Making changes

The container containing the kube-dns binary needs to be built for every architecture and pushed to the registry manually whenever the kube-dns binary has code changes. Every significant change to the functionality should result in a bump of the TAG in the Makefile.

Any significant changes to the YAML template for kube-dns should result a bump of the version number for the kube-dns replication controller and well as the version label. This will permit a rolling update of kube-dns.