SUMMARY In RHEL/Fedora, we use chrony by default instead of ntp/ntpsec. Can we please have the KCM work with Chrony instead, as we don't want to ship ntpd. SOFTWARE/OS VERSIONS Linux/KDE Plasma: Fedora Linux 35 (KDE Plasma) (available in About System) KDE Plasma Version: 5.23.2 KDE Frameworks Version: 5.87 Qt Version: 5.15.2 ADDITIONAL INFORMATION This was originally brought up downstream in: https://pagure.io/fedora-kde/SIG/issue/129
Makes sense.
The KCM calls org.freedesktop.timedate1.SetNTP which as I understand is implemented by systemd-timesyncd. Are you saying this doesn't work/can't be used on Fedora?
We don't have timesyncd enabled by default, no.
> We don't have timesyncd enabled by default, no. Is there a deeper reason for that? Just for my understanding, chrony would be an alternative to timesyncd here?
(In reply to Nicolas Fella from comment #4) > > We don't have timesyncd enabled by default, no. > > Is there a deeper reason for that? > > Just for my understanding, chrony would be an alternative to timesyncd here? We've been using chrony for years, and my understanding is that chrony is considered a better time server+client implementation than timesyncd.
Can you please describe what user-facing problem there currently is? I'm not sure I understand what we are fixing here. On my OpenSUSE Tumbleweed enabling automatic time settings seems to work despite the system apparently using chrony
I guess it's that you can't disable the feature since it only hooks into ntpd. So it's possible to uncheck the checkbox, but the time keeps getting set automatically, because the thing that does it (chrony) doesn't listen to that setting. The path of least resistance would be to just hide the checkbox when ntpd isn't available, though it might be nice to make the checkbox work on chrony too.
There seems to be some confusion here. The KCM uses org.freedesktop.timedate1/systemd-timesyncd if the DBus interface is available, which it is on Fedora(!). ntpd, which is a third option/alternative to chrony/systemd-timesyncd isn't really relevant here
(In reply to Nicolas Fella from comment #8) > There seems to be some confusion here. The KCM uses > org.freedesktop.timedate1/systemd-timesyncd if the DBus interface is > available, which it is on Fedora(!). ntpd, which is a third > option/alternative to chrony/systemd-timesyncd isn't really relevant here It shouldn't be, we don't even have timesyncd enabled and running by default. https://src.fedoraproject.org/rpms/fedora-release/blob/rawhide/f/90-default.preset#_4-5
Running 'qdbus --system org.freedesktop.timedate1' on Fedora 34 (Gnome) shows that is *is* available/getting activated
`timedate1` isn't the same thing as timesyncd. I believe `timedated` is a separate dbus-activated service.
aaah, yes, timedated and timesyncd are indeed separate things. That explains a lot of my confusion. The KCM is calling time*date*d to enable NTP, which then seems to activate the appropriate thing (chrony or systemd-timesyncd). On my OpenSUSE system 'systemctl status systemd-timedated' contains Nov 26 18:23:32 dumbledore systemd-timedated[6796]: chronyd.service: Enabling unit. Nov 26 18:23:32 dumbledore systemd-timedated[6796]: Set NTP to enabled (chronyd.service). On my Manjaro system (which seems to use systemd-timesyncd) it contains Nov 08 22:26:00 madeye systemd-timedated[4532]: Set NTP to enabled (systemd-timesyncd.service). So as far as I can tell things "just work", which brings me back to my question of what user-facing symptoms we are having in Fedora
Honestly, I'm not sure, Nate reported the problem to me in the Fedora KDE tracker, and I forwarded it back up here.
If everything is fine, feel free to close both :)