One of the issues I had in debugging this was that fake-hwclock does not appear to log what it does on boot.
Timesync does store the date/time in /var/lib/systemd/timesync/clock (https://wiki.archlinux.org/index.php/systemd-timesyncd) and the rate it does that is configurable.
Yes, timesync does need a network to sync, but I think on boot it initially loads the stored time if it is the newest time it has. This is no different to fake-hwclock. However, I found fake-hwclock to unreliably save the current datetime at shutdown.
I was only suggesting the purge if the daemon method was selected i.e. as part of the switch to the daemon method.
Yes it needs to be the only timesync present, but again, this assumes the time control is being handled by the DietPi config. If anyone fiddles with that they deserve issues!
Code: Select all
$ timedatectl show-timesync --all
That item on the RPi forum is quite old - systemd has moved on. Of note is that fake-hwclock does not set the time-sync.target (renamed to time-set in Buster it seems) so services that rely on time, cannot use that to control startup. That it only needs to complete before sysinit.target also implies fake-hwclock may start later.
All I can say is that since purging fake-hwclock and reducing the timesyncd poll time to 60s, I have had far fewer issues with time at startup!