Description of problem:

The overcloud director images for Red Hat OpenStack 10 - Maintenance Release 16 contain a hard-coded nameserver in the /etc/resolv.conf.  This presents an issue when deploying an overcloud as it results on delays with ssh connectivity into the overcloud nodes once deployed.  This issue also results in slow dns resolution as the 'correct' dns server(s) are appended to the /etc/resolv.conf during deployment. 


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

rhosp-director-images-10.0-20190103.1.el7ost.noarch


How reproducible: Always


Steps to reproduce below:


--------------------------------------------------------------
[stack@stn3-osp-dir01 debug]$ rpm -qf /usr/share/rhosp-director-images/overcloud-full-10.0-20190103.1.el7ost.tar
rhosp-director-images-10.0-20190103.1.el7ost.noarch


[stack@stn3-osp-dir01 debug]$ tar -C . -xvf /usr/share/rhosp-director-images/overcloud-full-10.0-20190103.1.el7ost.tar
overcloud-full.qcow2
overcloud-full.initrd
overcloud-full.vmlinuz

[stack@stn3-osp-dir01 debug]$ virt-cat -a ./overcloud-full.qcow2 /etc/resolv.conf
# Generated by NetworkManager
nameserver 192.168.122.1
--------------------------------------------------------------


The current workaround to fix this issue is for the manual customization of the image using virt-customize or guestfish to remove that hard-coded dns entry from /etc/resolv.conf.