[ntp:questions] New years glitch?

tylerwalden at gmail.com tylerwalden at gmail.com
Wed Jan 2 20:34:51 UTC 2013


The NTP servers for my environment sync off of 0-3.us.pool.ntp.org and we experienced a leap second jump on all of our servers:

2012-12-31T23:59:59.730993+00:00 localhost kernel: Clock: inserting leap second 23:59:60 UTC

This triggered on some systems without an upgraded kernel the linux hrtimer() bug which caused CPU spikes across my environments.  Happy New Years!

If I understand it correctly the problem is that certain pool servers are incorrectly reporting the leap=01 incorrectly? As the latest IERS bulletin specifies a leap second was NOT to be introduced.

What is the recommended approach to ensure this doesn't incorrectly happen again?  Once a server receives a leap=01 from an upstream server will it leap no matter what?

David, would it be possible to get the flags your sending to ntpq to check for the leap second flag in your program so we can run them using ntpq on a linux system ? I'd like to setup a monitor for this condition in our environment.


On Tuesday, January 1, 2013 4:19:31 AM UTC-8, steven Sommars wrote:
> For the past 6 months a number of NTP servers have been reporting leap=01.
> 
>  When 2013 began I saw a 1 second jump from NTP servers at University of
> 
> Minnesota, Apple, and a few other sites.
> 
> 
> 
> To my knowledge the underlying issue in ntpd has not been addressed.
> 
> 
> 
> 
> 
> 
> 
> 
> 
> On Tue, Jan 1, 2013 at 12:29 AM, David Taylor <
> 
> david-taylor at blueyonder.co.uk.invalid> wrote:
> 
> 
> 
> > On 01/01/2013 06:12, s.schaefer.at.acm at gmail.com wrote:
> 
> >
> 
> >> I'm using [0123].ntp.pool.org with my two internet accessing NTP
> 
> >> servers.  I missed the ball drop on new years because I was trying to
> 
> >> figure out why Nagios was complaining about my NTP servers dropping the two
> 
> >> internet accessing daemons and prefering each other, resulting in higher
> 
> >> and higher strata numbers.  My first reaction was to restart all these
> 
> >> non-internet accessing daemons, assuming they'd reset to the two
> 
> >> internet-accessing daemons, then honor the low stratum numbers and be sane.
> 
> >>
> 
> >> Then I noticed my two internet accessing daemons were reporting that they
> 
> >> were within milliseconds of their internet peers (multiple cnames sending
> 
> >> them to a non-overlapping set of real servers), BUT almost exactly one
> 
> >> second different from each other.  I first noticed this at a little after
> 
> >> midnight, but it went away by about 00:40 - I don't have status logs to say
> 
> >> exactly when.
> 
> >>
> 
> >> Huh?
> 
> >>
> 
> >
> 
> > My own plots show no glitch:
> 
> >
> 
> >   http://www.satsignal.eu/mrtg/**performance_ntp.php<http://www.satsignal.eu/mrtg/performance_ntp.php>
> 
> >
> 
> > Was something sending a Leap Second flag?  You could check with my program
> 
> > here:
> 
> >
> 
> >   http://www.satsignal.eu/**software/net.htm#NTPLeapTrace<http://www.satsignal.eu/software/net.htm#NTPLeapTrace>
> 
> >
> 
> > Recent versions of NTP should ignore a single leap-second flag, earlier
> 
> > versions may set a leap second even if only one of the upstream servers is
> 
> > indicating thus.
> 
> >
> 
> > Happy New Year and thanks to all those who have helped me in the past!
> 
> > --
> 
> > Cheers,
> 
> > David
> 
> > Web: http://www.satsignal.eu
> 
> >
> 
> >
> 
> > ______________________________**_________________
> 
> > questions mailing list
> 
> > questions at lists.ntp.org
> 
> > http://lists.ntp.org/listinfo/**questions<http://lists.ntp.org/listinfo/questions>
> 
> >



More information about the questions mailing list