On Mon, Jan 25, 2021 at 09:42:14AM +0100, Miroslav Lichvar wrote: > On Sun, Jan 24, 2021 at 01:45:34AM +0100, Kurt Roeckx wrote: > > chronyc> selectdata > > S Name/IP Address Auth COpts EOpts Last Score Interval Leap > > ======================================================================= > > T 192.168.0.2 N ----- ----- 28 1.0 -8012us +8034us N > > T 192.168.0.3 N ----- ----- 25 1.0 -7523us +7645us N > > T 192.168.0.6 N ----- ----- 9 1.0 -7278us +7859us N > > T 192.168.0.7 N ----- ----- 26 1.0 -7109us +7295us N > > T 192.168.0.21 N ----- ----- 10 1.0 -7846us +8107us N > > * excelsior.roeckx.be Y ----- --TR- 0 1.0 -6669us +7580us N > > The last source is trusted (T in EOpts). That comes from the nts > option. The intervals of the other sources don't fit inside the > interval of the trusted source, so they are not selectable. > > > chronyc> selectdata > > S Name/IP Address Auth COpts EOpts Last Score Interval Leap > > ======================================================================= > > T 192.168.0.2 N ----- ----- 18 1.0 -8449us +8306us N > > D 192.168.0.3 N ----- ----- 16 1.0 -6972us +7014us N > > D 192.168.0.6 N ----- ----- 0 1.0 -7151us +7432us N > > T 192.168.0.7 N ----- ----- 18 1.0 -8242us +8398us N > > T 192.168.0.21 N ----- ----- 0 1.0 -7279us +7404us N > > * excelsior.roeckx.be Y ----- --TR- 6 1.0 -7161us +7701us N > > Here, some sources fit inside that interval, but they don't pass the > selection for having a large root distance.
So the server needs to be better than the worst trusted server that's combined before it can also be combined? > > I'm not sure if I can get the value of the root distance, but it > > should be almost the same, surely not a factor of 3 difference. > > The length of the interval is the root distance at the time of the > selection. The "+/-" value in the sources output is the root distance > at the time of the measurement. That is also reported in the > measurements log. The 'D' state should clear up after 32 updates of > the selected source if the distance doesn't become too large. Are you saying that even when it was always smaller than the combine limit * selected peer's root delay, when it moved away from the T state, it needs to do that for 32 updates? I guess it changes too much to the T state. Note that the +/- is documented as the margin of error, which is not the same as the root delay for me. Kurt -- To unsubscribe email chrony-users-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-users-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.