I am creating this as a child of bug #1655207, which has was a customer escalation that ended up being something we don't want to code into TripleO or OpenStack. Basically the customer put themselves in a catch-22 position by hosting a service (DNS) on the cloud. The DNS service they hosted there was also the one that the undercloud and overcloud was configured to use. As soon as they had a problem in their cloud the DNS service hosted on the cloud stopped working, which then meant the cloud components couldn't do DNS lookups on each other. Bad Stuff Happened.
To reduce this to it's minimal viable expression: The cloud depended on working DNS. The DNS service depended on a working cloud. As soon as either had a problem, both were doomed.
I am not sure where we have a statement of generalized cloud architecture/deployment best practices, but wherever it is a statement to the effect that Thou Shalt Not Do This Thing should be added.
Affected versions: all
Engineering SME: Nate Johnston
*** Bug 1655207 has been marked as a duplicate of this bug. ***
The change has been made to the RHOSP 13 "Networking Guide." Customers can see
this change here: