[ntp:questions] NTP 4.2.8p10 released
Brian Inglis
Brian.Inglis at SystematicSw.ab.ca
Tue May 9 16:04:02 UTC 2017
On 2017-05-09 01:45, ashu6486 at gmail.com wrote:
> How can we add redundant server into NTP 4.2.8p10 ? If we try to
> edit the configuration file it's not switching to the redundant
> server?
After changing the /etc/ntp.conf file, you need to restart any
server ntp daemon using it to put changes into effect.
When dealing with ntp, patience is often required, to see if
changes have any obvious effect.
If you have insufficient truechimer sources (3, 5, ...) to form
a *majority* clique, the time will not be updated, as if you had
no sources.
The ntpd daemon looks at the time provided by all upstream sources
using its algorithms and calculates the correctness interval of the
majority clique survivors using averages weighted by the normalised
reciprocal root distance to give the smallest maximum error in both
offset and jitter.
If the survivor source with the smallest root distance (and thus
smallest maximum error), the candidate source, is not the same as
the previous system source and the offset difference between the
candidate source and the previous system source is less than the
clockhop threshold (initially 1ms), the system source stays the
same and the clockhop threshold is halved, so it could take more
than one poll interval to replace the system source if offsets
are very similar.
These values are updated every time another packet is received
from any upstream source and passes sanity checks, or the system
source is unreachable after eight poll intervals. This could take
128s (~2min) for LAN sources with minpoll 4, or 512s (~8min) for
network sources with minpoll 6; except at startup with sources
using iburst, when six packets are sent at poll 1 (2s intervals),
or after operating for a long period, with network sources at
default maxpoll 10/1024s, when it could take ~136min to become
unreachable.
For more details see:
https://www.eecis.udel.edu/~mills/ntp/html/prefer.html#clockhop
and linked/related pages.
--
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada
More information about the questions
mailing list