Very similar traffic pattern can be seen here: http://lx.ujf.cas.cz/ntp-lx/lx-pkt.html. The time is CET=GMT+1 at those graphs. I guess that there are clients who see the leap second flag but are unable to make this jump on its own. They lurk only for the right moment.
Karel Sandler > From: "Nelson Minar" <[email protected]> >I just noticed an interesting traffic pattern on my pool server's logs: > NTP traffic went up a huge amount right before the transition to January > 1, 2009. You can see the change here: > http://www.somebits.com/ntp/one%20month.html > > Requests / second went up from my usual 25 to nearly 300, a 12x increase. > > Some of the traffic may be new clients (I see a 3x spike in unique IPs), > but a lot of it looks like existing clients sending more requests. > Usually a single IP sends an average of 3 requests / 10 minutes, but > during the spike interval it goes to 9 requests / 10 minutes, for a 3x > increase. It's possible that's actually new clients sharing IP addresses > of a NAT router. > > The oddest thing is the duration of the spike. I'd expect it to be a > brief spike around 00:00 Jan 1 UTC as clients flubbed the leap second. > But in fact the increase starts almost two days before and peaks several > hours before 00:00. > > Any guesses what happened? Did other pool servers see the same spike? I > have more data that could be analyzed if someone's really excited. > > PS: the pool server monitor thinks my offset was below 10ms most of the > time over this time period. > > _______________________________________________ > timekeepers mailing list > [email protected] > https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers > _______________________________________________ timekeepers mailing list [email protected] https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
