Well. I sampled my 4.2.4p3 server continuously (As fast as possible.
About every 30ms, on average) from 15:52:51 PST to 16:05:53 PST, and at
one-second intervals thereafter until about 17:08:01. (My server, my
LAN, why not?) I have some 90MB of data to sift through. Some
interesting observations. All times are from the clock= variable.
I installed the leapsecond file about two weeks ago, and was showing
tai=33 in the rv command.
This changed to tai=34 at 16:00:30.267. It was also showing leap=01.
This changed to leap=00 at 16:36:04. It's not clear when/whether the
leap was applied to the host clock (unless it just played dead during
that time; I can't tell--what happens with a negative leapsecond?).
My refclock is a DeLorme Tripmate GPS receiver, without PPS (I dare
someone to find it on-chip and bring it out to a pin) and ntp never
chose another server to sync to (I use about 15) during this time, and
never reported an offset more than 5ms for it.
At 16:01:07.715 I polled another host and got an offset of 1002.34 ms.
Aha! This guy is showing a time 1s ahead of me. He didn't leap, I
did. For a while other polls were within reason, and a few more +1s
showed up (one is still there at this writing, two hours later. It
belongs to .ibone.comcast.net).
===
I was also watching a friend's server. Spectracom 8170 WWVB clock (I
don't think it groks leap seconds, but of course WWVB does). No
leapsecond file, no stratum 1s to get leapinfo from. At about 16:02:44
one of the ISPs servers returned a -987.18ms offset. At about 16:06:47
it noticed its reflock offset -999.88ms (not bad for a clock of that
vintage to sync up again) and chose another server that hadn't leapt
yet. At 16:02:26 a third -994.20ms offset was recorded. No more peers.
At 16:19:40 ntp gave up and started over. At 16:25:12 it selected the
ISP's stratum 2 to sync with. At 16:27:07 it returned to using the
refclock. Crisis over. Nothing to see here, move along.
===
It's interesting to note that the human-readable times reported from the
clock= variable are local to the client; the WWVB server is in a
timezone two hours ahead of me.
--
Jeff Woolsey {woolsey,[email protected],[email protected]
Nature abhors a straight antenna, a clean lens, and unused storage capacity.
"Delete! Delete! OK!" -Dr. Bronner on disk space management
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers