[ntp:questions] ntp synchronisation lost

RichardR randjunk at gmail.com
Wed Mar 9 09:52:27 UTC 2005


Hi.
I know you guys have surely already seen these ntp errors occured. I
have read threads from some of you. So I have done things that have
been told but it still occurs. I am using a v20z sunfire product and
running on RHEL/WS3 using kernel 2.4.21-20.ELsmp.
As you can guess by seeing my logs and some ntpq results, my problems
are about losing synchronisation with our internal ntp server.
looking on my 'tickadj', it seems to be correct '10000' but we when I
run 'calc_tickadj' script (calculating drift from tickadj. perl-script
I have taken on the net) I just find out that my /etc/ntp/drift always
differed and is never synchronized with the given result of the
calculation. I have read in this mailing list a reply from David
Wolley that, it might be a hardware problem or processes may trying to
set the time from the CMOS clock.
I just dont know what to do now guys. I am definitely sure its not NTP
problem. but actually how could I bug this out? forums answers are too
basics and some of them are too foggy. so can I say its a hardware
problem? kernel problem because of an unstable patch? or something
else I have tested before I open an incident to
return my product back.

thanks in advance for your help guys.
Cheers,

here are my results and tests:
-- calc_tickadj.pl
-- http://darwinsource.opendarwin.org/10.0.3/ntp-13/ntp/scripts/calc_tickadj
#!/usr/bin/env perl
#
# drift of 104.8576 -> +1 tick.  Base  of 10000 ticks.
#
# 970306 HMS Deal with nanoseconds.  Fix sign of adjustments.
$df="/etc/ntp/drift";
# Assumes a 100Hz box with "tick" of 10000
# Someday, we might call "tickadj" for better values...
$base=10000;    # tick: 1,000,000 / HZ
$cvt=104.8576;  # 2 ** 20 / $base
$v1=0.;
$v2="";
if (open(DF, $df))
  {
    if ($_=<DF>)
      {
        ($v1, $v2) = split;
      }
    while ($v1 < 0)
      {
        $v1 += $cvt;
        $base--;
      }
    while ($v1 > $cvt)
      {
        $v1 -= $cvt;
        $base++;
      }
  }
printf("%.3f (drift)\n", $v1);
printf("%d usec; %d nsec\n", $base, ($base + ($v1/$cvt)) * 1000);
~
--
[root at jor log]# ntpq -p
*ntp.cdas.augern .GPS.            1 u  174  128  276    7.812   -3.664   7.812
--
 7 Mar 11:23:12 ntpd[32266]: signal_no_reset: signal 17 had flags 4000000
 7 Mar 11:23:12 ntpd[32263]: running as uid(38)/gid(38) euid(38)/egid(38).
 7 Mar 11:27:39 ntpd[32263]: kernel time discipline status change 41
 7 Mar 11:28:43 ntpd[32263]: kernel time discipline status change 1
 7 Mar 11:31:46 ntpd[32263]: ntpd exiting on signal 15
 7 Mar 11:31:46 ntpd[629]: signal_no_reset: signal 17 had flags 4000000
 7 Mar 11:31:46 ntpd[626]: running as uid(38)/gid(38) euid(38)/egid(38).
 7 Mar 11:36:09 ntpd[626]: kernel time discipline status change 41
 7 Mar 11:37:25 ntpd[626]: kernel time discipline status change 1
 7 Mar 13:32:16 ntpd[626]: time reset -0.225222 s
 7 Mar 13:32:16 ntpd[626]: synchronisation lost
 7 Mar 13:47:26 ntpd[626]: time reset 0.248564 s
 7 Mar 13:47:26 ntpd[626]: synchronisation lost
 7 Mar 14:51:07 ntpd[626]: time reset -0.232833 s
 7 Mar 14:51:07 ntpd[626]: synchronisation lost
 7 Mar 15:06:22 ntpd[626]: time reset 0.271954 s
 7 Mar 15:06:22 ntpd[626]: synchronisation lost
 7 Mar 15:45:11 ntpd[626]: time reset -0.239905 s
...
--
[root at jor log]# ntpq -c rv
status=0684 leap_none, sync_ntp, 8 events, event_peer/strat_chg,
version="ntpd 4.1.2 at 1.892 Tue Feb 24 06:31:19 EST 2004 (1)",
processor="x86_64", system="Linux2.4.21-20.ELsmp", leap=00, stratum=2,
precision=-7, rootdelay=7.812, rootdispersion=32.742, peer=308,
refid=ntp.cdas.augernet,
reftime=c5d941ab.58abd5dc  Wed, Mar  9 2005  9:34:35.346, poll=7,
clock=c5d94202.890385c6  Wed, Mar  9 2005  9:36:02.535, state=4,
offset=3.039, frequency=-133.287, jitter=11.930, stability=21.915

-- 
Richard RANDRIA
CNRS/IN2P3/LPNHE Jussieu - Paris VI
IT Soft/System Engineer Researcher
--



More information about the questions mailing list