Help !! NTPD time sync steps instead of slews
 

I have ntp time synchronization setup between several Redhat 7.3 client machines (default install of everything with NO updates) which sync to 2 local ntpd timeservers which are sync'd to 3 stratum 2 servers on the net.

Everything Sync's fine but I notice that on most of my client machines the time is stepped instead of slewed a couple of times a week. This causes my application software to fail in different ways depending on how big the change is and if the change is forwards or backwards. Can I prevent the ntpd client from EVER stepping the time ? I step change the time at bootup using ntpdate (I am already doing this) and then when ntpd starts, the machines start polling every 1024 seconds (17 min) against my time servers and should stay synchronized. There should never be a reason for ntpd to step-change the time. Also ....what are the fudge statements used for.

The error messages I see in the /var/log/messages file on just about every machine (about 30 machines) appear at random and look like this :

Mar 24 10:21:15 server04 ntpd[807]: synchronisation lost
Mar 24 10:25:41 server04 ntpd[807]: time reset 1.916137 s
Mar 24 10:25:41 server04 ntpd[807]: synchronisation lost
Mar 24 10:45:10 server04 ntpd[807]: time reset -3.469102 s
Mar 24 10:45:10 server04 ntpd[807]: synchronisation lost
Mar 24 11:08:03 server04 ntpd[807]: time reset 14.546810 s

The "time reset "XXX.XXX s" is what is killing me since it steps the time instead of skewing.

A "ps -ef |grep ntp" on both the server and the client reveals :

ntp 807 1 0 Jan02 ? 00:00:00 ntpd -U ntp

My config file on my stratum 3 servers that synch to the internet stratum 2 servers is as follows :

server 127.127.1.0 # local clock
fudge 127.127.1.0 stratum 10
server rolex.usg.edu # Stratum 2
server sundial.columbia.edu # Stratum 2
server ntp1.cs.wisc.edu # Stratum 2
driftfile /etc/ntp/drift
broadcastdelay 0.008
authenticate no

My client configuration is as follows :

server 127.127.1.0 # local clock
fudge 127.127.1.0 stratum 10
server NTPhost1 # Primary Company NTP server
server NTPhost2 # Secondary Company NTP server
driftfile /etc/ntp/drift
broadcastdelay 0.008
authenticate no


I have tested all of the ntp time servers at stratum 1 and stratum2 from my location and have picked the 3 most reliable.