It turn out that there was a hacked VM on another host on the same switch - the 70Mb/s of warez traffic was overloading the switch and causing dropped packets which in turn seems to have caused some ntp clients to reduce their poll interval (bad design not sure which client does that - it's not the stock ntpd)
What I don't get is why it's still spiking up, the packet loss is gone but I was seeing 440 packets a second at one point this morning (speed set to 1000Mb/s). John Todd Eddy wrote: > http://rikku.vrillusions.com/ntp/ > > Mine looks normal. I have noticed the actual offset is a lot more > chaotic after upgrading the server from ubuntu 6.06 to 8.04 and I know > there was a new ntpd verison. Full restarts of the server can knock it > +/- 60msec or more and take 2 or 3 days to get to +/- 1 msec. But it's > always been around 20 packets/sec with it occasionally going over 30. > This is with the netspeed set to 100mb/sec btw. > > John Pettitt wrote: > | > | > | I have seen a spike in ntp traffic on my co-lo box from ~150 requests > | per second to well over 300 (peak 393). > | > | http://echelon.no-such-agency.net/mrtg/echelon_ntp_reqsec.html > | > | I'm not seeing a similar rise in the number of client addresses > | > | http://echelon.no-such-agency.net/mrtg/echelon_ntp_users.html > | > | However it correlates with some ntp instability on my end when I > | upgraded to the latest version and my ISP had a route flap/packet loss > | issue. If nobody else is seeing a spike I'm going to put it down the > | ntp clients reducing their poll time when things got flaky. So the > | question - > | > | Am I the only one seeing a spike? > | > | John > | . > | _______________________________________________ > | 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
