Bug 1779182

As mentioned in https://bugzilla.redhat.com/show_bug.cgi?id=1770030#c1 host-deploy doesn't alter hostname at all. Dominik, could you please take a look if this is not some new NM on RHEL 8 feature?

According to https://docs.fedoraproject.org/en-US/Fedora/18/html/System_Administrators_Guide/s1_Using_Hostnamectl.html transient hostname is maintained by kernel, so if you should make it persistent by calling hostnamectl, no? Or am I missing something?

(In reply to Martin Perina from comment #6)
> According to
> https://docs.fedoraproject.org/en-US/Fedora/18/html/
> System_Administrators_Guide/s1_Using_Hostnamectl.html transient hostname is
> maintained by kernel, so if you should make it persistent by calling
> hostnamectl, no? Or am I missing something?

We never touch the hostname with hostnamectl. In rhel7 we didn't saw such behavior at all. hostname was persisted all time without doing anything specific.
It seems to be some difference between rhel 7 and rhel8

Looks similar to bug 1419906 .

Hi,

initially the system doesn't have a hostname. Then we get a connection
on enp1s0f0 and the hostname is updated via reverse lookup on enp1s0f0:

 07:15:25 <info>  [1575375325.8736] device (enp1s0f0): state change: secondaries -> activated (reason 'none', sys-iface-state: 'managed')
 07:15:25 <info>  [1575375325.8748] policy: set 'enp1s0f0' (enp1s0f0) as default for IPv4 routing and DNS
 07:15:25 <info>  [1575375325.9400] device (enp1s0f0): Activation: successful, device activated.
 07:15:25 <info>  [1575375325.9904] policy: set-hostname: set hostname to 'vega08.qa.lab.tlv.redhat.com' (from address lookup)

Later, someone issues a reload of connections and enp1s0f0 becomes
unmanaged. Since there is no longer an active connection on enp1s0f0,
the hostname that was got via reverse lookup on the interface is
withdrawn:

 07:59:43 <info>  [1575377983.1764] audit: op="connections-reload" pid=4763 uid=0 result="success"
 07:59:43 <info>  [1575377983.3085] audit: op="connections-load" args="/etc/sysconfig/network-scripts/ifcfg-enp1s0f0" pid=4864 uid=0 result="success"
 ...
 08:02:51 <info>  [1575378171.1263] policy: set-hostname: set hostname to 'localhost.localdomain' (no IP config)
 08:02:51 <info>  [1575378171.1621] device (enp1s0f0): state change: activated -> unmanaged (reason 'unmanaged', sys-iface-state: 'managed')

I don't think this behavior changed from RHEL 7.

Benjamino, the RHEL8 behavior is different from RHEL7 behavior for sure.
Do you consider this as a bug, or is the changed behavior acceptable for you?

(In reply to Dominik Holler from comment #12)
> Benjamino, the RHEL8 behavior is different from RHEL7 behavior for sure.
> Do you consider this as a bug, or is the changed behavior acceptable for you?

Do you have logs from a RHEL7 system? I did a quick test on RHEL 7.7 and the NM behavior (under the same conditions, i.e. the device that has network connectivity gets NM_CONTROLLED=no) is the same.

Hello, I would like to investigate on Anaconda side on possible change of behaviour which seems to happen based on different content of /etc/hostname mentioned in comment #14.

Could you please post a link to rhv installation image (iso) that would reproduce the issue and to another one (is there any such RHEL 8 ?) in which the behaviour is same as expected? (Also any related configuration, eg kickstart etc..)