Re: -EXT-Re: [chrony-users] Re: chrony losing sync with timeserver and never recovers |
[ Thread Index |
Date Index
| More chrony.tuxfamily.org/chrony-users Archives
]
- To: chrony-users@xxxxxxxxxxxxxxxxxxxx
- Subject: Re: -EXT-Re: [chrony-users] Re: chrony losing sync with timeserver and never recovers
- From: Miroslav Lichvar <mlichvar@xxxxxxxxxx>
- Date: Fri, 20 Oct 2017 09:59:16 +0200
- Authentication-results: ext-mx01.extmail.prod.ext.phx2.redhat.com; dmarc=none (p=none dis=none) header.from=redhat.com
- Authentication-results: ext-mx01.extmail.prod.ext.phx2.redhat.com; spf=fail smtp.mailfrom=mlichvar@xxxxxxxxxx
- Dmarc-filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 0934E81DF6
On Thu, Oct 19, 2017 at 03:52:17PM -0400, Bryan Seitz wrote:
> [seitz@storage ~]$ chronyc sources
> 210 Number of sources = 7
> MS Name/IP address Stratum Poll Reach LastRx Last sample
> ===============================================================================
> ^* ntp-gps1.fiber.house 1 7 377 122 +799ns[-2226ns] +/- 1252us
> ^- some.server 1 10 377 995 +32us[-3108ns] +/- 35ms
> ^- another.server 1 10 377 927 -3831us[-3869us] +/- 15ms
>
> Where ntp-gps1.fiber.house was dead/down/not powered up, lastRx was 10h but it remained 'synced' to it :(
> Seems like correct operation would mark this server as offline, but continue to poll and mark it online again
> once it comes back up? I know this is how 'ntpd' works.
chrony doesn't work like that. The ntp-gps1.fiber.house source is
about 10 times better than the next best source (comparing the +/-
value). When it stops responding, chronyd will not switch to another
source until the estimate of the maximum local error becomes worse
than the estimate of error with another source, or all samples of the
source with * are older than all samples of the other sources. You
could increase the maxclockerror option to speed up the former, or
decrease the maxsamples option to speed up the later.
With a GPS refclock the difference in accuracy may be 3 or more orders
of magnitude. When the GPS signal was lost only for few minutes, I
think it would be a bad idea to switch to an NTP source.
> On Thu, Oct 19, 2017 at 05:04:14PM +0000, Parker, Michael D. wrote:
> > I had a similar type of problem using the release Chrony 2.x under RHEL 6 using only 2 time sources.
> > All looked good at the start but eventually the problem showed up.
> > If I recall my research, it had something to do some type of time source window situation.
> > If the sources' +/- differences windows did not overlap chrony quit trying to go back into sync.
> > I was polling much more frequently than you were.
That works as expected. If there are only two sources and they don't
agree with each other (they are marked with the x symbol), chronyd
knows that at least one of them is wrong, but it doesn't know which
one that is, so it must give up on synchronization until they agree
again. That's the reason why 3 is the minimum recommended number of
sources.
--
Miroslav Lichvar
--
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.