[ntp:questions] large dispersion

Richard B. Gilbert rgilbert88 at comcast.net
Wed May 30 17:05:06 UTC 2007


Dave wrote:
> On May 25, 9:28 am, Dave <dave.deconsult... at gmail.com> wrote:
> 
>>On May 24, 8:09 pm, "Richard B. Gilbert" <rgilber... at comcast.net>
>>wrote:
>>
>>
>>
>>
>>>Dave wrote:
>>>
>>>>On May 23, 1:16 pm, "Richard B. Gilbert" <rgilber... at comcast.net>
>>>>wrote:
>>>
>>>>>Dave wrote:
>>>>
>>>>>>I'm experiencing large values of dispersion when I use 'ntpq -p':
>>>>>
>>>>>>3:22pm:ntp>ntpq -p
>>>>>>    remote           refid      st t when poll reach   delay
>>>>>>offset    disp
>>>>>>==============================================================================
>>>>>>*10.2.100.10     .GPS.            1 u   25   64  377     0.46
>>>>>>943.557  439.36
>>>>>>3:22pm:ntp>
>>>>>
>>>>>>but when I look at the peerstats log file, my dispersion is low:
>>>>>
>>>>>>3:22pm:ntp>tail /var/ntp/ntpstats/peerstats.log
>>>>>>54243 54804.018 10.2.100.10 9634 0.617563 0.00046 0.31154
>>>>>>54243 54868.013 10.2.100.10 9634 0.813226 0.00043 0.32210
>>>>>>54243 54932.018 10.2.100.10 9634 0.099169 0.00046 0.55225
>>>>>>54243 54996.010 10.2.100.10 9634 0.300057 0.00047 0.28690
>>>>>>54243 55060.014 10.2.100.10 9634 0.584358 0.00044 0.31111
>>>>>>54243 55124.009 10.2.100.10 9634 0.781632 0.00047 0.32353
>>>>>>54243 55188.014 10.2.100.10 9634 0.065932 0.00043 0.55287
>>>>>>54243 55252.013 10.2.100.10 9634 0.657674 0.00044 0.37859
>>>>>>54243 55316.018 10.2.100.10 9634 0.943557 0.00046 0.43936
>>>>>>54243 55380.013 10.2.100.10 9634 0.139388 0.00044 0.59772
>>>>>>3:23pm:ntp>
>>>>>
>>>>>>I'm also concerned with the large amount of maximum/estimated error
>>>>>>when I use the xntpdc command 'kerninfo':
>>>>>
>>>>>>xntpdc> kerninfo
>>>>>>pll offset:           0 us
>>>>>>pll frequency:        -391.137 ppm
>>>>>>maximum error:        524704 us
>>>>>>estimated error:      408368 us
>>>>>>status:               0089
>>>>>>pll time constant:    2
>>>>>>precision:            1 us
>>>>>>frequency tolerance:  512 ppm
>>>>>>pps frequency:        0.000 ppm
>>>>>>pps stability:        512.000 ppm
>>>>>>pps jitter:           200 us
>>>>>>calibration interval: 4 s
>>>>>>calibration cycles:   0
>>>>>>jitter exceeded:      0
>>>>>>stability exceeded:   0
>>>>>>calibration errors:   0
>>>>>>xntpdc>
>>>>>
>>>>>>Here is the output of the pstats command:
>>>>>
>>>>>>xntpdc> pstats 10.2.100.10
>>>>>>remote host:          10.2.100.10
>>>>>>local interface:      10.2.100.5
>>>>>>time last received:   3s
>>>>>>time until next send: 61s
>>>>>>reachability change:  2947s
>>>>>>packets sent:         59
>>>>>>packets received:     59
>>>>>>bad authentication:   0
>>>>>>bogus origin:         0
>>>>>>duplicate:            0
>>>>>>bad dispersion:       15
>>>>>>bad reference time:   0
>>>>>>candidate order:      1
>>>>>>xntpdc>
>>>>>
>>>>>>And lastly, my ntp.conf file:
>>>>>
>>>>>>3:25pm:inet>more ntp.conf
>>>>>>server 10.2.100.10             # NTP server
>>>>>
>>>>>>driftfile /etc/ntp.drift          # Drift available for next restart
>>>>>>logfile /var/ntp/ntp.log       # NTP logging
>>>>>
>>>>>>statsdir /var/ntp/ntpstats/
>>>>>>statistics loopstats peerstats clockstats
>>>>>>filegen loopstats file loopstats.log type day link enable
>>>>>>filegen peerstats file peerstats.log type day link enable
>>>>>>filegen clockstats file clockstats.log type day link enable
>>>>>>3:25pm:inet>
>>>>>
>>>>>>Anyone have any ideas? I have a GPS signal coming in to a Brandywine
>>>>>>NTA-100, which is configured at 10.2.100.10. Thanks in advance!
>>>>>
>>>>>If you are using "X"ntpdc you would appear to be using a version that
>>>>>may be as much as ten years old!   Just what are you using and what are
>>>>> you running it on?
>>>>
>>>>Wow you're right. Heres the version printout:
>>>>xntpdc 3-5.93e Mon Sep 20 15:47:24 PDT 1999 (1)
>>>
>>>>I'm running this on a Sun Fire 4200, Solaris 10. Something this old
>>>>comes installed on Solaris 10?
>>>
>>>Yup!
>>
>>>I think it has something to do with the fact that there is, as yet, no
>>>RFC for NTP V4.  There is a committee, God help us, working on one.  I
>>>think it has been about a year now with no visible results!
>>
>>>I'd suggest grabbing a more recent version of the code from somewhere.
>>>Sun Freeware and Blastwave sites should both have Solaris versions more
>>>recent than what Sun ships.  There have been quite a few fixes and
>>>enhancements since 3-5.93e.
>>
>>>If you want/are able to build your own, try the ntp.org web site; there
>>>are links there to download the source to the stable and development
>>>versions which are, I believe, at 4.2.something.
>>
>>I just downloaded ntp v4.2.4 from sunfreeware.com and I'll give that a
>>try. I'll let you know how it goes!
> 
> 
> 
> Ok so I have the new version loaded and its been running for about 24
> hours now. I still have a large amount of jitter (reported on by ntpq -
> p):
> 2:50pm:freadd_user>ntpq -p
>      remote           refid            st t when poll   reach
> delay    offset     jitter
> ========================================================
> *10.2.100.10     .GPS.            1 u    3    64     377     0.461
> 91.914   385.406
> 
<snip>
> 
> Do I need to worry about the jitter value from ntpq -p, or does
> everything look okay?
> 

Looking a little more carefully at your ntpq banner it looks as if the 
delay (round trip delay) to your server is unreasonably large.  You're 
using an RFC-1918 address which is a private network but the delay 
suggests that the server is a couple of thousand miles away or that you 
are using tin cans and string to communicate with it.

The potential error in transmitting the time from server to client is 
one half the round trip delay.  This, in turn, suggests that minimizing 
the round trip delay is a good idea.  I don't know how to account for a 
delay like this in what is presumably a small LAN.  Can you explain it?





More information about the questions mailing list