This directory holds an example of how to run SkyDNS in a Kubernetes cluster.
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. This maps well to DNS, which has a long history of clients that, on purpose or on accident, do not respect DNS TTLs.
The DNS server itself runs as a Kubernetes Service. This gives it a stable IP
address. When you run the SkyDNS service, you can assign a static IP to use for
the Service. For example, if you assign DNS_SERVER_IP (see below) as
10.0.0.10, you can configure your docker daemon with the flag --dns 10.0.0.10.
Of course, giving services a name is just half of the problem - DNS names need a
domain also. This implementation uses the variable DNS_DOMAIN (see below).
You can configure your docker daemon with the flag --dns-search.
The first thing you have to do is substitute the variables into the
configuration. You can then feed the result into kubectl.
DNS_SERVER_IP=10.0.0.10
DNS_DOMAIN=kubernetes.local
DNS_REPLICAS=2
sed -e "s/{DNS_DOMAIN}/$DNS_DOMAIN/g" \
-e "s/{DNS_REPLICAS}/$DNS_REPLICAS/g" \
./contrib/dns/skydns-rc.yaml.in \
| ./cluster/kubectl.sh create -f -
sed -e "s/{DNS_SERVER_IP}/$DNS_SERVER_IP/g" \
./contrib/dns/skydns-svc.yaml.in \
| ./cluster/kubectl.sh create -f -SkyDNS depends on etcd, but it doesn't really need what etcd offers when in
Kubernetes mode. SkyDNS finds the Kubernetes master through the
kubernetes-ro service, and pulls service info from it, essentially using
etcd as a cache. For simplicity, we run etcd and SkyDNS together in a pod,
without linking the etcd instances into a cluster.
DNS resolution does not work from nodes directly, but it DOES work for containers. As best I can figure out, this is some oddity around DNAT and localhost in the kernel. I think I have a workaround, but it's not quite baked as of the this writing (11/6/2014).