[ntp:questions] failing-over flaky upstream servers
tik-tok at bodosom.net
Thu Mar 12 14:55:35 UTC 2015
On Thu, Mar 12, 2015 at 8:36 AM, Paul <paul-ntp-questions at lookmumnohands.net
> I would like my ntpd to continue serving time, gracefully choosing
> from the "best" available upstream servers.
> 3. PPS signal derived from GPS. Excellent accuracy, but only
> available say 90% of the day due to satellite visibility.
You should get a better (timing) receiver. I have an 18x-LVC in the
"window" of my interior office in a rehab'ed machine room (they did cut out
some exterior windows) and I get a "valid" signal 97% of the time.
Admittedly it's probably all multipath but it manages sub-millisecond
offsets from more reliable clocks.
How do I best configure ntpd to get the best available time under
> changing circumstances? I also like to be able to "unplug the
> Internet" without ntpd getting upset. If I recall correctly, in order
> for PPS to function it needs a "prefer" peer, and I am unsure what the
> best selection is.
Versions prior to 4.2.8 allowed you to use PPS + LOCAL. This is a
sensible configuration because after you've "initialized" the seconds with
a correct source you do have a "discliplined" system clock -- at least
until the power fails. Similar results can be achieved with a pps kernel
but I suspect the intent there is to have a primary(ish) standard, where a
GPSDO is "ish" compared to your Cesium standard, rather than a simple
receiver. A new GPSDO can be had for much less than a Cesium (Cs) or
Rubidium (Rb) standard if you do want to go that way. Even less from Ebay
although older units like RFTGs and Z38NNs have old receivers that need
better antenna siting.
Another approach is the NIST/USNO ACTS_MODEM driver which I've yet to get
working reliably but that's a personal problem.
In each case you probably want to use orphan mode.
More information about the questions