Description of problem:

I'm filing a bug based on the feedback in bz 1900500 .
In bz 1900500 we observed that neutron detects hypervisor name incorrectly
and that result in issue with syncing resource provider record.

The current inconsistency is caused by the following facts.

- socket.gethostname() returns short name

- canonical name is FQDN because of records in /etc/hosts\
~~~
172.17.1.17 compute-0.redhat.local compute-0
~~~

Feedback from nova's perspective is that we should not use FQDN but should use
a short name as canonical name because of nova's expectation.
Note that it is required by nova to have a unique short name because that is
used to determine service host name (*1), thus short name is always unique.

(*1)
In TripleO setup this is not true, though, since we override host parameter
using FQDN and service host record doesn't rely on socket.gethostname().

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


How reproducible:
Always

Steps to Reproduce:
1. Deploy overcloud
2. Check canonical name


Actual results:
Canonical name is FQDN

Expected results:
Canonical name is a short name

Additional info: