Description of problem:

If default port for DNS in OpenShift master is changed, the Kubernetes DNS service exposes incorrect port for DNS.

Version-Release number of selected component (if applicable):
3.1.0.4

How reproducible:
Very

Steps to Reproduce:
1. Edit /etc/origin/master/master-config.yaml

2. Change value for dnsConfig: bindAddress: from 0.0.0.0:53 to (say) 0.0.0.0:8053

3. restart atomic-openshift-master


Actual results:

# oc get svc kubernetes
NAME         CLUSTER_IP   EXTERNAL_IP   PORT(S)                 SELECTOR   AGE
kubernetes   172.30.0.1   <none>        443/TCP,8053/UDP,8053/TCP   <none>     22d

Note use of port 8053 which is useless for most DNS clients

S2I and other pods will hang and terminate with error

dial tcp: i/o timeout

Expected results:

# oc get svc kubernetes
NAME         CLUSTER_IP   EXTERNAL_IP   PORT(S)                 SELECTOR   AGE
kubernetes   172.30.0.1   <none>        443/TCP,53/UDP,53/TCP   <none>     22d

Note use of port 53 which is required by libresolv and most DNS clients

S2I processes should run to completion

Additional info:

The following script can be used after restarting the master to fix the Kubernetes service definition

#!/bin/bash --

oc get svc kubernetes -n default -o yaml | \
sed -e '/- name: dns/,+1s/port: [0-9][0-9]*$/port: 53/' | \
oc replace -f -