ntp offset very high

[Rick Garland]

Hi all:

Working on an L3000 w/ 11.00 and trying to sync this server with the others in the network. All other servers have sync'd with no trouble but this system has an offset of 220+. Been looking for clues on how to get this offset lower but cannot find much. The actual time between this server and a stratum 2 server is less than 1 sec (UTC).

I have tried the ntpdate -B and can see the offset get lower but then when I stop the ntpdate and start the ntp daemon, it says the offet is 220+ again.

How to get that offset lower so I can sync the time

[Bill Hassell · Accepted solution]

ntpdate -B will slew the time in fractions of a second (ie, a few minutes offset will require hours to sync). The reason it is so slow is to prevent the loss/gain of any one second period. Essentially, ntpdate makes the clock run just a bit slower or faster as needed.

Now this is only needed in a production system where databases and time-critical programs will be able to account for every second (or not see the same second more than once). Now if the server is not running any appications, you can step change the time with ntpdate -b rather than -B.

Also test your servers (note: plural) by using:

ntpdate -p

You will see a list of all the time servers used by this system. Then query each of the servers with ntpq -p server_name to see that each of them are working correctly. The key column is: reach and should be 377 which means a high quality connection.

[James R. Ferguson]

Hi Rick:

I presume that you are looking at the output of 'ntpq -p'.

Do you have a (relatively) high 'disp'ersion value for the server in question? This would reflect poor(er) network conditions for that server. At an offset of 220 (msec) I would also expect to see synchronization lost messages in the 'syslog'.

Regards!

...JRF...

[Rick Garland]

Trying the ntpdate -b as well, no luck. the ntpdate -p shows all OK. The reach for the other servers is 377 so I am good there. The reach for the server in question is less than 40, not good at all. The disp will start at 16000 and with every unicast default of 64 sec will decrease until it gets down to ~500 and then goes back to 16000.

The server I am trying to sync with is a stratum 2 server which is sync'd with a stratum 1 server no problems.

[James R. Ferguson]

Hi (again) Rick:

Have you checked your network configuration on the troublesome server? Auto-negotiation should be *off* and duplex probably nailed to full on your side as well as at the network switch. Does 'lanadmin' show any significant errors?

Regards!

...JRF...

[Rick Garland]

The LAN is set to 100FD with AUTO-NEGOTIATE OFF.

I agree with JRF in that it could very well be a networking issue but I am at a loss in finding it.