On Fri, 8 Dec 2006, Arthur Konovalov wrote:

> Hi!
>
> Noticed periodical messages in logfile:
>
> Dec  8 11:33:38 time ntpd[7000]: kernel time sync disabled 2307
> Dec  8 11:33:55 time ntpd[7000]: kernel time sync enabled 2107
> Dec  8 11:42:56 time ntpd[7000]: kernel time sync disabled 2307
> Dec  8 11:43:13 time ntpd[7000]: kernel time sync enabled 2107
> Dec  8 11:48:21 time ntpd[7000]: kernel time sync disabled 2307
> Dec  8 11:48:36 time ntpd[7000]: kernel time sync enabled 2107
>
> From status bits I understand that it means "PPS signal jitter exceed",
> but why?
> Is it normal (safe)? How to found reason?
>
> Using Garmin GPS18 LVC, FreeBSD 4.2, ntpd 4.2.0-a
>
> time# ntpq -p
>     remote           refid      st t when poll reach   delay   offset
> jitter
> ==============================================================================
> *GPS_NMEA(1)     .PPS.            0 l    6   16  377    0.000    0.001
> 0.004
>
>
> Regards,
> AK
>
> _______________________________________________
> timekeepers mailing list
> [email protected]
> https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
>

Hello Arthur,

I have no experience at all with stratum 0 devices like GPSs.

But as my two cents, here is how I understand it :

kernel time sync disabled: ntpd doesn't sync kernel time anymore.
kernel time sync enabled: ntpd is back to syncing kernel time.

I had repetitive kernel time sync outputs in my logs while testing 
with bad source servers that returned too much offset for ntpd to keep 
syncing kernel time.

So, as far I understand it, it doesn't mean your GPS setup is wrong. It 
just means that ntpd quit syncing the kernel time for a while.

I believe it is not fatal to loose kernel sync once in a while. It mostly 
depends on how long you loose kernel sync for. Say loosing kernel sync 
for a few seconds shouldn't hurt much the reliability of your time server 
unless ntpd quit serving requests when that happens, which I am not sure 
of...

During those seconds, the kernel just keeps time on 
itself, which should be fine.

My main ntp server ALWAYS loose kernel sync for a few seconds when 
restarted, after connecting to stratum 1 servers. Before connecting to the 
stratum 1 servers it took its time from the local clock which was too far 
away from the average result returned by the stratum 1 servers. (I use 
iburst on my ISP servers)

Also, since you have no backup stratum 1 servers for ntpd to rely on. Ntpd 
should loose kernel sync at the smallest transitory glitch on the link 
between the GPS and the computer or the GPS device itself. With backup 
stratum 1 servers configured, ntpd would rely on them to keep kernel sync 
(no more loosing kernel sync) when a small GPS glitch occurs.

Others with deeper knowledge of ntpd can correct me if I am wrong and 
they can also be more specific. This was just my two cents, hope it 
helped.

I am quite positive although than adding backup stratum one servers to 
your config would help you, unless your onboard oscillator generating 
the interrupts for the "loop" is defective, of course ;-)

Again, others are invited to correct me if I am wrong !

Louis
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to