Latest snap NTP issues

jimp

Just pushed a fix, gitsync or wait for a new snap dated after this post.

jcyr

Yes, commit 3289b42 fixes the problem but…

Any changes made to the timeserver list requires a reboot. The kill and restart stuff for ntpd in /etc/inc/system.inc does not seem to work!

After such a change /var/etc/ntp.conf is correctly updated, but an 'ntpq -c peers' at the command line shows the old upstream servers.

The -p parameter seems to be missing from the boot time ntpd startup

[2.1-BETA1][root@srvrrouter.dillobits.lan]/root(8): ps -ax | grep ntpd
45218 ?? Ss 0:00.18 /usr/local/bin/ntpd -g -c /var/etc/ntpd.conf

So the /var/run/ntpd.pid file is never created, hence the kill fails, and new ntpd sees that its ports are held (by the old ntpd that wasn't killed) and exits!

jimp

Edit/Save on System > General.
That's all I had to do.

Make sure you gitsync again to pull in my last change, the first commit was bad.

Then ntpq -pn showed all my peers.

jcyr

Err… no

Do an ls on /var/run after a reboot. You'll see there's no valid ntpd.pid there. So the kill code in system.inc after you've recreated the .conf file won't actually kill the running ntpd.

Sooo... after that when you start ntpd again, it finds port 123 in use (since the initial ntpd is still running) and exits. The old ntpd keeps running with the old .conf file.

What you need is:

*** ntpdate_sync_once.sh Sun Feb 10 22:27:54 2013
--- ntpdate_sync_once.sh.orig Sun Feb 10 22:27:05 2013

*** 29,33 ****

if [ -f /var/etc/ntpd.conf ]; then
echo "Starting NTP Daemon." | /usr/bin/logger -t ntp;
! /usr/local/bin/ntpd -g -c /var/etc/ntpd.conf -p /var/run/ntpd.pid
else
echo "NTP configuration file missing, not starting daemon." | /usr/bin/logger -t ntp;
–- 29,33 ----

if [ -f /var/etc/ntpd.conf ]; then
echo "Starting NTP Daemon." | /usr/bin/logger -t ntp;
! /usr/local/bin/ntpd -g -c /var/etc/ntpd.conf
else
echo "NTP configuration file missing, not starting daemon." | /usr/bin/logger -t ntp;

This will cause ntpd to create a proper pid file the first time it is launched at boot time. Without this, either the pid file doesn't exist, or it contains an invalid pid number from previous launches.

grandrivers

also cant add more than one server in general page it gives validation errors

jcyr

With Jimp's latest fix, I add multiple servers (separated with spaces) without any problems. Only problem left is the restart issue which I had to patch locally.

fragged

I just updated to snapshot from 10th Feb and NTP shows as running under services, but under Status - NTP it shows "No peers found, is the ntp service running?.". With the last snapshot I was running (5th Feb) I had issues with NTP crashing "Core dumped" if I had multiple NTP servers set in Setup -> General.

ntpq -p shows:
No association ID's returned

fragged

Updated to Sun Feb 10 22:07:45 EST and NTP seems to work again. I'm going to monitor it and see if it works with multiple peers.

jcyr

Try changing the server list in Setup->General, and see if ntpd picks up the changes.

fragged

@jcyr:

Try changing the server list in Setup->General, and see if ntpd picks up the changes.

I added a third host/pool to NTP server list and it picked up the change immediately. My Windows workstation had issues syncing to pfSense's NTP before I restarted Win32time, but that's working again too.