I can't answer your full question; the jitter seems strange to me. A few thoughts, though:
- NTP prefers lower-strata clocks. Thus it likes the "darkcity.cerius" host, a stratum 1, more than Blue, at stratum 2. (Throw ntp.conf the "prefer" keyword on red to make it, well, prefer Blue's time.) - NTP doesn't know the "right" time, per se, but it sees that the "darkcity" host has a +3ms offset, as does the "europium" Canonical host, and the other preferred one is +5ms. Thus Blue's very small negative offset is an "outlier," correct or not. (You've got to love when a 4ms difference constitutes an outlier!) Essentially, each host is just 'guessing' (through scientific means) which is the most accurate and holds it as the standard. Your two systems sync to a (partially) different list of servers, so they're choosing different clocks to hold as correct, hence them not quite agreeing. - Blue's IP, 192.43.244.18, is not a conventional "non-routable" internal IP. The delay amounts to about 14ms latency, so I have a strong hunch that you're going out the router and back in, or at least passing through a router that's introducing some latency. I'd try a traceroute from red to blue to verify. Hope this helps! -- Matt (N1ZYY) On Wed, May 7, 2008 at 5:06 PM, Kipton Moravec <[EMAIL PROTECTED]> wrote: > I have two Ubuntu computers. A server (Blue) and a workstation (red). > > I thought I would set up Blue as a NTP server, and have red get time > from it and the pool. > > Blue says its time is good, but red says blue is an outlier. > > Blue is a 3.0 GHz Pentium, with 512M RAM and not doing anything else at > the moment (not even file serving). They are connected together with an > 8 port switch. So there is no distance between them. > > Here is Blue: > > [EMAIL PROTECTED]:~$ ntpq -p > remote refid st t when poll reach delay offset > jitter > ============================================================================== > +europium.canoni 192.36.133.17 2 u 507 1024 377 116.412 3.837 > 1.207 > -server1-a.your. 64.202.112.75 2 u 526 1024 377 32.929 6.629 > 0.734 > +ntp1.us.grundcl 164.67.62.212 2 u 443 1024 377 46.272 3.858 > 0.653 > -ntp.LogicX.net 18.103.0.198 2 u 266 1024 375 55.304 -1.035 > 0.644 > *time.nist.gov .ACTS. 1 u 478 1024 377 32.773 2.198 > 0.201 > -ntp3.tamu.edu 128.194.254.7 3 u 414 1024 377 46.361 13.276 > 1.496 > > [EMAIL PROTECTED]:~$ ntptrace > localhost: stratum 2, offset 0.002716, synch distance 0.041578 > time.nist.gov: stratum 1, offset 0.000000, synch distance 0.002650, > refid 'ACTS' > [EMAIL PROTECTED]:~$ > > Here is red: > > [EMAIL PROTECTED]:~ $ ntpq -p > remote refid st t when poll reach delay offset > jitter > ============================================================================== > +europium.canoni 192.36.133.17 2 u 91 1024 377 116.228 3.783 > 0.903 > -Blue 192.43.244.18 2 u 72 1024 377 0.144 -0.415 > 2.072 > *darkcity.cerias .GPS. 1 u 315 1024 377 56.111 3.591 > 11.436 > -tesla.fireduck. 198.82.1.201 3 u 122 1024 377 53.209 -2.050 > 0.208 > -clock.trit.net 192.12.19.20 2 u 108 1024 377 43.304 -10.235 > 0.827 > +server.donkeyfl 18.26.4.105 2 u 125 1024 377 56.097 5.461 > 0.514 > > [EMAIL PROTECTED]:~ $ ntptrace > localhost: stratum 2, offset 0.002221, synch distance 0.062901 > 128.10.252.7: timed out, nothing received > ***Request timed out > [EMAIL PROTECTED]:~ $ > > Why does red not use the data from Blue? > > What would cause Blue to have so much jitter when they are so close > together? Do I have something set wrong? > > Kip > -- > Kipton Moravec KE5NGX > "Always do right; this will gratify some people and astonish the rest." > --Mark Twain > > > _______________________________________________ > 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
