[ntp:questions] Facing some issue in the ntp version Ver. 4.2.6p1
martin.burnicki at meinberg.de
Wed Aug 1 20:33:14 UTC 2012
Sorry for jumping in so late.
In my opinion the basic concept of the local clock driver was good, easy to
configure, sufficient for small, simple networks, and it just worked as
expected in older versions of ntpd.
As David Woolley has pointed out in one of his comments, the local clock can
perfectly be used with stratum 0 if the system time is disciplined
accurately outside of NTP, so ntpd is just used to make the disciplined
time available on the LAN. There *are* networks out there where this
configuration is used.
Of course orphan mode works better than a simple local clock setup if you
have a network with several redundant servers and/or refclocks which need
to keep time very reliably under all circumstances, but this requires a
version of ntpd running on those orphan nodes where orphan mode really
works as expected, and is IMO very much harder to configure for
So if you have just a simple network where one NTP node synchronizes to a
single refclock or one or more internet NTP servers, and you just want your
clients to keep the same time if your refclock is temporarily
unsynchronized or your internet connection is temporarily down, then the
local clock driver used to be the perfect solution. I know from many
customers that they consider such a solution as good enough.
The original poster said in one of his comments:
> My machine is the server for other nodes[blades] in the cluster.
> So it is not a leaf machine.
So it looks like he was just trying to install this simple setup.
If in current versions of ntpd the local clock stays selected as system peer
after the daemon has started even though good upstream time sources are
available, or if it "eats up the -g flag" after startup so ntpd terminates
itself when remote servers become available and the initial time offset
exceeds the sanity limit, then in my opinion the local clock driver is just
broken in the current version of ntpd and should be fixed. I don't know if
what the OP has found out could already be part of a fix.
As said above, this used to work perfectly with older versions, and it
should be up to the user if they are satisfied with a simple, easy
configuration, or if they want a more enhanced configuration requiring much
more basic knowledge to get it working as expected.
More information about the questions