Due to company organization, customer cannot use the default Nutanix nameservers, so he configured a NMState policy to define custom ones.

NMState works as expected (file /var/run/NetworkManager/resolv.conf correctly updated), but the file /etc/resolv.conf file is not changed. We troubleshooted this issue to the configuration contained in the 00-master and 00-worker MCs:

~~~
# /etc/NetworkManager/conf.d/99-nutanix.conf"
[main]
rc-manager=unmanaged
~~~

(I replicated this with Baremetal IPI with VirtualIPs.)


Customer worked-around this issue by creating a MC that over-writes the file above:

~~~
# /etc/NetworkManager/conf.d/99-nutanix.conf"
[main]
rc-manager=file
~~~

Everything is working as the customer desires, and they are happy.

The ask is: Is this supported?


According to our documentation [1], we can read:

"IMPORTANT
Red Hat supports the Kubernetes NMState Operator in production environments on bare-metal, IBM Power®, IBM Z®, IBM® LinuxONE, VMware vSphere, and OpenStack installations."

[1] https://docs.openshift.com/container-platform/4.14/networking/k8s_nmstate/k8s-nmstate-about-the-k8s-nmstate-operator.html


So, we (myself and the customer TAM, Arne Gogala) just want to be sure that this change does not impact other components.


In addition, I reproduced the NMState configuration with the following results:

4.14.10 == Issue as above.
4.15.20 == Issue as above.
4.16.1  == NO Issue!!!

How can we explain this inconsistent behaviour with 4.16.1 ?

Thank you

    4.14.10 / 4.15.20 - NM/NMState (no more versions were tested)

    Always.

    1. Install Baremetal IPI/Nutanix cluster.
    2. Install NMState operator.
    3. Define policy to change the nameservers.
    

    Nameservers are not changed.

    Nameservers are updated according to NMState policy.