[ntp:questions] Stratum goes back to 16 all the time

Tom Smith smith at cag.zko.hp.com
Sat Dec 30 19:37:58 UTC 2006


independence at blinkenlights.se wrote:
> Hi, I've set up a ntpd on my linux box and at first it worked fine and
> got a stratum of 2 after a while. Then, after a couple of minutes, it
> got a stratum of 16 again. Now it hops back and forth and doesn't seem
> to be very stable at all.
> I sat and watched the output from peers in ntpq for a while, and first
> everything is fine but after a while all the remote servers get *+- etc
> in front of their names, I've tried to read what these mean but I'm
> don't really understand.
> 
> Here's the output from ntpq peers:
> ntpq> peers
>      remote           refid      st t when poll reach   delay   offset
> jitter
> ==============================================================================
>  nissan.ifm.liu. .GPS.            1 u   16   64   17   27.721  -479.06
> 115.831
>  timmy.lysator.l .GPS.            1 u   28   64   17   28.089  -380.49
> 72.967
>  timehost.lysato .GPS.            1 u   40   64   17   28.833  -462.36
> 111.578
>  Time1.Stupi.SE  .PPS.            1 u   23   64   17   24.651  -313.99
> 115.770
>  ntp1.sth.netnod .PPS.            1 u   28   64   17   26.199  -475.23
> 119.470
>  ntp2.gbg.netnod .PPS.            1 u   51   64   17   32.007  -455.40
> 123.874
 >    .
 >    .
 >    .
> ntpq> peers
>      remote           refid      st t when poll reach   delay   offset
> jitter
> ==============================================================================
>  nissan.ifm.liu. .GPS.            1 u  112   64   36   27.721  -479.06
> 115.831
> *timmy.lysator.l .GPS.            1 u   58   64   37   28.089  -380.49
> 118.008
> +timehost.lysato .GPS.            1 u    8   64   77   27.253  -522.91
> 141.296
> -Time1.Stupi.SE  .PPS.            1 u   53   64   37   24.651  -313.99
> 169.327
> +ntp1.sth.netnod .PPS.            1 u   58   64   37   26.199  -475.23
> 116.326
> +ntp2.gbg.netnod .PPS.            1 u   15   64   77   31.388  -633.99
> 250.372
> 
> Why might does this happen?
> 
> ntptrace shows this:
> localhost: stratum 16, offset 0.000000, synch distance 0.010290
> 

You are starting with an offset of almost 1/2 second from your
sources. After 5 polls of the sources, your client selects a set
of servers to use as a reference. Those are the "*" and "+"
characters that appear in the first column. The "-" character
indicates a source that was rejected for one reason or another.
The source with an "*" is the "synchronization source".

After a period of time during which the consensus offset is
deemed to be stable, the client will reset its clock to
try to bring the offset to 0. When this happens, the client
stratum goes back to 16 (unsynchronized) until it synchronizes
again. The same synchronization sequence then repeats
indefinitely.

You should expect this to happen repeatedly for several
hours or even for a day or 2 until your client computes its
characteristic drift rate and is able to cruise along
between polls at approximately the same rate as the
clocks you are using as sources.

You can accelerate the process as follows:

1) Stop ntpd
2) Add "iburst" to the end of each of your server lines in ntp.conf
3) make sure you have a (valid) driftfile declaration in ntp.conf
4) Delete any existing ntp.drift file
5) Use "ntpdate -b [server]" to set your clock to the
    correct time. This should be part of your boot or ntpd start sequence.
    It is configured differently for different Linux distributions.
    Red Hat, for example, uses the file /etc/ntp/step-tickers to list
    the servers that ntpdate should use on initial startup.
6) Start ntpd.
7) Let ntpd run for at least a day without disturbing it

-Tom




More information about the questions mailing list