Network time service only has one instance, however like HPUx on same network has two

[Kevin M Meyers]

Our HPUx in Texas only has one network time service listed
# *** check Network Time service:
# /usr/sbin/ntpq -p
Enter to continue:
remote refid st t when poll reach delay offset disp
==============================================================================
*ftwvdns.verizon pearldns.verizo 2 u 22 64 377 42.85 -0.644 0.85
however other HPUx servers on same network have two listed.

# *** check Network Time service:
# /usr/sbin/ntpq -p

Enter to continue:
remote refid st t when poll reach delay offset disp
==============================================================================
*autoproxy.veriz pearldns.verizo 2 u 43 64 377 96.94 1.049 1.05
+sacvdns.verizon pearldns.verizo 2 u 18 64 377 30.70 6.358 0.95

Why is that?
Thanks
Kevin M Meyers

[James R. Ferguson]

Hi Kevin:

It is good practice to have more than one time source configured in 'etc/ntp.conf'. You don't in this particular case.

You should configure multipe (two or three) servers for high-availabliity.

See the manpages for 'xntpd' and/or chapter-7 of the "Installing and Administering Internet Services" guide for more information:

http://www.docs.hp.com/en/B2355-90685/index.html

Regards!

...JRF...

[A. Clay Stephenson]

I'm gonna go way out on a limb and say because the ntp.conf's are different on the two boxes. Normally, the minimum number of timeservers you want is about 3. You need to add more timeservers to the np.conf file and bounce xntpd.

[Bill Hassell]

There is an old mariner's (paraphrased) saying:

Never go to sea with two chronometers (clocks). After all, which one is correct?

NTP has a very complex set of rules to maintain accurate time -- it does not simply jam the clock forward or backward without lots of tests. The 'correct' method to set NTP is to pick 3 or more servers so that loss of one or two will not affect the time. Now, that being said, in many companies, the company firewall/router has NTP built-in (possibly not configured though), so a single source may be appropriate.

Now loss of all NTP services is not a critical problem since the system just free runs until the service is restored then slowly adjusts the clock by a few milliseconds until it is back in sync. NTP will ignore any time server that is dozens of minutes out of sync.

All this being said, all servers within a company should have a standard for time sync hierarchy, thus defining what ntp.conf should have as servers. See also the man page for ntpq to evaluate NTP servers.

[Steven E. Protter]

Shalom Kevin,

Adding to the Olympic parade.

Either the ntp.conf files are different or the firewall/network/ISP configuration is different, blocking access to the second time source.

After confirming that the ntp.conf files are the same, check with the firewall administrator or isp.

Or simply find two or three more good time sources and add them to the ntp.conf file. I've always found the US navy is a good primary or secondary source.

SEP

[rick jones]

Not to challenge an olympian :) but ntp will not always slew the clock, it can do a stepwise adjustment, altering the time in one swell foop.

I'd configure multiple time sources even if one's firewall etc has NTP support.

Pointers about public timeservers can be found off of www.ntp.org.