Re: [chrony-users] chrony switch to low stratum very slow

something interesting:

2019-08-20 23:40:40 PPS 1 2.645 0.079 -9.830e-07 N 1 7.856e-07 -2.762e-08 1.000e-09 1.058e-05 2.998e-05
2019-08-20 23:40:56 PPS 1 2.639 0.098 -8.795e-07 N 1 1.459e-06 1.809e-07 1.000e-09 1.008e-05 2.882e-05
2019-08-20 23:41:12 127.127.1.0 2 2.621 0.124 4.482e-06 N 1 1.895e-06 5.462e-07 3.487e-05 3.167e-05 2.876e-05
2019-08-20 23:41:28 PPS 1 2.603 0.243 -3.006e-05 N 1 5.709e-06 -1.859e-11 1.000e-09 1.400e-05 7.117e-05
2019-08-20 23:41:44 PPS 1 2.580 0.411 9.807e-06 N 1 6.495e-06 -6.043e-10 1.000e-09 1.041e-05 6.926e-05
===================================================================================================================================
Date (UTC) Time IP Address St Freq ppm Skew ppm Offset L Co Offset sd Rem. corr. Root delay Root disp. Max. error
===================================================================================================================================
2019-08-20 23:42:03 127.127.1.0 2 2.568 0.145 1.146e-05 N 1 3.293e-06 2.805e-11 3.487e-05 2.425e-05 4.995e-05
2019-08-20 23:42:16 PPS 1 2.528 0.301 -2.630e-05 N 1 6.101e-06 1.261e-10 1.000e-09 1.077e-05 6.772e-05
2019-08-20 23:42:32 PPS 1 2..521 0.405 1.276e-05 N 1 6.037e-06 -4..840e-10 1.000e-09 1.279e-05 6.072e-05
2019-08-20 23:42:48 PPS 1 2.514 0.416 -2.454e-07 N 1 7.098e-06 3.986e-11 1.000e-09 7.712e-06 4.812e-05
2019-08-20 23:43:04 PPS 1 2.492 0.364 -2.954e-06 N 1 6.634e-06 2.396e-07 1.000e-09 1.408e-05 3.066e-05
2019-08-20 23:43:20 PPS 1 2.526 0.250 5.345e-06 N 1 4.849e-06 2.168e-06 1.000e-09 9.852e-06 3.918e-05

127.127.1.0 is the local clock with stratum 10 in chrony.conf

This might be caused by missing PPS signal and it switched server from PPS to local clock just a few seconds after the PPS signal missing and switched back to PPS in the similar manner.

Comparing to switch the server unreachable over 10 minute, the behavior here is different.

Frank Huang

On Wed, Aug 14, 2019 at 11:39 AM chuang213 <chuang213@xxxxxxxxx> wrote:

updates
settings
stratumweight 0.0001
reselectdist 0.000001
and results
2019-08-14 18:10:21 172.19.40.23 2 -14.122 0.098
-4.465e-06 N 5 3.006e-05 -1.439e-05 2.429e-04 3.733e-05 2.611e-04
2019-08-14 18:11:26 172.19.40.23 2 -14.125 0.082
-1.042e-06 N 5 1.967e-05 -7.703e-06 2.435e-04 3.740e-05 2.402e-04
2019-08-14 18:26:43 192.168.10.9 3 -14.126 0.001
3.704e-05 N 4 6.061e-05 4.936e-10 3.443e-04 5..172e-05 1.159e-03
2019-08-14 18:27:16 192.168.10.9 3 -14.126 0.001
1.390e-06 N 4 5.970e-05 -3.650e-05 3.443e-04 5.707e-05 2.935e-04
===================================================================================================================================
Date (UTC) Time IP Address St Freq ppm Skew ppm
Offset L Co Offset sd Rem. corr. Root delay Root disp. Max. error
===================================================================================================================================
2019-08-14 18:27:48 192.168.10.9 3 -14.126 0.001
8.495e-06 N 4 4.816e-05 -2.210e-05 3.443e-04 6.272e-05 2.992e-04
2019-08-14 18:28:20 192.168.10.9 3 -14.126 0.004
1.395e-05 N 3 3.780e-05 -1.755e-05 3.459e-04 2.490e-05 2.990e-04
2019-08-14 18:28:52 192.168.10.9 3 -14.125 0.015
8.798e-06 N 3 2.973e-05 -1.407e-05 3.459e-04 2.667e-05 2.654e-04
2019-08-14 18:29:25 192.168.10.9 3 -14.107 0.065
4.334e-06 N 3 2.703e-05 -1.105e-05 3.494e-04 2.321e-05 2.596e-04
2019-08-14 18:29:57 192.168.10.9 3 -14.031 0.115
6.838e-06 N 3 1.811e-05 -9.552e-06 3.494e-04 2.806e-05 2.512e-04

much better than the default setting. it took 15 minutes to switch the
server and the local clock drift during this period is around 30 us.

thanks for the advice to get this.

Frank Huang

On Tue, Aug 13, 2019 at 12:01 PM chuang213 <chuang213@xxxxxxxxx> wrote:
>
> Thanks for the advice and pointing it out the policy that chrony used.
> I'll try them soon.
>
> chrony is good at quick convergence to the servers, while in the
> mobile environment drop of time server is inevitable.
>
> Frank Huang
>
> On Mon, Aug 12, 2019 at 12:15 AM Miroslav Lichvar <mlichvar@xxxxxxxxxx> wrote:
> >
> > On Fri, Aug 09, 2019 at 11:38:23AM -0700, chuang213 wrote:
> > > I changed "stratumweight 0.00010" and it did help, the switching time
> > > reduced to 21 minutes, though still too long. And the side effect is
> > > that this change caused the chrony to use local clock for a long
> > > period of time. the local clock is a stratum 10 one.
> >
> > Have you tried also setting reselectdist to 0? You can also increase
> > maxclockerror to 10 or even more to indicate the local clock is less
> > stable, which should shorten the free running interval.
> >
> > Generally, we don't want to be switching between sources too
> > frequently to minimize the frequency error.
> >
> > --
> > Miroslav Lichvar
> >
> > --
> > To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
> > with "unsubscribe" in the subject.
> > For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
> > with "help" in the subject.
> > Trouble? Email listmaster@xxxxxxxxxxxxxxxxxxxx.
> >