Re: clock running fast

[Peter Jeremy]

On Thu, 2004-Dec-30 15:59:10 -0500, Harlan Stenn wrote:
>> Remove /var/db/ntpd.drift before you reboot or restart your ntpd.
>
>Why would you want to do this?

On a number of occasions, I've had the ntpd PLL start oscillating and
winding up saturated at +500ppm or -500ppm.  Removing ntpd.drift lets
it recover.  Admittedly, this was mostly on old versions of ntpd - I'm
not sure if the problem is still present.

-- 
Peter Jeremy
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Harlan Stenn]

I'd appreciate it if somebody would add this information to:

 http://ntp.isc.org/Support/KnownOsIssues

H
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[dtalk-ml]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Kevin Kinsey wrote:
Have you tried all possible values for kern.timecounter.hardware?
#sysctl kern.timecounter.choice
My most recent encounter with this issue, a K6-2/500 on an
Asus P5A mobo, needed "i8254".
Thank you vociferously.  A value of i8254 for kern.timecounter.hardware 
seems to have had the desired effect.

Cheers ... -d
- -- 
David Talkington
[EMAIL PROTECTED]
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.2.6 (FreeBSD)

iD8DBQFB1O/d5FKhdwBLj4sRAkAcAKCkG4sD4+8sWszGSsocj0gpIAwYrQCfVNwV
24g6apBqGznNfM28YjaWZMI=
=P4Pa
-END PGP SIGNATURE-
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[David Magda]

On Dec 30, 2004, at 15:58, Federico Galvez-Durand Besnard wrote:
Your NTPD will never be stable with a wrong localtime setting.
NTP does not care about local time. All values that NTP uses are in 
UTC: local time is a function of the operating system and is not used 
when calculating time values.

Another place to ask questions would be the Usenet group:
comp.protocols.time.ntp
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Kevin Kinsey]

> [ rc.conf and ntp.conf snipped ]
> 
> Thank you for the suggestions.  Unfortunately, I've been through all 
> that, including the rude values of minpoll and maxpoll, using multiple 
> servers, and starting with a fresh drift file.  I'm pretty sure ntpd 
> isn't the problem.  In addition, the hardware clock itself appears to be 
> plenty accurate, as it is always correct within a second or two when I 
> check it directly in BIOS ... and two other 5.3-STABLE hosts on the same 
> network, with the same ntpd configuration, but on different hardware, do 
> not have this problem, which began when I updated (reinstalled) to 
> 5.3-STABLE from 5.2.
> 
> - -- 
> David Talkington
> [EMAIL PROTECTED]

I apparently misposted this earlier, my apologies.

Have you tried all possible values for kern.timecounter.hardware?

#sysctl kern.timecounter.choice

My most recent encounter with this issue, a K6-2/500 on an
Asus P5A mobo, needed "i8254".

HTH,

Kevin Kinsey

___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Federico Galvez-Durand Besnard]

Oops, I meant:
cp  /usr/share/zoneinfo/Etc/GMT+8/etc/localtime
if you want  Pacific Time in GMT format (don't use GMT-8).
Harlan Stenn wrote:
No way.
ntpd operates in gmt/utc only; localtime is completely out of the question.
H
--
 

Hmmm, are you using the same localtime in all machines?
I remember having the same problem several years ago, in 3.x, with a 
server. The clock kept walking. Hardware was OK. It came back to normal 
after setting localtime correctly.
Remember, if you are in Pacific Time you are GMT+8 (some people wrongly 
uses GMT-8), it means you have to add 8h to your localtime in order to 
get GMT. Your NTPD will never be stable with a wrong localtime setting.
Fico//
   

 

___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Harlan Stenn]

No way.

ntpd operates in gmt/utc only; localtime is completely out of the question.

H
--
> Hmmm, are you using the same localtime in all machines?
> I remember having the same problem several years ago, in 3.x, with a 
> server. The clock kept walking. Hardware was OK. It came back to normal 
> after setting localtime correctly.
> Remember, if you are in Pacific Time you are GMT+8 (some people wrongly 
> uses GMT-8), it means you have to add 8h to your localtime in order to 
> get GMT. Your NTPD will never be stable with a wrong localtime setting.
> Fico//
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Federico Galvez-Durand Besnard]

Harlan Stenn wrote:
Remove /var/db/ntpd.drift before you reboot or restart your ntpd.
   

Why would you want to do this?
And in config files, iburst is your friend.
See http://ntp.isc.org/Support/ for more informaion.
H
 

True only when NTPD is running in stable steady state.
In this case, NTPD is not performing... so drift file can be recording 
very messy values.
Better get it from scratch than having a very bad initial condition.
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Harlan Stenn]

> Remove /var/db/ntpd.drift before you reboot or restart your ntpd.

Why would you want to do this?

And in config files, iburst is your friend.

See http://ntp.isc.org/Support/ for more informaion.

H
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Federico Galvez-Durand Besnard]

Hmmm, are you using the same localtime in all machines?
I remember having the same problem several years ago, in 3.x, with a 
server. The clock kept walking. Hardware was OK. It came back to normal 
after setting localtime correctly.
Remember, if you are in Pacific Time you are GMT+8 (some people wrongly 
uses GMT-8), it means you have to add 8h to your localtime in order to 
get GMT. Your NTPD will never be stable with a wrong localtime setting.
Fico//

[EMAIL PROTECTED] wrote:
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Federico Galvez-Durand Besnard wrote:
Set /etc/localtime to your correct timezone before doing anything in 
your machine. Remove /var/db/ntpd.drift before you reboot or restart 
your ntpd.

[ rc.conf and ntp.conf snipped ]
Thank you for the suggestions.  Unfortunately, I've been through all 
that, including the rude values of minpoll and maxpoll, using multiple 
servers, and starting with a fresh drift file.  I'm pretty sure ntpd 
isn't the problem.  In addition, the hardware clock itself appears to 
be plenty accurate, as it is always correct within a second or two 
when I check it directly in BIOS ... and two other 5.3-STABLE hosts on 
the same network, with the same ntpd configuration, but on different 
hardware, do not have this problem, which began when I updated 
(reinstalled) to 5.3-STABLE from 5.2.

- -- David Talkington
[EMAIL PROTECTED]
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.2.6 (FreeBSD)
iD8DBQFB1FwQ5FKhdwBLj4sRArSqAJwK8MAvUfB69ixoHNzu8700Pvd52QCgl0dD
07gb7ipg0ENIcUN/PPHhXpw=
=Lv+T
-END PGP SIGNATURE-
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Harlan Stenn]

> But isn't that what the drift file is for, to improve the accuracy of
> the local clock during those times when the configured NTP servers aren't
> available?

Yes.

H
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Federico Galvez-Durand Besnard]

Brooks Davis wrote:
On Thu, Dec 30, 2004 at 11:37:04AM -0800, Javier Henderson wrote:
 

/usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
- -f /var/db/ntpd.drift
   

It's generally recommended that you never trust your own clock since
it's completely crap as time sources go.  It's also recommended that you
use at least 5 time sources to avoid problems with bad clocks.
 

But isn't that what the drift file is for, to improve the accuracy of
the local clock during those times when the configured NTP servers aren't
available?
   

Yes, but it can only do so much.  The clock crystals in your average PC
are OK as clocks, but much better as thermometers so the drift
calculations are only approximate.
-- Brooks
 

Yeap, but only if your NTPD is running fine and stable, otherwise 
drift-file can be very messy.
Check your dmesg for NTPD log before you trust drift-files.

___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Brooks Davis]

On Thu, Dec 30, 2004 at 11:37:04AM -0800, Javier Henderson wrote:
> > > /usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
> > > - -f /var/db/ntpd.drift
> > 
> > It's generally recommended that you never trust your own clock since
> > it's completely crap as time sources go.  It's also recommended that you
> > use at least 5 time sources to avoid problems with bad clocks.
> 
> But isn't that what the drift file is for, to improve the accuracy of
> the local clock during those times when the configured NTP servers aren't
> available?

Yes, but it can only do so much.  The clock crystals in your average PC
are OK as clocks, but much better as thermometers so the drift
calculations are only approximate.

-- Brooks

-- 
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgp5BEU1OFD1v.pgp
Description: PGP signature

Re: clock running fast

[dtalk-ml]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Federico Galvez-Durand Besnard wrote:
Set /etc/localtime to your correct timezone before doing anything in 
your machine. Remove /var/db/ntpd.drift before you reboot or restart 
your ntpd.
[ rc.conf and ntp.conf snipped ]
Thank you for the suggestions.  Unfortunately, I've been through all 
that, including the rude values of minpoll and maxpoll, using multiple 
servers, and starting with a fresh drift file.  I'm pretty sure ntpd 
isn't the problem.  In addition, the hardware clock itself appears to be 
plenty accurate, as it is always correct within a second or two when I 
check it directly in BIOS ... and two other 5.3-STABLE hosts on the same 
network, with the same ntpd configuration, but on different hardware, do 
not have this problem, which began when I updated (reinstalled) to 
5.3-STABLE from 5.2.

- -- 
David Talkington
[EMAIL PROTECTED]
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.2.6 (FreeBSD)

iD8DBQFB1FwQ5FKhdwBLj4sRArSqAJwK8MAvUfB69ixoHNzu8700Pvd52QCgl0dD
07gb7ipg0ENIcUN/PPHhXpw=
=Lv+T
-END PGP SIGNATURE-
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Federico Galvez-Durand Besnard]

Set /etc/localtime to your correct timezone before doing anything in 
your machine.
Remove /var/db/ntpd.drift before you reboot or restart your ntpd.

This is what I use in /etc/rc.conf:
+++
ntpd_enable="YES"   # Run ntpd Network Time Protocol (or NO).
ntpd_program="/usr/sbin/ntpd"   # path to ntpd, if you want a different one.
ntpd_flags="-p /var/run/ntpd.pid -f /etc/ntp/drift -l /var/log/ntp.log"
+++
Check ntpd messages with dmesg.
If it does not work, this is one ntp.conf (change server to 
time.u.washington.edu or whatever is closer to you) that worked well for me:

+ BEGIN /etc/ntp.conf +++
# NTP configuration file
#
# Thu Nov 11 20:06:40 GMT+1 2004
restrict192.168.1.0 mask 255.255.255.0 notrust nomodify notrap
restrict127.0.0.1
# servers to query
#---
server  clock.tix.ch
restrictclock.tix.chmask 255.255.255.255 nomodify notrap 
noquery

+ END /etc/ntp.conf +++
Once you get synchronised with your time-server, you can consider 
changing minpoll  maxpoll to polite values.

Fico//
David Talkington wrote:
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Federico Galvez-Durand Besnard wrote:
what do you have in :
/etc/ntp.conf

Only:
server time.u.washington.edu
server  127.127.1.0 fudge   127.127.1.0 stratum 10
Command line:
/usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
- -f /var/db/ntpd.drift
Thank you ... -d
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Javier Henderson]

> > /usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
> > - -f /var/db/ntpd.drift
> 
> It's generally recommended that you never trust your own clock since
> it's completely crap as time sources go.  It's also recommended that you
> use at least 5 time sources to avoid problems with bad clocks.

But isn't that what the drift file is for, to improve the accuracy of
the local clock during those times when the configured NTP servers aren't
available?

-jav
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Re: clock running fast

[Brooks Davis]

[Please don't top post].

On Thu, Dec 30, 2004 at 11:13:55AM -0800, David Talkington wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
> 
> Federico Galvez-Durand Besnard wrote:
> 
> >what do you have in :
> >/etc/ntp.conf
> 
> Only:
> 
> servertime.u.washington.edu
> server  127.127.1.0 
> fudge   127.127.1.0 stratum 10
> 
> 
> Command line:
> 
> /usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
> - -f /var/db/ntpd.drift

It's generally recommended that you never trust your own clock since
it's completely crap as time sources go.  It's also recommended that you
use at least 5 time sources to avoid problems with bad clocks.

-- Brooks

> >-BEGIN PGP SIGNED MESSAGE-
> >Hash: SHA1
> >
> >
> >Salutations --
> >
> >I have an old board with an installation of STABLE whose time keeps 
> >running away.  ntpd is not helping.  I can manually set the time using 
> >ntpdate, and my ntpq queries look fine, yet the clock persists in running 
> >very fast (seems to be about 1.5x speed).  The BIOS clock remains correct, 
> >and the machine therefore has the correct time at each reboot, but quickly 
> >runs away.
> >
> >Data points:
> >
> >- - I have set kern.timecounter.hardware=TSC, and this has not helped. - - 
> >Two other hosts on the same network, using the same ntp server, do not 
> >have this problem.
> >- - This box did not display this behavior under FreeBSD 5.2. - - I have 
> >replaced the board on the problem box (with the same age and type), and 
> >the problem persists.
> >- - dmesg for this machine, an old Dell Optiplex 590, is below.
> >
> >What other information might help diagnose the problem?
> >
> >Thank you ... -d
> >
> >- --
> >David Talkington
> >dtalk-ml at prairienet.org 
> >
> >
> >
> >
> >___
> >freebsd-stable@freebsd.org mailing list
> >http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> >To unsubscribe, send any mail to "[EMAIL PROTECTED]"
> >
> >
> 
> - -- 
> David Talkington
> Computing and Communications
> University of Washington
> 206-543-2144
> - --
> [EMAIL PROTECTED]
> - --
> PGP key: http://staff.washington.edu/dtalk/004B8F8B.asc
> -BEGIN PGP SIGNATURE-
> Version: GnuPG v1.2.6 (FreeBSD)
> 
> iD8DBQFB1FN55FKhdwBLj4sRAknTAJ9Wq/I64DpzBoblMmH2JMl2UR08NgCgkLuL
> qlB7DOSkWlOmR2OLZ32O1NA=
> =8B5z
> -END PGP SIGNATURE-
> ___
> freebsd-stable@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "[EMAIL PROTECTED]"
-- 
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgpt56ACz439C.pgp
Description: PGP signature

Re: clock running fast

[David Talkington]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Federico Galvez-Durand Besnard wrote:
what do you have in :
/etc/ntp.conf
Only:
server 		time.u.washington.edu
server  127.127.1.0 
fudge   127.127.1.0 stratum 10

Command line:
/usr/sbin/ntpd -c /etc/ntp.conf -p /var/run/ntpd.pid \
- -f /var/db/ntpd.drift
Thank you ... -d

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Salutations --
I have an old board with an installation of STABLE whose time keeps running 
away.  ntpd is not helping.  I can manually set the time using ntpdate, and my 
ntpq queries look fine, yet the clock persists in running very fast (seems to 
be about 1.5x speed).  The BIOS clock remains correct, and the machine 
therefore has the correct time at each reboot, but quickly runs away.

Data points:
- - I have set kern.timecounter.hardware=TSC, and this has not helped. - - Two 
other hosts on the same network, using the same ntp server, do not have this 
problem.
- - This box did not display this behavior under FreeBSD 5.2. - - I have 
replaced the board on the problem box (with the same age and type), and the 
problem persists.
- - dmesg for this machine, an old Dell Optiplex 590, is below.

What other information might help diagnose the problem?
Thank you ... -d
- --
David Talkington
dtalk-ml at prairienet.org 



___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

- -- 
David Talkington
Computing and Communications
University of Washington
206-543-2144
- --
[EMAIL PROTECTED]
- --
PGP key: http://staff.washington.edu/dtalk/004B8F8B.asc
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.2.6 (FreeBSD)

iD8DBQFB1FN55FKhdwBLj4sRAknTAJ9Wq/I64DpzBoblMmH2JMl2UR08NgCgkLuL
qlB7DOSkWlOmR2OLZ32O1NA=
=8B5z
-END PGP SIGNATURE-
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"