Hello Anssi
On 30.06.2012 20:14, Anssi Johansson wrote:
Fabian Wenk kirjoitti:
I am curios if your numbers are really packets per second.
Yes, those really are packets per second and I'm using the same method
as you for stats.
$ (ntpdc -c iostats; sleep 10; ntpdc -c iostats) | grep "packets sent"
packets sent: 882188250
packets sent: 882396875
$ echo $[ (882396875-882188250)/10 ]
20862
Wow, impressive numbers. Before my e-mail yesterday, I did also
count my packets manually and they are really packets / minutes
(as in my graphs). My Servers are located in Switzerland, so they
are in the ch zone. When I compare Turkey [1] with Switzerland
[2], then your large number of clients make sense.
[1] http://www.pool.ntp.org/zone/tr
[2] http://www.pool.ntp.org/zone/ch
For Turkey are only 8 (IPv4) and 3 (IPv6) servers and for
Switzerland are 72 (IPv4) and 24 (IPv6) servers in the Pool. This
is a huge discrepancy, when we also compare it with the
population of each country, Turkey ~75 Million [3] and
Switzerland only ~8 Million [4].
[3] http://en.wikipedia.org/wiki/Turkey
[4] http://en.wikipedia.org/wiki/Switzerland
I'm not planning to change my NTP server configuration because of this,
because I have an educated guess that the traffic will return to normal
levels after we've passed the leap second. The traffic surge started at
around 03:00 local time, at which time my NTP server got to know of the
upcoming leap second and started announcing it in the replies. Once the
leap second has passed, those announcements will stop and hopefully the
traffic will also return to the previous levels.
Yes, it went down again, I just checked your graph. :)
Could be some kind of device (e.g. access routers) which are
mostly deployed in Turkey and using the Pool to sync, which show
this behavior during the leap second day.
I've also seen some peaks on my other NTP servers today, for example
http://leopardi.miuku.net/stats/ntppackets.html and I believe those are
also of primarily Turkish origin. I'm not particularly concerned about
those either.
But here you have the same short peaks which I have seen on my
servers. So this is could have been something different on
'kameli'. Is this server also uses in some "internal" environment
with a large number of local clients (e.g. Windows computers)
which are doing a lot more ntp requests during leap second day? I
am just speculating.
bye
Fabian
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
