Re: ntp problem
On Tue, 24 Jun 2014, Pierre Frenkiel wrote:
> ==> ntpq -p
> remote refid st t when poll reach delay offset jitter
> ==============================================================================
> outbound-smtp.n 192.93.2.20 2 u 53 64 377 29.858 -189901 2903.84
> solsocog.de 131.188.3.222 2 u 28 64 377 32.191 -190155 2883.27
> merlin.ensma.fr 192.93.2.20 2 u 46 64 377 30.728 -189979 2843.02
> loin.ploup.net 213.251.128.249 2 u 48 64 377 28.665 -189960 2854.50
> oods.roflcopter 195.83.222.27 2 u 16 64 377 24.955 -190301 2868.25
> excalibur.exsv. 194.57.169.1 3 u 27 64 377 25.564 -190194 2918.50
ntpd will refuse to skew a clock that went that badly out of sync by
default.
start ntpd with the '-g' option to fix that (add it to /etc/default/ntp).
However, beware that it will skip (set) the clock, which can annoy badly
written applications if they're running when the clock changes.
Once ntpd is running, it should be able to discipline the clock and check
for systematic drift correction information (refer to ntpd's driftfile
option).
However, I don't know if ntp can make *use* of that systematic drift
correction. And the same goes to Debian itself:
We used to apply drift correction (stored in /etc/adjtime) when we still ran
hctosys / systohc during system boot/shutdown (refer to
/etc/init.d/hwclock*). I am unsure whether we still run that properly.
So, if you cannot live with the "ntpd -g" skip on boot/resume, you'll likely
have to set up the adjtimex package manually, and configure the hwclock
package (also manually).
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: