[time-nuts] Holdover, RTC for Pi as NTP GPS source
mlewis000 at rogers.com
Wed Nov 1 17:18:58 EDT 2017
This GPS/Pi server, with possibly a second copy, are intended to be the
sole NTP sources for a local network. No going out to the internet.
My intent was to discipline the RTC, so it's current when needed. If I
had to do a cold boot without GPS available, I'd want the RTC to be current.
I hadn't considered 'just riding through using the NTP estimate of the
current system clock frequency', having assumed that a disciplined RTC
would be a better source.
Hadn't heard about that issue with NTPsec.
I haven't looked at chrony. Something else to read tonight.
On 01/11/2017 4:57 PM, Chris Caudle wrote:
> On Wed, November 1, 2017 3:17 pm, MLewis wrote:
>> hadn't got there yet
> Your RTC is not likely to be tightly synchronized to NTP time, so there is
> a high probability that trying to use RTC as a secondary time source will
> actually make the system worse than just riding through using the NTP
> estimate of the current system clock frequency. If ntpd will even use it
> as a secondary clock, and not reject it because it is too far off from the
> preferred GPS source.
> Is this system not connected to other NTP servers over the network?
> NTPsec does not work very well in the case of only one GPS device
> available to set time, that is documented as one of the use cases which
> NTPsec does not currently handle well. If there are no other servers
> available for comparison you will probably want to use chrony. If there
> are other servers available then trying to switch over to the RTC will
> definitely make your ntp server perform worse than just leaving well
> enough alone.
More information about the time-nuts