Fixes for register_dns

[detiber]

• no longer attempt to register the same name multiple times if multiple roles on the same host are configured with the same hostname
• update the load balancer dns in the same nsupdate as the host dns entries

addresses the following error when running vagrant-openshift based devenvs:

    err: /Stage[main]/Openshift_origin::Register_dns/Exec[Register openshift.ec2.internal]/returns: change from notrun to 0 failed:   (

      echo server 10.33.160.203
      echo update delete openshift.example.com A
      echo update add openshift.example.com 180 A 10.33.160.203

      echo server 10.33.160.203
      echo update delete openshift.example.com A
      echo update add openshift.example.com 180 A 10.33.160.203

      echo server 10.33.160.203
      echo update delete openshift.example.com A
      echo update add openshift.example.com 180 A 10.33.160.203

      echo server 10.33.160.203
      echo update delete openshift.example.com A
      echo update add openshift.example.com 180 A 10.33.160.203

      echo server 10.33.160.203
      echo update delete openshift.example.com A
      echo update add openshift.example.com 180 A 10.33.160.203

      echo send
    )| nsupdate -y HMAC-MD5:example.com:slPkQt2ESB7R/dMs50WoXn5gGltjUIhFm83tJfHpGwF5Qha7OTsnalXk28de3Ov/ZpByw8Y+2gUyjahHm9eKIg==
     returned 1 instead of one of [0] at /etc/puppet/modules/openshift_origin/manifests/register_dns.pp:32

other resource ordering fixes:

• avoid restorecon for nameserver if not needed
• ensure that nameserver role is applied before other roles on co-located hosts
• ensure that datastore and activemq roles are applied before the broker role on a co-located host
• contain classes where needed using the anchor pattern

[detiber]

[test]

[openshift-bot]

Origin Test Results: FAILURE (https://ci.openshift.redhat.com/jenkins/job/test_pull_requests/3078/)

[openshift-bot]

Evaluated for origin up to 1952c1a6b860c53561faaa415baa9ced9dfa88ef