[LEAPSECS] June 30 not leap

[David Malone]

> The plot for the end of 2012 is very telling, for it means that our

> one Linux system with out-of-the-box configuration managed to be

> listening to one of the very few NTP servers that were still aberrent.


Indeed - I think I posted a graph at the time, and it seemed like
the number of extra announcements was close to being noise.


> David, do you have the numbers for 2012-07-01 and 2012-08-01 when

> there were several incidents of Linux systems locking because their

> NTP allowed leaps at the end of any month?


Unfortunately not. I started recording this data as part of my
efforts to record what was going on for the 2008 leap second. I
have data doing back to November 2008.


> Do the numbers give an indication of how many servers are in the

> NTP pool?


The three groups that I query are Stratum 1 (based on the public
Stratum 1 list), Stratum 2 (based on the public Stratum 2 list) and
Pool (based on making a few different queries to the continental
pool groups). I only update the list of machines queried infrequently.
Consequently, for the pool, I may be querying servers that are not
currently active in the pool. You can see how may servers I was
sending queries to at:

	http://www.maths.tcd.ie/~dwmalone/time/leaps/ntp-leapbits-totals.png

Short drops represent problems with my network connectivity. The
recent decrease was when I pruned all the servers that hadn't
responded to any queries recently. The fraction I use is the fraction
of responding servers. You can see the number of responding servers
over time at:

	http://www.maths.tcd.ie/~dwmalone/time/leaps/ntp-leapbits-responding.png

Again, you can see step changes when I add new servers to the list.


> I'm wondering what happened to clean things up in the past year,

> and I haven't seen anything written about the difference.

> E.g., to what extent did Ask and the NTP pool admins cull the aberrent

> servers, or to what extent did they get the admins to upgrade their

> NTP daemons, etc?


I've started thinking about this, but I don't have a clear picture
about what changed yet, but I'd be interested to know. It has become
easier to configure the leapseconds file in more recent versions
of ntpd, so I don't know if that had any impact.

	David.