All of our servers are running NTP and all have different times!

[Stuart Abramson_2]

I recently got to looking at NTP on our servers.

They all have different times:

Wed Mar 31 13:10:07 EST 2004
Wed Mar 31 13:09:41 EST 2004
Wed Mar 31 13:11:06 EST 2004
Wed Mar 31 13:10:59 EST 2004
Wed Mar 31 13:10:59 EST 2004
Wed Mar 31 13:11:00 EST 2004
Wed Mar 31 13:11:55 EST 2004
Wed Mar 31 13:10:52 EST 2004
Wed Mar 31 13:11:07 EST 2004

BUT, they all run xntpd, and they are all configured like this:

# grep -v ^# /etc/ntp.conf
server ntp1.wcc.wabtec.com version 3 prefer
server ntp2.wcc.wabtec.com version 3 prefer
broadcastclient yes

Is this configuration wrong? Why isn't it working? Should I do a "peer" thing, to keep them all together?

When I run "ntpq -p" on them, several have big "offset" and "disp" values:

remote refid st t when poll reach delay offset disp
==============================================================================
wcsadc001.ad.wa 172.20.208.6 2 u 37 64 0 0.23 -5.571 16000.0
wcsadc002.ad.wa 172.20.208.6 2 u 41 64 0 0.26 2.917 16000.0
saroot@acshps03:/home/saroot/delivery/logs# grep wcs log.ntp
wcsadc001.ad.wa 172.20.208.6 2 u 19 64 0 0.32 -10.134 16000.0
wcsadc002.ad.wa 172.20.208.6 2 u 26 64 0 0.41 -15.076 16000.0
wcshps06
wcshps07
wcsadc001.ad.wa wcsadc001.ad.wa 2 u 51 64 0 0.26 43133.8 16000.0
wcsadc002.ad.wa wcsadc001.ad.wa 2 u 32 64 0 0.27 43210.6 16000.0
wcshps08
wcsadc001.ad.wa 172.20.208.6 2 u 12 64 0 0.26 -7.922 16000.0
wcsadc002.ad.wa 172.20.208.6 2 u 21 64 0 0.31 2.638 16000.0
wcshps09
wcsadc001.ad.wa 172.20.208.6 2 u 37 64 0 0.23 -5.571 16000.0
wcsadc002.ad.wa 172.20.208.6 2 u 41 64 0 0.26 2.917 16000.0

[Tom Smith_9 · Accepted solution]

Hi Stuart,
Are you seeing any errors in your syslog related to NTP?

Have you tried ntpdate to see if you can manually sync.

We were having some network problems a while back and I noticed that NTP was failing because the offset between my servers and my time source was too large so no update would occur. If I manually sync'd with ntpdate after the network problems were solved, NTP went back to updating properly.

Hope this helps.

[Steven E. Protter]

I don't know when you started ntpd.

But if your actual time server is accurate, all these systems should eventually synch right up.

If not, suspect and change the time source.

Microsoft's NT Server time server was not Unix compliant and could not give times to Unix ntp requestors. You'd need to upgrade the time service on your Microsoft boxes in that event.

Give me a little more and I can be a little more specific.

SEP

[A. Clay Stephenson]

You appear to have terrible network problems. The dispersion values are awful and reach is near zero. Note that you can't synchronize to a time source until dispersion falls below 1000ms and you are nowhere near that.

I suggest that you print /usr/share/doc/Net_Primer.txt. It is an excellent tutorial on debugging and setting up NTP.

[RAC_1]

You configuration is OK. When did you start NTP, if it has been started recently, depending upon the time difference between NTP master and client, clients may may take some time before they sync.

And how did you check the times on different NTP clients? Time gaps between checking times may also give you different results. I use
telnet client1 13;telnet client2 13 and compare the results.

Anil

[Stuart Abramson_2]

I stopped xntpd, reset the time with ntpdate, and within a few minutes we were all synched back up.

The problem was, as you all suggested, that we have had some LAN problems, and we moved our servers (one at a time, usually) from 100 BT to 1000 BT, and our "dispersion" became greater than 1000 ms, and the xntpd-s stopped synching.

Thanks all.

[Angus Crome]

I'm not sure exactly what two prefered servers would mean. It may have the effect that time is constantly being pulled and pushed by two (non-synced) systems.
I would drop the prefer on one or the other of those hosts.