Re: [chrony-users] Multi-PPS+RTC Setup Questions |
[ Thread Index |
Date Index
| More chrony.tuxfamily.org/chrony-users Archives
]
- To: chrony-users@xxxxxxxxxxxxxxxxxxxx, Miroslav Lichvar <mlichvar@xxxxxxxxxx>
- Subject: Re: [chrony-users] Multi-PPS+RTC Setup Questions
- From: Matt Corallo <ntp-lists@xxxxxxxxxxxxxxx>
- Date: Tue, 29 Sep 2020 10:27:51 -0400
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mattcorallo.com; s=1601388070; t=1601389671; bh=Yw1AP7KE1tk6bMwO36hROfbYrP/qeNQkruZS9+DGDjk=; h=To:References:From:Subject:Date:In-Reply-To:From; b=KlF7JJynY9uoGCj2uuLohGXRM8iMURTYg8v84QNAIGwkN+EKgWXTKYkXa9hToJYTk qiNeijDBHAe7j6WehZ464eROFjZWkA5emZK7QRSPlXP60yMU7hPtU4vhQSKUT3LKh6 XabSzWz1Sdb1psHGRhBcUaHvkUhZX+EuLF0mh8lu4pZeBpDnu7bn9MwuYHGPjrnI6F LXz/vbOWn+rfJ5C3eZSxL1WilcynCZ8cTTpibvDlDkqDZn9N4KKl9rDNNbDFqAgcBk 6LNBIdQ3MS86tKQjqXzdEe26Lt84S0ubTieVLowneVFq2rpY77Sgev9pLXcleBa24c qQnAyEQcUVwXw==
Relevant selectdata on 4.0 (which seems to be working great) is:
P RTC1 N --TR- --TR- 22 1.0 -3326us +4679us (since changed to +/-4ms, not 2ms)
* GPS1 N -P--- -P--- 15 1.0 -22us -8175ns
+ GPS2 N -P--- -P--- 8 2.5 -9093ns -2063ns
Indeed, RTC1 is an external daemon that reads in from an RTC, checks that chrony isn't using it as its timesource and
then roughly disciplines it according to chrony's time, all before feeding samples back into chrony over sock. Indeed,
the two GPS refclocks are of different model, and do adjust slightly differently as their internal RTCs drift, but all
within noise, or only slightly outside of it (at least once I compensate for them being persistently off by 30us, which,
uhhh...wtf u-blox?).
With 3.4, I had to add "combinelimit 100" to get chrony to look at both refclocks instead of picking one and deciding
that the other one was persistently off (from the first?) by just enough that it wasn't worth looking at (which does
seem like a bug?), and its still in my chrony.conf, though I haven't tested whether I can remove it with 4.0.
Thanks,
Matt
On 9/29/20 4:29 AM, Miroslav Lichvar wrote:
On Sun, Sep 27, 2020 at 12:48:02PM -0400, Matt Corallo wrote:
Ah! I'd hit send on this before reading the chrony 4.0 update list - initial
testing shows latest git HEAD to not have the issue of selecting the RTC.
Interesting. I'm not sure which change in 4.0 would be related to
that (assuming it's not related to the trust option). It might help if
we could see the sources report, or selectdata in 4.0. I'd suspect
there is an occasional disagreement between the two GPS refclocks,
which would be fixed by setting a delay for them to reliably get an
intersection.
What exactly is the RTC refclock? Is it an external application
feeding chronyd via SHM or SOCK?
Finally, it would be nice to get a warning if "maxlockage" is not set
when rate is >1 - if I'm understanding it correctly it doesn't make
sense to have a rate > 1 if the lock source doesn't have a negative
dpoll (which should roughly be the common setting for eg a GPS set to
1HZ PPS, as in my setup).
Ok, I think I see what you mean. This could work only when locked to a
polling refclock (i.e. not SOCK). I'll look into that after 4.0.
They are locked to ntpshm fed from gpsd, so, indeed, polling, if I understand correctly.
--
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "help" in the subject.
Trouble? Email listmaster@xxxxxxxxxxxxxxxxxxxx.