ntp sync appears to bounce between values

[Dave Chamberlin]

I have ntp configure on 6 servers. One of them continually loses sync - the others rarely. All are running  hpux11.31. The server with the issue has logs like below. The time resets are opposite - +3.4 followed by -3.4 - like it cant make up its mind. All servers have identical configuration files, only this one has the issue......any ideas - btw - what does stratum of 8 mean?

Feb 10 08:00:29 jupiterv xntpd[6332]: synchronized to 10.32.255.1, stratum=8

Feb 10 08:00:32 jupiterv xntpd[6332]: time reset (step) 3.395968 s

Feb 10 08:00:32 jupiterv xntpd[6332]: synchronisation lost

Feb 10 08:04:49 jupiterv xntpd[6332]: synchronized to 10.32.255.2, stratum=8

Feb 10 08:04:46 jupiterv xntpd[6332]: time reset (step) -3.402812 s

Feb 10 08:04:46 jupiterv xntpd[6332]: synchronisation lost

Feb 10 08:10:05 jupiterv xntpd[6332]: synchronized to 10.32.255.1, stratum=8

Feb 10 08:10:08 jupiterv xntpd[6332]: time reset (step) 3.401069 s

Feb 10 08:10:08 jupiterv xntpd[6332]: synchronisation lost Feb 10 08:13:19 jupiterv su: + tc root-clr Feb 10 08:14:25 jupiterv xntpd[6332]: synchronized to 10.32.255.2, stratum=8

Feb 10 08:14:22 jupiterv xntpd[6332]: time reset (step) -3.404882 s

Feb 10 08:14:22 jupiterv xntpd[6332]: synchronisation lost

Feb 10 08:19:41 jupiterv xntpd[6332]: synchronized to 10.32.255.1, stratum=8

Feb 10 08:19:44 jupiterv xntpd[6332]: time reset (step) 3.404139 s

Feb 10 08:19:44 jupiterv xntpd[6332]: synchronisation lost

P.S. This thread has been move from HP-UX > System Administration to HP-UX > networking. - Hp Forum Moderator

[Bill Hassell]

The problem machines is trying to make sense of the ntp responses. Those massively large steps (more than 3 seconds) indicate a major issue with the NTP server(s) you are using.

Stratum levels start at 0 which is the most accurate source of time, typically the Naval Observatory reference clock. A stratum 1 connection must have direct connect (not networked) to this reference clock and provides NTP time sync over the network to stratum 2 servers which then provide syncs for stratum 3, etc. The higher the stratum level (like 8), the farther away it is. Most internet NTP sources are either stratum 1 or 2.

Now to answer your question, we would need to see the output from ntpq -p from both your problem server as well as the ones that are working OK. We also need the content from /etc/ntp.conf. To get rid of the comments, do this:

awk 'NF && ! /^[[:space:]]*#/' /etc/ntp.conf

 Best practice in the data center would have the network firewall(s) setup to obtain NTP sync from reliable stratum 1 and 2 servers. The best source is to use the NTP server pool provided by ntp.org. Here's more information:

http://www.pool.ntp.org/en/

Then all the systems in the data center would point to your firewall for NTP service and all the systems would be in sync with each other.