[Pool] Leap Indicator bits

Ask Bjørn Hansen ask at ntppool.org
Fri Jul 3 18:06:21 UTC 2015


> On Jul 3, 2015, at 7:42 AM, Miroslav Lichvar <mlichvar at redhat.com> wrote:
> 
> This is very interesting. Thanks for sharing the data.
> 
> How close to the leap second were the "23" data collected?

In the hour before and up until midnight (the leap second).

Looking at the data minute by minute (just glancing with a human eye) I don’t see the number of leap announcements going up in a meaningful way inside the last hour; but the data is pretty noisy because it’s basically just ~2-3 measurements per server within each hour at a random minute.

> I was monitoring a subset of IPv4 servers from the pool and I saw a
> larger percentage of servers not announcing the leap second, so I'm
> wondering how that last hour looked like. In the first measurements
> after the leap second about 10% of them was off by one second. Maybe
> the IPv6 servers were better on average?

Depending on how you chose your subset you might have run into some countries having worse data because the popular stratum 1s didn’t do it right (as was indicated on the list here for .nl for example).

At a glance (see below) it doesn’t look like IPv4 vs IPv6 servers did significantly differently (as expected). There are roughly 3x more IPv4 than IPv6 servers in the pool currently.


Ask

mysql> select date(ts) d, s.ip_version ip, count(distinct(server_id)) servers from log_scores inner join servers s on (s.id=log_scores.server_id) where ts > '2015-06-29 19:00' and step > 0 and attributes is not null  and attributes = '{"leap":1}' group by d, s.ip_version order by d;
+------------+----+---------+
| d          | ip | servers |
+------------+----+---------+
| 2015-06-29 | v4 |      43 |
| 2015-06-29 | v6 |      13 |
| 2015-06-30 | v4 |    2764 |
| 2015-06-30 | v6 |    1016 |
| 2015-07-01 | v4 |     391 |
| 2015-07-01 | v6 |      88 |
| 2015-07-02 | v4 |     107 |
| 2015-07-02 | v6 |      35 |
| 2015-07-03 | v4 |      94 |
| 2015-07-03 | v6 |      32 |
+------------+----+---------+
10 rows in set (14.66 sec)



More information about the pool mailing list