Hi Dan, Everything seems pretty much back to normal this morning. I’ll have to take another look at yesterdays data to see when that happened. I’ll add a suitable graph to my site as an example of wide area routing induced offset glitches. Mike > Le 11 janv. 2017 à 00:14, dan@drown.org a écrit : > > On Jan 10, you wrote: >> I saw 5ms offset jumps on most of the internet ntp servers that I am using at about 01:00UTC today. The servers are as widespread as NIST Boulder, NPL UK, Hungary, Scotland, Spain, France, Czechoslovakia. > > Did the RTT change at the same time? 5ms offset change might show up as a 10ms increase or decrease in latency. > > I saw a jump a few hours later reaching utcnist2.colorado.edu: > > ======================================================================================================================== > Date (UTC) Time IP Address L St 123 567 ABCD LP RP Score Offset Peer del. Peer disp. Root del. Root disp. > ======================================================================================================================== > 2017-01-10 02:57:48 128.138.141.172 N 1 111 111 1111 10 13 1.00 -1.260e-02 4.826e-02 1.003e-06 2.441e-04 4.883e-04 > 2017-01-10 03:15:02 128.138.141.172 N 1 111 111 1111 10 13 1.00 -9.482e-03 5.012e-02 1.003e-06 2.441e-04 4.883e-04 > 2017-01-10 03:32:19 128.138.141.172 N 1 111 111 1111 10 13 1.00 -1.173e-02 4.629e-02 1.003e-06 2.441e-04 4.883e-04 > 2017-01-10 03:49:35 128.138.141.172 N 1 111 111 1111 10 13 1.00 -1.268e-02 4.887e-02 1.003e-06 2.441e-04 4.883e-04 > 2017-01-10 04:06:51 128.138.141.172 N 1 111 111 1111 10 13 1.00 -1.256e-02 4.922e-02 1.003e-06 2.441e-04 4.883e-04 > --- RTT (Peer del.)/Offset changes here > 2017-01-10 04:41:15 128.138.141.172 N 1 111 111 1111 10 13 0.88 -5.550e-04 2.369e-02 1.002e-06 2.441e-04 4.883e-04 > 2017-01-10 04:58:23 128.138.141.172 N 1 111 111 1111 10 13 0.78 -1.634e-03 2.340e-02 1.687e-06 2.441e-04 4.883e-04 > > https://dan.drown.org/vps3/latest/remote-statistics.128.138.141.172.png > > You can see from the graph this offset change was most likely a change in both routes (request and response) between my server and the utcnist2 server. > > I saw one other server change latency at around the same time. It's possible one of the backbone networks was doing maintenance work. "The power of accurate observation is commonly called cynicism by those who have not got it. » George Bernard Shaw