I'm seeing some strange behavior with NTP. Running "ntpq -p" on one of my client systems is showing the following:

root@cc:~# ntpq -p
remote refid st t when poll reach delay offset jitter
==============================================================================
rt-ora1 LOCAL(0) 11 u 136 1024 377 0.544 24155.8 469.994
*LOCAL(0) .LOCL. 12 l 33 64 377 0.000 0.000 0.001

I'm confused about the large offset (24155) that's huge. When I run the date command on the systems there is about a 24 second variation between the NTP server and its client.

I've run NTP on many systems before and have never seen such large offsets. Does anyone have experience with this and any idea why this would happen. I have about 25 clients and all are experiencing much the same results.

Lastly does anyone have experience with NTP running on a VM? I was wondering if putting a server on a VM would be problematic.

Any help will be greatly appreciated.
Thank you

From the NTPQ docs:

peers
Obtains a current list peers of the server, along with a summary of each peer's state. Summary information includes the address of the remote peer, the reference ID (0.0.0.0 if this is unknown), the stratum of the remote peer, the type of the peer (local, unicast, multicast or broadcast), when the last packet was received, the polling interval, in seconds, the reachability register, in octal, and the current estimated delay, offset and dispersion of the peer, all in milliseconds.

So, 24155.8 msecs is about 24 secs which matches what you see using date.


As far as using NTP on a VM there should not be any issue there.

1 members found this post helpful.

There were (are?) known issues with VMware and RedHat. They may have fixed them, but in the past, there were combinations of RHEL and Vmware that would cause the time to drift wildly. You had to not only patch VMware, but also the Linux kernel with a VMware provided patch, to get NTP to work. Otherwise, the clock would drift so bad, the NTP server would die after about an hour.

They MAY have fixed it, but the OP doesn't say what version/distro of Linux or what kind of VM

1 members found this post helpful.

Thank you for the information, though what I was concerned about is that the offset and time difference between the systems was so large. In the past I've rarely seen an offset greater than 1000, they were generally between 300 and 700. My clients kept close track with the server.

My guess is the clocks on my NTP servers are rather bad and must be fluctuating wildly and the clients can't keep track. Possibly I should look at one of the other systems to act as the server. My previous servers were actual network clocks. I don't have one here and was hoping I picked a fairly accurate system as it was the newest.

For my NTP servers I have
Redhat on a Dell R710, no VMs, running Oracle 10G (picked because it is the newest)
Ubuntu running KVM with the NTP server on a VM on a Dell 2950 (also the DNS server)

Thanks for setting me and the OP straight. I was not aware of the issues and fortunately have not run into them.


thllgo - sorry for being off the mark with my answer, I misunderstood what you asked. As TBone said there are indeed issue that can occur with NTP and VMs, so that may account for some of your time drift.


regards,
z1p

thllgo: why not make your own radio clock and hook it up to one of your servers?

I like that idea, I'm guessing there are instructions someplace on the web?

Yep...and another possibility is to use a cheap GPS receiver. I grabbed a $69 USB GPS receiver, plugged it in, and configured NTP to use the generic NEMA data. Works like a charm, and I have a very reliable stratum-1 time server at home. Use the same thing at work too, but with 5 GPS units in a 'pool', but all running Linux NTP.

I documented it on here before, give a quick search. I think there's even a link to the NTP documentation that tells you how to configure the server line for the different hardware devices.

1 members found this post helpful.

I'm definitely going to try the GPS idea. My company won't give up the cash to buy one of the professional ones, though I'm sure I can get enough money to buy two or 3 simple GPS units.

Thanks a bunch.

No problems, glad to have helped (if I did. ). The thread I did on this is here:
http://www.linuxquestions.org/questi...th-gps-824129/

Obviously, if you're not using a bluetooth GPS, just replace the bluetooth device ID with the USB one, such as /dev/ttyUSB0. A note though...I've seen some instances where the GPS software expects to find a /dev/gps0. If that's the case (may not be any longer), you can just easily make a symlink.