f



advice on hardware/OS/refclock for NTP server

Greetings,

I'd like to assemble an NTP server and need some advice on the hardware 
and operating system.  FreeBSD and Linux have kernel PLL code, so would 
it be safe so say they would be the optimal OSes to run NTP?

What GPS receiver or interface card, performs best off the shelf?

Any motherboard recommendations?  Asus, or Intel make use particularly 
good crystals?

Given that maintainability, security, reliability, and cost are factors, 
what do the experts suggest?

Thanks,
Bill

0
William
7/4/2003 1:18:33 AM
comp.protocols.time.ntp 4895 articles. 2 followers. Post Follow

16 Replies
995 Views

Similar Articles

[PageSpeed] 32

William Holden wrote:

> I'd like to assemble an NTP server and need some advice on the hardware 
> and operating system.  FreeBSD and Linux have kernel PLL code, so would 
> it be safe so say they would be the optimal OSes to run NTP?

I'd say FreeBSD.

> What GPS receiver or interface card, performs best off the shelf?

http://www.synergy-gps.com/

> Any motherboard recommendations?  Asus, or Intel make use particularly 
> good crystals?

See below -- the stock xtals on the Soekris are excellent.  They are,
however, also excellent thermometers.  You could replace the 33.333MHz
xtal with a 32.768MHz OCXO if you -- for some reason -- need microsecond
accuracy.

> Given that maintainability, security, reliability, and cost are factors, 
> what do the experts suggest?

http://phk.freebsd.dk/soekris-pps/

http://www.soekris.com/

If you need to place the GPS/antenna far from the box (not likely),
you'll need an RS-422 card instead of the onboard RS-232 on the Soekris.

http://www.quatech.com/catalog/rs422_upci.php

0
Michael
7/4/2003 1:45:17 AM
In article <azWdnfgBBtSzQZmiU-KYvg@speakeasy.net>,
	Michael Sierchio <kudzu@tenebras.com> writes:
> William Holden wrote:
> 
>> I'd like to assemble an NTP server and need some advice on the hardware 
>> and operating system.  FreeBSD and Linux have kernel PLL code, so would 
>> it be safe so say they would be the optimal OSes to run NTP?
> 
> I'd say FreeBSD.

Except that FreeBSD does not support mlock calls so that if you run
low on real memory and start swapping then everything will degrade.


>> What GPS receiver or interface card, performs best off the shelf?
> 
> http://www.synergy-gps.com/

Where in the world are you?


>> Any motherboard recommendations?  Asus, or Intel make use particularly 
>> good crystals?
> 
> See below -- the stock xtals on the Soekris are excellent.  They are,
> however, also excellent thermometers.  You could replace the 33.333MHz
> xtal with a 32.768MHz OCXO if you -- for some reason -- need microsecond
> accuracy.
> 
>> Given that maintainability, security, reliability, and cost are factors, 
>> what do the experts suggest?
> 

You need more than one time source. If you are in range of some
of the LW radioclocks then

    http://www.buzzard.org.uk/radioclock.html

make excellent low cost receivers. I have had two (one MSF one DCF77)
running for 18 months without any trouble whatsoever.

JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan at buzzard.me.uk
Northumberland, United Kingdom.       Tel: +44 1661-832195
0
jonathan
7/4/2003 8:33:06 AM
Jonathan Buzzard wrote:

> Except that FreeBSD does not support mlock calls so that if you run
> low on real memory and start swapping then everything will degrade.

Why make assertions on a subject you know nothing about?

Absolutely false, wrong, stupidly misinformed.  FreeBSD does not
implement mlockall(), which even in the case of Linux does not
offer an ironclad guarantee.  See the archives.

MLOCK(2)                  FreeBSD System Calls Manual                 MLOCK(2)

NAME
      mlock, munlock - lock (unlock) physical pages in memory

LIBRARY
      Standard C Library (libc, -lc)

SYNOPSIS
      #include <sys/types.h>
      #include <sys/mman.h>

      int
      mlock(const void *addr, size_t len);

      int
      munlock(const void *addr, size_t len);

DESCRIPTION
      The mlock() system call locks into memory the physical pages associated
      with the virtual address range starting at addr for len bytes.

Modern operating systems don't swap, they page.

In any case, you're not well-informed.  My stratum 1 server is a Soekris
net4501 box that runs with a read-only root filesystem and no swap.

It's cheap enough to dedicate to timekeeping.

> You need more than one time source.

24 satellites?

0
Michael
7/4/2003 4:57:03 PM
I'll second (no pun intended) that. I get <1ms stability which would 
probably be better if my local crystal was not suffering from
fits. When well behaved it is -17,4ppm at around 20�C. THat's
a long long way from that available from the cheapest TCXO, but
I can't see how to safely make the swap on the main board.

0
mike
7/4/2003 7:58:43 PM
In article <sRydnRczEKN9LJiiXTWJjQ@speakeasy.net>,
	Michael Sierchio <kudzu@tenebras.com> writes:
> Jonathan Buzzard wrote:
> 
>> Except that FreeBSD does not support mlock calls so that if you run
>> low on real memory and start swapping then everything will degrade.
> 
> Why make assertions on a subject you know nothing about?

Because I have had several emails discussion on exactly this topic,
from people wanting to port my radioclkd program to FreeBSD last year
and finding that mlockall is not supported that is why.  Yes it exists
but it is a stup that does nothing.

If this has changed in the very latest FreeBSD then I stand corrected,
but it is not something I know nothing about.


> Absolutely false, wrong, stupidly misinformed.  FreeBSD does not
> implement mlockall(), which even in the case of Linux does not
> offer an ironclad guarantee.  See the archives.

Well I disagree totally, it is not absolutely false, wrong or
stupidly misinformed.


[SNIP]

> Modern operating systems don't swap, they page.

You are being pedantic here.

> In any case, you're not well-informed.  My stratum 1 server is a Soekris
> net4501 box that runs with a read-only root filesystem and no swap.
> 
> It's cheap enough to dedicate to timekeeping.

Not everyone has the space to dedicate a whole machine to time keeping,
or is happy wasting the energy required to run a machine for just
time keeping either.


>> You need more than one time source.
> 
> 24 satellites?
> 

Go read the NTP FAQ's. You do not have 24 satellites, you have a
single GPS receiver. These can and do have hardware and firmware
bugs/faults that can cause them to pass false time information into
ntpd. If ntpd has no other means to verify the time then it will
start passing this bad time information onto other machines.

This is why it is recommened that you have three time sources
so that if one goes bad, the other two can vote it down.

JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan at buzzard.me.uk
Northumberland, United Kingdom.       Tel: +44 1661-832195
0
jonathan
7/4/2003 11:05:06 PM
Jonathan Buzzard wrote:

> This is why it is recommened that you have three time sources
> so that if one goes bad, the other two can vote it down.

I don't want to join the rest of this "debate" but want to comment on 
this particular item.

I've seen stability problems with only 3 NTP sources. I've found that 4 
works better and even more works better still.

IMO NTP is quite "fragile" in terms of how it behaves if the input 
sources have very much offset or jitter at all. I often see poor 
behavior over my 768/128 DSL line. This is stock NTP under various 
recent versions of FreeBSD 4.x. I don't know what version, I just run 
what comes with the OS.


0
Bohdan
7/5/2003 7:07:56 PM
Jonathan Buzzard wrote:

>>>Except that FreeBSD does not support mlock calls so that if you run
>>>low on real memory and start swapping then everything will degrade.
>>
>>Why make assertions on a subject you know nothing about?
> 
> 
> Because I have had several emails discussion on exactly this topic,
> from people wanting to port my radioclkd program to FreeBSD last year
> and finding that mlockall is not supported that is why.  Yes it exists
> but it is a stup that does nothing.

You made the assertion that FreeBSD does not support mlock.  This
is simply false.  'mlockall()' is not supported,  but this is
mostly out of intellectual honesty -- no OS supports mlockall()
in such a way as to honor the guarantee that it seems to make.

>>Modern operating systems don't swap, they page.
 >
> You are being pedantic here.

And you are in need of instruction, so it seemed appropriate.

> Not everyone has the space to dedicate a whole machine to time keeping,
> or is happy wasting the energy required to run a machine for just
> time keeping either.

There's no waste of energy -- either electrical or otherwise. It's
a bulletproof appliance that draws less than 10 Watts, and may be
place atop a pole in an enclosure with the GPS unit, with all the
necessary power supplied via PoE.

> Go read the NTP FAQ's. You do not have 24 satellites, you have a
> single GPS receiver. These can and do have hardware and firmware
> bugs/faults that can cause them to pass false time information into
> ntpd. If ntpd has no other means to verify the time then it will
> start passing this bad time information onto other machines.

It makes little sense to attach more than one reference clock to
a machine, however.  It makes more sense to deploy multiple
timekeepers and have the stratum 1 clocks peers with each
other.

0
Michael
7/5/2003 8:08:50 PM
In article <GHSdnUQUAovOrZqiXTWJgA@speakeasy.net>,
	Michael Sierchio <kudzu@tenebras.com> writes:
> Jonathan Buzzard wrote:
> 
>>>>Except that FreeBSD does not support mlock calls so that if you run
>>>>low on real memory and start swapping then everything will degrade.
>>>
>>>Why make assertions on a subject you know nothing about?
>> 
>> 
>> Because I have had several emails discussion on exactly this topic,
>> from people wanting to port my radioclkd program to FreeBSD last year
>> and finding that mlockall is not supported that is why.  Yes it exists
>> but it is a stup that does nothing.
> 
> You made the assertion that FreeBSD does not support mlock.  This
> is simply false.  'mlockall()' is not supported,  but this is
> mostly out of intellectual honesty -- no OS supports mlockall()
> in such a way as to honor the guarantee that it seems to make.

Well pardon me if I made a small typing mistake. Further more Linux
makes a very good job of keeping the program in memory. Without
mlockall my radioclkd program shows lots of jitter under heavy
memory usage, with mlockall, it works a treat.

Let's face it a small program only depending on libc can easily
be locked into memory.


>>>Modern operating systems don't swap, they page.
> >
>> You are being pedantic here.
> 
> And you are in need of instruction, so it seemed appropriate.

99.9% of people refer to the process of paging in passing as swaping,
if nothing from historical usage.

>> Not everyone has the space to dedicate a whole machine to time keeping,
>> or is happy wasting the energy required to run a machine for just
>> time keeping either.
> 
> There's no waste of energy -- either electrical or otherwise. It's
> a bulletproof appliance that draws less than 10 Watts, and may be
> place atop a pole in an enclosure with the GPS unit, with all the
> necessary power supplied via PoE.

10W is 10W, if it is not needed.

>> Go read the NTP FAQ's. You do not have 24 satellites, you have a
>> single GPS receiver. These can and do have hardware and firmware
>> bugs/faults that can cause them to pass false time information into
>> ntpd. If ntpd has no other means to verify the time then it will
>> start passing this bad time information onto other machines.
> 
> It makes little sense to attach more than one reference clock to
> a machine, however.  It makes more sense to deploy multiple
> timekeepers and have the stratum 1 clocks peers with each
> other.

Not quite true. It is more robust to have three or more machines,
but to stop false times getting onto the system more than one
clock on a machine is perfectly acceptable. Sure if you have three
clocks on a machine and it dies, then you have no time source but
at least you are not going to have bad time stamps. The machines
will just start drifting till you bring it back online.

Have three seperate identical machines, and some firmware bug and
you are going to have bad time stamps.

JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan at buzzard.me.uk
Northumberland, United Kingdom.       Tel: +44 1661-832195
0
jonathan
7/5/2003 10:14:57 PM
Just out of curiosity:  If the accuracy of PC hardware clocks varies so
much with temperature, couldn't ntpd analyze the drift vs. temperature
(assuming the hardware had a temperature sensor for the CPU case or
something) and further refine corrections for the clock in consequence?

I assume that clock drift induced by temperature varies smoothly with
that temperature, and so it seems some sort of tracking should be
possible, and it might improve the free-running accuracy of time on the
server substantially, increasing the autonomy of the server if a more
accurate reference is not available.

-- 
Transpose hotmail and mxsmanic in my e-mail address to reach me directly.
0
Mxsmanic
7/6/2003 9:55:33 AM
In article <n6sfgv0pcinrq480re86615jgn53gf3f20@4ax.com>,
	Mxsmanic <mxsmanic@hotmail.com> writes:
> Jonathan Buzzard writes:
> 
>> Well pardon me if I made a small typing mistake.
> 
> It would require quite a coincidental cascade of typing mistakes to make
> such an assertion.

No I typed mlock instead of mlockall, hardly a large mistake.

>> 99.9% of people refer to the process of paging
>> in passing as swaping, if nothing from historical usage.
> 
> Paging and swapping are not the same thing.

I am perfectly well aware of this. However the process is often
referred to as swapping, mainly for historical reasons. You tend
to say things like "my machine is swapping like mad" etc.

I will to further hammer the point home that he was being a pedantic
idiot that on Linux, it "pages" as he puts to a "swap area". From
"man swapon" 

    Swapon is used to specify devices on which paging and
    swapping are to take place.

    Swapoff disables swapping on the specified devices and
    files, or on all swap entries in /etc/fstab when the -a
    flag is given.

> Paging involves segments of a process.  Swapping involves the entire
> process.  Swapping is the only option on systems without hardware
> support for paging.  On systems with such support, swapping is still an
> option, but only one of several; many operating systems designed to run
> on paging hardware using both paging and swapping.

Well that is one take on the issue. I could equally argue that it
is "swapping" of pages and the entire process is can still correctly
be called swapping, and paging is just one type of swapping.

I am sure if I trawlled the lkml I could find Linus or Alan Cox
or some other notible kernel hacker refering to it as swapping.


JAB.

-- 
Jonathan A. Buzzard                 Email: jonathan at buzzard.me.uk
Northumberland, United Kingdom.       Tel: +44 1661-832195
0
jonathan
7/6/2003 3:35:03 PM
"Maarten Wiltink" <maarten@kittensandcats.net> wrote in message news:<3f082573$0$49111$e4fe514c@news.xs4all.nl>...
> Tim Shoppa wrote in message ...
> >Bohdan Tashchuk <tashchuk@easystreet.com> wrote in message
>  news:<vge8g6arifjkad@corp.supernews.com>...
> >>            ...my 768/128 DSL line.
>  [...]
> >The NTP design implicitly assumes symmetric network delays, yet your
> >network connection advertises itself by how asymmetric it is :-(.
> 
> 
> Those numbers are asymmetric bandwidth. Nothing is said about the delay.
> In theory, I suspect the technology is the same both ways, and the only
> difference is in the throttling at the ISP, so as long as your line isn't
> saturated, it should have the same delay both ways. In theory anyway.

Well, yeah, that's the assumption of NTP.  Maybe if you
have more download bandwidth than upload bandwidth AND you really download
more than you upload everything may work out well.  Reverse the load
situation and then it's not so good.

If your round trip delays get much above the step threshold offset
(by default 128ms?) then you have trouble brewing.  Large downloads
over modems or ADSL or SDSL will give you round-trip-times several
times this.

In practice one of the few ways to measure asymmetric delays in
both directions (as opposed to simple round-trip) is to set
both nodes up with good stratum-0 references and compare timestamps
both ways.  Some folks have been doing this for years... see

 http://www.ripe.net/test-traffic/

Tim.
0
shoppa
7/6/2003 10:15:47 PM
>I'll second (no pun intended) that. I get <1ms stability which would 
>probably be better if my local crystal was not suffering from
>fits. When well behaved it is -17,4ppm at around 20�C. THat's
>a long long way from that available from the cheapest TCXO, but
>I can't see how to safely make the swap on the main board.

The absolute value of the drift doesn't matter much.  ntpd is
very good at correcting for that.

What does matter is the temperature coefficient and/or the temperature
stability of the environment where your system is running.

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

0
hmurray
7/7/2003 4:29:29 PM
>Just out of curiosity:  If the accuracy of PC hardware clocks varies so
>much with temperature, couldn't ntpd analyze the drift vs. temperature
>(assuming the hardware had a temperature sensor for the CPU case or
>something) and further refine corrections for the clock in consequence?

http://www.ijs.si/time/temp-compensation/


>I assume that clock drift induced by temperature varies smoothly with
>that temperature, and so it seems some sort of tracking should be
>possible, and it might improve the free-running accuracy of time on the
>server substantially, increasing the autonomy of the server if a more
>accurate reference is not available.

The hard part is measuring the temperature of the crystal.  (rather
than the CPU)

The crystal itself is reasonably linear within a small temperature
range.  If you are interested in a wide enough temperature range,
then the curve becomes second or third order.

Manfacturers/designers have some control over the curve by
picking the angle that the block is cut relative to the axis
of the physical quartz "crystal".

Sometimes you can get data from manufacturers.  Sometimes
they cut things so there is a flat spot where you want it.

Sometimes they will only tell you "x PPM over initial
manufacturaring error and temperature and voltage".
They often make the curves have flat spots a long way from
normal operations since that fits the specs.  That usually
puts a steep curve where you would like a flat spot.

http://www.toyocom.co.uk/crystal/appcry.htm
http://www.corningfrequency.com/library/isaac.html
http://www.mancini99.freeserve.co.uk/Omega_folder.html

There is lots and lots of data/info on this area.

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

0
hmurray
7/8/2003 6:27:00 AM
Hal Murray writes:

> The hard part is measuring the temperature of the
> crystal.  (rather than the CPU)

If you have the case temperature, as many PCs do, wouldn't that track
the crystal temperature fairly closely?  The CPU temperature certainly
would not, but the case temperature overall probably would.

-- 
Transpose hotmail and mxsmanic in my e-mail address to reach me directly.
0
Mxsmanic
7/8/2003 7:32:27 AM
This is a multi-part message in MIME format.
--------------070407010609070702060504
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Mxsmanic wrote:

>Just out of curiosity:  If the accuracy of PC hardware clocks varies so
>much with temperature, couldn't ntpd analyze the drift vs. temperature
>(assuming the hardware had a temperature sensor for the CPU case or
>something) and further refine corrections for the clock in consequence?
>
>I assume that clock drift induced by temperature varies smoothly with
>that temperature, and so it seems some sort of tracking should be
>possible, and it might improve the free-running accuracy of time on the
>server substantially, increasing the autonomy of the server if a more
>accurate reference is not available.
>
>  
>
Since most PCs, especially those being used for a time server, live in 
relatively stable environments and since the clocks in the PC are only 
necessary when you've lost GPS sync for extended periods of time, and 
since the PC is generally not moving a good location for the GPS antenna 
should not be difficult to find:

    * The GPS should not go out of sync much
    * The software should calibrate the internal clocks to the GPS
      signals to determine their true clock rate
    * And since the temperature should not change much the clock
      frequencies should remain stable

Using software to calibrate the internal clocks and stable temperatures 
I can easily keep my PC to within 100 us of drift per day.  If you're 
loosing GPS signal for more than an hour, you've got major problems with 
your GPS.

So, the stability and accuracy of the PC's on-board clocks are not that 
critical unless you're using it in an unusual situation.

Mike


--------------070407010609070702060504
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
  <meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
  <title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
Mxsmanic wrote:<br>
<blockquote type="cite"
 cite="midhcsfgvkq5khoc0lqbd11usc8r8op091nbe@4ax.com">
  <pre wrap="">Just out of curiosity:  If the accuracy of PC hardware clocks varies so
much with temperature, couldn't ntpd analyze the drift vs. temperature
(assuming the hardware had a temperature sensor for the CPU case or
something) and further refine corrections for the clock in consequence?

I assume that clock drift induced by temperature varies smoothly with
that temperature, and so it seems some sort of tracking should be
possible, and it might improve the free-running accuracy of time on the
server substantially, increasing the autonomy of the server if a more
accurate reference is not available.

  </pre>
</blockquote>
Since most PCs, especially those being used for a time server, live in
relatively stable environments and since the clocks in the PC are only
necessary when you've lost GPS sync for extended periods of time, and
since the PC is generally not moving a good location for the GPS
antenna should not be difficult to find:<br>
<ul>
  <li>The GPS should not go out of sync much</li>
  <li>The software should calibrate the internal clocks to the GPS
signals to determine their true clock rate</li>
  <li>And since the temperature should not change much the clock
frequencies should remain stable <br>
  </li>
</ul>
Using software to calibrate the internal clocks and stable temperatures
I can easily keep my PC to within 100 us of drift per day.&nbsp; If you're
loosing GPS signal for more than an hour, you've got major problems
with your GPS.<br>
<br>
So, the stability and accuracy of the PC's on-board clocks are not that
critical unless you're using it in an unusual situation.<br>
<br>
Mike<br>
<br>
</body>
</html>

--------------070407010609070702060504--

0
Michael
7/8/2003 8:16:36 PM
"Michael Klos" <mike@klos.com> wrote in message
news:EEFOa.11851$EQ5.11037@twister.nyroc.rr.com...

> So, the stability and accuracy of the PC's on-board clocks are not that
> critical unless you're using it in an unusual situation.

    If the GPS only provides a pulse every second, what do you think the
time server uses to tell at what point between two seconds you are?
Short-term stability of the PC clock is *very* important.

    DS


0
David
7/8/2003 8:48:11 PM
Reply:

Similar Artilces:

can AIX 5.x pull NTP time from a win2000 server with the NTP server enabled?
We have a win2k sever connected via serial timetap to a spectracom netclock/2. can an AIX 5.1 server pull NTP time from this..if so how. TIA. me@home.com wrote: > We have a win2k sever connected via serial timetap to a spectracom > netclock/2. can an AIX 5.1 server pull NTP time from this..if so how. > TIA. Yes, it can, but you'd probably prefer to go the opposite way, or connect that Spectracom clock to a surplus Pentium running FreeBSD. Using NT, if that clock isn't directly supported by NTP on NT, then you'd need to configure a local clock, and use the supplied ...

NTP client syncs successfully with NTP server within local network but not with the NTP pool servers
I am using ntp 4.2.6p5 on linux. I setup an NTP server on a machine in another subnet and tried to sync my machine with the server and it works well. But when I try to use the pool.ntp.org servers (I tried 0.pool.ntp.org, 1.pool.ntp.org, etc. and just pool.ntp.org servers), it does not sync. nslookup on pool.ntp.org as well as 0.pool.ntp.org gives proper DNS resolution and ping works on the IPs returned by nslookup as well. Output of iptables grepped for UDP is as follows: [Thu Feb 10 12:39:14 root@root-ubuntu:~]# iptables -L -n -v | grep udp 92 13001 ACCEPT udp -- * *...

Sow to sync system time to BIOS automatic after ntpd updated os time from remote ntp server?
Hi, all, I meet below problem on SUSE11 sp1 64bit with ntp4.2.4p6: 1. Set os date and bios date 7 minutes earlier than current date time; and wait ntpd to sync from remote ntp server to fix the time , after some minutes, both OS time and bios time fixed. 2. but when I try to test 30 minutes instead of 7minutes, only os time synced with upper level ntp server, the bios time still 30 minutes earlier than os time. hwclock can sync time to bios. Is there any option to control the "sync to bios" action? Or I need to set a cron job to run hwclock ? Thanks. -- Colin > > Is th...

Re: [ntp:questions] Unable to get time from NTP server
----- Original Message Follows ----- > That worked. I had a number of peers who's access policy had changed. > Also, it seems that changing notrust to nopeer seems to have corrected > the problem. Thanks for your help. > You shouldn't use notrust unless you know what you are doing. Its meaning changed betweeen 4.1.x and 4.2.0 and almost certainly was the cause of your problems. Danny ...

NTP Server Synchronizing with the incorrect NTP server
Problem: I configured my NTP server in Linux to use the US Pool of NTP Servers in /e= tc/ntp.conf. I restarted the NTP service many times and the actual server m= any times, but it was not using the pool. I was verifying this by using the= ntpq -p command. It was using a time server it picked up via DHCP when it = was plugged into a completely different network. I had no idea where it was= getting this information. Solution: After running strace (strace /etc/init.d/ntp restart) I was able to see it = was picking up the configuration from /var/lib/ntp/ntp.conf.dhcp . I delet...

A ntp client is running ahead of time of our netowork ntp server.
Hi List I have an issue with a ntp client running Linux SuSE 9.3 Our Network ntp server is a Red Hat Linux Advanced Server release 2.1A. *I initialy had an* error regarding seeing* *"Bad file descriptor" errors in /var/log/ntp* *then made sure that only one instance of ntpd* *is running socklist command and ps -auxxx|egrep ntpd thus run * *kill -9 PID's and restart ntpd /etc/init.d/ntpd restart. This seemed to fix things and when I run date command volia I thought I fixed it So thus to test after weekend. * *Thus I have checked this morning using date command and I the ...

Re: [ntp:questions] Re: Unable to get time from NTP server
----- Original Message Follows ----- > > 4.2.0: > > notrust > > Deny service unless the packet is cryptographically > > authenticated. nopeer > > Deny packets which would result in mobilizing a new association. > > This includes > > broadcast and symmetric active packets when a configured > > association does not exist. > > > It's not specified in the man page, which is probably why I missed it. > > -- > Daniel Rudy > The man pages are not authorative. Always look at the HTML documentation. Danny D...

LAN and a local NTP server (time protocol)
Is there a way to have a group of networked computers access one of the computers in the LAN to synchronize their time? I don't want these computers to have internet access, but I would like them all to be set to the same time (even if it isn't exactly the same as the atomic clock at the Naval Observatory). Can I set up one computer to be an NTP server and the others access it for setting each computers time? I'm currently using OSX jaguar. I used to have a client-server database that would set the time on each client as it accessed the database. However with OSX only the root...

On NTP server, how to check offset between client and NTP server?
Dear All I have stratum 2 NTP server. My client request time from my NTP server. On my NTP server, how to know offset of between my client and my server ? I found ntp_script software suite from http://www.schlitt.net/scripts/ntp/. It's not solve my problem directly. Thanks Chong On 2008-10-09, chowalit.lab Chowalit Lab Linux <chowalit.lab@gmail.com> wrote: > I have stratum 2 NTP server. My client request time from my > NTP server. On my NTP server, how to know offset of between my > client and my server ? I found ntp_script software suite from > http://www.schlit...

NTP server can not synchronize with external NTP server
Dear All, I got the problem, my NTP server can not synchronize with external NTP server. *root@ntpserver:~# ntpdate -d 2.id.pool.ntp.org 7 Jan 10:24:37 ntpdate[1779]: ntpdate 4.2.4p6@1.1549-o Thu Oct 22 21:58:39 UTC 2009 (1) transmit(110.92.72.250) transmit(202.162.32.12) transmit(110.92.72.250) transmit(202.162.32.12) transmit(110.92.72.250) transmit(202.162.32.12) transmit(110.92.72.250) transmit(202.162.32.12) transmit(110.92.72.250) transmit(202.162.32.12) 110.92.72.250: Server dropped: no data 202.162.32.12: Server dropped: no data server 110.92.72.250, port 123 stratum 0, precision 0,...

Adding NTP time servers (Mac OS X)
Is there a way I can add my own NTP time server under Mac OS X? -- It does not seem to work, adding it in the box System Preferences/Date & Time/Date & Time/Set Time & Date Automatically It only has some Apple time servers listed. -- Hans Aberg Hans Aberg <haberg@math.su.se> wrote: > Is there a way I can add my own NTP time server under Mac OS X? -- It does > not seem to work, adding it in the box If I check my "Set time autmatically" checkbox, the field becomes editable - assuming I have Admin privileges. Make sure the padlock at lower left is unlo...

Local time and NTP server time
Hello when I specify for example a ntp that is located in a different Time Zone, how ntp records my local time? thx Melanie Pfefer <melanie_pfefer@yahoo.co.uk> wrote: > Hello > > when I specify for example a ntp that is located in a different Time Zone, how ntp records my local time? > It doesn't as ntp always uses UTC. Conversion to local time is up to the system. -- Jim Pennino Remove .spam.sux to reply. Melanie Pfefer wrote: > Hello > > when I specify for example a ntp that is located in a different Time Zone, how ntp records my local time? > ...

Let ntp server not synchronize time from other servers
Hello, experts! Is there any way to configure my NTP server *not* to synchronize time from other servers? I have read the man pages about ntpd and googled a lot, but I didn't find any good solution. Any comments are welcome! Thanks! Hello WANG Cong, WANG Cong wrote: > Hello, experts! > > Is there any way to configure my NTP server *not* to synchronize time > from other servers? Just don't put any server lines into your ntp.conf file ;-) If you want some upstream servers to be polled but don't want to synchronize to them you can simply append the noselect keyword ...

[rfc-dist] RFC 5908 on Network Time Protocol (NTP) Server Option for DHCPv6
A new Request for Comments is now available in online RFC libraries. RFC 5908 Title: Network Time Protocol (NTP) Server Option for DHCPv6 Author: R. Gayraud, B. Lourdelet Status: Standards Track Stream: IETF Date: June 2010 Mailbox: richard.gayraud@free.fr, blourdel@cisco.com Pages: 9 Characters: 17452 Updates/Obsoletes/SeeAlso: None I-D Tag: draft-ietf-ntp-dhcpv6-ntp-opt-06.txt URL: ...

Web resources about - advice on hardware/OS/refclock for NTP server - comp.protocols.time.ntp


Graphics hardware - Wikipedia, the free encyclopedia
integrated and dedicated. An integrated graphics card, usually by Intel for use in their computers, is bound to the motherboard and shares RAM ...

Investors support Metcash bid for Home Timber & Hardware
Metcash revealed its interest in Woolworths' Home Timber &amp; Hardware business as long ago as the middle of 2015, sources close to the listed ...

LG admits G4 bootloop problem is a hardware fault, will repair affected devices
... LG. The official response has come through and it should make G4 owners happy: LG has acknowledged the LG G4 bootloop issue is a hardware defect ...

​What’s So Creative About Hardware?
... so much more to the convergence of technology and creative industries though. While Buildit accelerator was originally set up as a pure hardware ...

UC Berkeley profs lambast new “black box” network monitoring hardware
... that a Fidelis XPS system be installed at all 10 campuses at a total estimated cost of at least a few million dollars. The Fidelis hardware ...

Xerox Slips on WSJ Report It Will Split into Hardware, Services Cos.
... late trading after The Wall Street Journal sent headlines just before the close saying the company plans to split in two, forming one a hardware ...

Microsoft is treading water when it comes to hardware, but it could take off next year
... and 950XL, new Microsoft-made flagship Windows 10 Mobile phones that failed to get anyone excited about the platform , despite powerful hardware ...

Storied Xerox to split hardware and services into two companies
Over 50 years ago, Xerox introduced photocopiers based on a dry process called xerography that largely did away with the requirement of carbon ...

Apple’s rumored March event: catch our hardware expectations on The CultCast
This week: new hardware, incoming! We’ll tell you our hardware expectations of Apple’s rumored March 2016 event, and there’s a lot to discuss. ...

Commentary: Apple expected to see decreasing profits from hardware
Apple is expected to see difficulties maintaining strong hardware sales, as sales of iPhone, iPad and Mac-series products during its fiscal first ...

Resources last updated: 2/6/2016 3:39:09 AM