[ntp:questions] NTP+GPS gone haywire
David Lord
snews at lordynet.org
Sun Jan 29 01:27:22 UTC 2012
A C wrote:
> I finally rebuilt my NetBSD box after having lost the partition table on
> the disk. I installed 4.2.7-p236 again and everything seemed fine all
> day yesterday. Suddenly I get this:
>
>> $ ntpq -pn
>> remote refid st t when poll reach delay
>> offset jitter
>> ==============================================================================
>>
>> x127.127.22.0 .PPS. 0 l 2 16 377 0.000 -244.18
>> 351.567
>> 127.127.28.0 .GPSD. 0 l 65 128 377 0.000 -131866
>> 2837.34
>> 74.118.152.85 69.36.224.15 2 u 185 512 377 42.681 -128790
>> 9459.57
>> 64.16.211.38 142.3.100.15 3 u 328 512 377 86.417 -122725
>> 6592.50
>> 173.244.211.10 131.107.13.100 2 u 196 512 377 55.746 -123465
>> 6503.82
>> 130.207.165.28 130.207.244.240 2 u 114 512 377 78.525 -131743
>> 11757.9
>> 131.144.4.10 130.207.244.240 2 u 488 512 377 87.644 -129888
>> 11647.1
>
> This happened sometime late last night (I'll have to look at the peer
> and loop files to see when). No cron jobs fired off during this period,
> they had already run yesterday afternoon. Before this I had offsets for
> the remote servers in the few milliseconds range and an offset on the
> PPS signal of only a couple microseconds.
>
> Anyone have any ideas? Note that the GPSD refclock is set as "noselect"
> currently. I'm trying to test the static navigation feature of the GPS
> to see if the time offsets calm down on that. PPS is supplied by the
> kernel itself.
I had problems when I tried 4.2.7-* recently so I backed out to
either 4.2.6p3 or p5. I was trying 4.2.7.p98 without problem and
also a slightly later version but 4.2.7p239 wasn't usable and I'm
not willing to experiment too much as it might affect my online
servers.
I don't use gpsd either.
Sure GPS (but no different from Garmin 18x-LVC).
NetBSD 5.1_stable, ntp-4.2.6p5
PPSb offset -0.001 jitter 0.003
but offset can increase to 30 usec during cron jobs at 02:00
and 03:00 but is then quickly back to < 5 us.
David
More information about the questions
mailing list