Edit3: Edit1 and Edit2 seems to resolve the issue.
After struggling with a small openstack environment I decided to use a fresh UCS installation with two networks:
192.168.2.0/24 with an internet gateway and
10.0.0.0/24 without internet connection.
Now the UCS should act as an AD-controller in the 10.0.0.0 network.
A virutal machine (Windows 10 1809) gets the IP 10.0.0.22 from the UCS and I configured the UCS-DHCP to also deliver DNS, WINS and routing configuration to the vm. That works fine.
But: The vm can’t join the domain. I spend two days, but I didn’t found any solution. What I figured out, is that UCS changes DNS-records to values, that could be the reason for the join errors. It inserts the IP 192.168.2.1 (the IP of the UCS in the 192.168.2.0-network) into the Forward Lookup Zone, DomainDnsZones, ForestDnsZones, gc._msdcs and emmett (the name of the UCS-server).
nslookup of “mvz.intranet” and “emmett.mvz.intranet” on the vm gives me 10.0.0.1 and 192.168.2.1. The last record is wrong.
In LDAP intranet.mvz:/computers/dc/emmett only has the IP-address 10.0.0.1.
Edit1: Using samba/interfaces/bindonly to disable the 192.168.2.0-network for samba also is not possible because then some linux devices within the 192.168.2.0-network couldn’t join the domain.
Edit2: I removed the “yes” from samba/interfaces/bindonly