[ntp:questions] how to read and understand ntpq -pcrv output - pre-wiki

Jim Cromie jim.cromie at gmail.com
Mon May 15 18:42:29 UTC 2006

in the hope that its seen as more useful, Im framing questions
in a way Id like to have found reading the wiki.
Once I understand the answers, I can add them, assuming thats desirable.


Shortly after you've got ntpd running, you'll want to diagnose
your setup to see if it looks sane .. optimal.

ntpq is your friend:
#> ntpq -pcrv
     remote           refid      st t when poll reach   delay   offset  
-harpo   3 u  151 1024  377    1.219   -0.124   
+mail01.tjgroup.    2 u  156 1024  377  206.401  -13.005   
*adsl.remco.org  .PG0A.           1 u  172 1024  377  195.164   -5.140   
+ntp.aramiska.ne      2 u  137 1024  377  186.166  -10.974   
-ntp1.us.grundcl    2 u    5 1024  377   99.248  -16.884   
 LOCAL(0)        LOCAL(0)        13 l   33   64  377    0.000    0.000   
assID=0 status=0664 leap_none, sync_ntp, 6 events, event_peer/strat_chg,
version="ntpd 4.2.0a at 1:4.2.0a+stable-8-r Fri Oct 28 15:39:49 CEST 2005 
processor="i586", system="Linux/2.6.17-rc3-mm1-tunehrt-sk", leap=00,
stratum=2, precision=-18, rootdelay=195.164, rootdispersion=41.201,
peer=36230, refid=,
reftime=c8131944.fec918e3  Mon, May 15 2006  8:14:12.995, poll=10,
clock=0xc81319f0.c20913a4, state=4, offset=-6.457, frequency=-0.384,
noise=3.925, jitter=5.886, stability=31.015
#> uptime
 08:21:08 up 20:25,  1 user,  load average: 0.00, 0.00, 0.00

* I have 4 timeserver entries;

* they have somewhat hi delays - not ideal, but a function of the dns 
lookup of *.pool.ntp.org
    (Ive just switched to 0-2.us.ntp.org, but not for these measurements)

* time-source (table) jitter is nice and low
* poll period is 1024, nice and long, indicates ntpd is 'happy' with 
that time-source


? refid - Ive seen other listings with different stuff here,
    PGOA ?
    PPS - pulse per second - usually off an atomic source, maybe GPS
    they mostly look like IPs,
    are they some kind of hierarchy id, ie hash on path of clock-refs / 
system peers?

? delay, offset, jitter are in  milliseconds ?  (if rtfm, I havent seen 
it, pointers welcome)
good values are _______

WRT the state variables, (the -crv part of the above listing)

rootdispersion - lower is better
    maximum error of the local clock relative to the reference clock.
rootdelay - recap of *entry in peers table

? units of each field ?
? noise - lower is better ?  relationship to jitter, wander ?
? stability - higher is better ?

add post-note: for a fuller understanding (re-)read the rfc

? *frequency* ?    I have special interest here ..

this var tells how much ntpd has pushed your timesource away from
its natural free-running frequency.

Ive written a 'clocksource' driver for linux-2.6.17.rc3-mm1 to use a 32 
bit counter,
attached to a 27 mhz XO, and Ive been 'sanity' checking it with ntpd.

Ive found that ntp said my frequency=18.fuzz, so it looked like my clock 
ran fast.
To test what I could, I included a 'ppm' mod-param to tell the driver
the actual frequency of the XO.  (in integer hz correction to default 
1.000 mhz)

Before setting ppm = -18,  ntpq -pcrv reported a freq of -19.fuzz, and 
now its -0.384.  
Allowing for temp diffs over the duration of the tests ( 20 hrs)
that seems to match nicely with expectations, but Id appreciate hearing 
you all agree with my interpretation of things (I could be biased;)

thanks in advance
Jim Cromie

More information about the questions mailing list