Comment 4 for bug 1593182

Actually, I think the fix for this is to not run yum update with the services running: https://review.openstack.org/#/c/315683/

The package update restarts services automatically, and since at that point we haven't run new puppet to configure the newer versions correctly this sort of breakage is unavoidable. If we instead change the flow to be 1) stop all services 2) yum update 3) openstack undercloud install then that problem should go away (eventually openstack undercloud upgrade will just do those steps for you).

Note that this is the flow I was testing in my local undercloud upgrade test job because otherwise the undercloud upgrade was taking upwards of an hour (sometimes two!) due to services breaking. With stopping the services it takes more in the range of 20 minutes start to finish.