[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: NTP.conf pool vs server



On Sun 11 Jun 2017 at 10:57:51 -0500, David Wright wrote:

> On Wed 07 Jun 2017 at 17:26:30 (+0100), Brian wrote:
> > On Wed 07 Jun 2017 at 10:30:54 -0500, John Hasler wrote:
> > 
> > > Remove Ntp and install Chrony.
> > 
> > Too easy. There would be nothing to rant about. :)
> > 
> > https://chrony.tuxfamily.org/comparison.html
> 
> I read this and did exactly that:

It's good to see someone testing things.
 
> Start-Date: 2017-06-07  12:58:22
> Install: […], chrony:i386 (1.30-2+deb8u2), […]
> 
> I looked at, but didn't change, the configuration file.

Three machines running chrony here. One (which is always on) has an
original chrony.conf and the only change to the other one (a laptop
which is suspended overnight) is to use my ISP's time servers. The
third one has just been booted after not being used for fourteen
days; it uses a single one of my ISP's time servers.
 
> Two days later, the laptop¹ was still running about five seconds
> slow, so:
> 
> Start-Date: 2017-06-09  17:39:56
> Purge: […], chrony:i386 (1.30-2+deb8u2), […]
> 
> Start-Date: 2017-06-09  17:43:19
> Install: […], ntp:i386 (4.2.6.p5+dfsg-7+deb8u2)
> 
> I looked at, but didn't change, the configuration file.
> 
> By the time I had brought up two swissclocks (from my server
> and other laptop) all three second hands were marching in step
> (and within one second of the radio wall clock).
> 
> ¹ had been "running free".

Machines number 1 and 2 are in step with my radio clock. Machine
number 3 came up three seconds slow. Within 6 minutes it agreed
with the other machines. I cannot account for this because I've
never had cause to examine chrony's configuration or internal
workings. Perhaps it has to do with frequency of polling? At the
moment my motivation to find out why is not high.

All my other machines use systemd-timesyncd on the basis it is
already available on them. The first three have used chrony for
nearly ten years without a discernable problem.

-- 
Brian.


Reply to: