Once upon a time, Miroslav Lichvar <mlichvar@xxxxxxxxxx> said: > On Tue, Jan 05, 2021 at 08:30:39AM -0600, Chris Adams wrote: > > Once upon a time, Zbigniew Jędrzejewski-Szmek <zbyszek@xxxxxxxxx> said: > > > #2517 F34 Change: ntp replacement > > > https://pagure.io/fesco/issue/2517 > > > APPROVED (+5, 0, 0) > > > > I changed my stratum 1 server from ntpd to chronyd+ntp-refclock... had a > > few small bumps to figure out (getting the NTP, PPS, and chrony services > > correct), but it is working. > > If the bumps were due to insufficient documentation (e.g. the > ntp-refclock man page), would you have any suggestions on how to > improve it, or submit a patch? Since I'm using a refclock that doesn't handle PPS, I have chronyd configured to use /dev/pps0. In that case, the example systemd unit in the ntp-refclock man page can't work, because chronyd expects /dev/pps0 to exist when it starts (or it'll just exit). The ordering has to be PPS setup, then chronyd, then ntp-refclock. I did this by making a separate service for PPS, pps-setup@.service: ***** [Unit] Description=PPS serial setup Before=chronyd.service [Service] Restart=on-failure ExecStartPre=/usr/bin/setserial /dev/%I low_latency ExecStart=/sbin/ldattach -d 18 /dev/%I [Install] WantedBy=multi-user.target ***** then these udev rules for my device (USB serial adapter for the data stream, ttyS0 for the PPS ticks): ***** # - TSIP serial stream over USB KERNEL=="ttyUSB*", ATTRS{interface}=="DLP2232M", ATTRS{bInterfaceNumber}=="01", SYMLINK+="palisade0", MODE:="0660", GROUP:="ntp-refclock", RUN+="/usr/bin/chcon -t clock_device_t /dev/$name" # - PPS over serial SUBSYSTEM=="pps", MODE:="0660", GROUP:="chrony" KERNEL=="ttyS0", TAG+="systemd", ENV{SYSTEMD_WANTS}="pps-setup@%k.service" ***** and this service for ntp-refclock: ***** [Unit] Description=GPS Resolution clock After=chronyd.service BindsTo=chronyd.service [Service] ExecStart=/usr/sbin/ntp-refclock -s /var/run/chrony-GPS.sock 127.127.29.0 mode 5 time1 0.060 flag2 1 RestartOn=failure RestartSec=5s [Install] WantedBy=multi-user.target ***** It would be nicer if ntp-refclock just included the systemd unit and used a config to get the options (e.g. /etc/sysconfig/ntp-refclock), and maybe even the PPS setup unit. Obviously the udev rules are device specific, but it might be good to include an example in the docs directory. > > It might be useful for others to document how I have it working - where > > would be a good place for that? Submit a README example or something? > > Is it meant to be a description of an alternative to switching to > ntpsec wrt to the ntp retirement? Fedora-specific information can be > included in the package's documentation, but I'm not sure how many > people will look for that. Generally applicable information would be > best upstream. ntpsec either has or plans to drop the refclock I use. I picked up a used Trimble Resolution T a while back, and running it in TSIP mode works better than NMEA mode, but uses the Palisade refclock. It works fine, but it seems the ntpsec goal is to support as little as possible. > If you didn't use gpsd because you hit a bug there, it would be nice > to send a bugreport upstream. A debug output (-D 9 option) might be > enough to get it fixed. Last I looked, gpsd operates Trimble TSIP devices differently than the ntpd refclock code. The Resolution T is designed to be a timing device, assuming it's in a fixed position, and the ntpd Palisade refclock code puts it in "0-D" mode to support that. Once it gets a position fix, it can tick accurate time with only a single satellite lock (which is nice for my setup since I don't have the best antenna position). Since gpsd is designed first for positioning, it sets TSIP devices to a regular "3-D" mode, where they'll stop ticking if they don't have an active position fix (at least 3 satellites in view). That makes sense and I wouldn't call it a bug in gpsd. -- Chris Adams <linux@xxxxxxxxxxx> _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
