Hi,
I'm running debian/unstable on my laptop and often the LAN/WLAN is not
connected (yet)
when the system is starting up - including NTPD.

NTPD then seems to discard all unreachable server entries and ends up
with no peers left.
In some googled doc I found the ntp.conf option "dynamic" to tell it
that some peers may become available later on,
but the debian ntpd doesnt't seem to understand it - or I don't know how
to use it correctly.

What is the best way to configure the ntpd in such an environment ?

TIA,
Bruno


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Tue, Jan 23, 2007 at 01:49:55AM +0100, Bruno Voigt wrote:
> I'm running debian/unstable on my laptop and often the LAN/WLAN is not
> connected (yet)
> when the system is starting up - including NTPD.
>
> NTPD then seems to discard all unreachable server entries and ends up
> with no peers left.
> In some googled doc I found the ntp.conf option "dynamic" to tell it
> that some peers may become available later on,
> but the debian ntpd doesnt't seem to understand it - or I don't know how
> to use it correctly.
>
> What is the best way to configure the ntpd in such an environment ?

How about calling the init.d script from /etc/network/interfaces?

I mean, there is little need for having the deamon running while
offline and providing that your clock does not drift too much, that
should do it.

jacques

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFFtb5CoYczEVPGNVERAgVhAJ9QAcwVDF+J0UbVly0lY3 +PcFnBNQCeMWOy
w9+CcD26U4ONOj6qCjn0bl8=
=NC/1
-----END PGP SIGNATURE-----

Jacques Normand wrote:
>> I'm running debian/unstable on my laptop and often the LAN/WLAN is not
>> connected (yet)
>> when the system is starting up - including NTPD.
>>
>> NTPD then seems to discard all unreachable server entries and ends up
>> with no peers left.
>> In some googled doc I found the ntp.conf option "dynamic" to tell it
>> that some peers may become available later on,
>> but the debian ntpd doesnt't seem to understand it - or I don't know how
>> to use it correctly.
>>
>> What is the best way to configure the ntpd in such an environment ?
>>
>
> How about calling the init.d script from /etc/network/interfaces?
>
> I mean, there is little need for having the deamon running while
> offline and providing that your clock does not drift too much, that
> should do it.
>
That is hard to configure reliable because of stuff like openvpn that
also needs a varying timespan to settle etc..
So I would really like to know a way to tell ntpd to not give up on a
server entry if it is not reachable/resolvable on the first try:

Jan 23 21:06:27 europa ntpd_initres[5255]: host name not found:
0.debian.pool.ntp.org │
Jan 23 21:06:27 europa ntpd_initres[5255]: couldn't resolve
`0.debian.pool.ntp.org', giving up on it │


Bruno


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Tue, Jan 23, 2007 at 08:53:55PM +0100, Bruno Voigt wrote:
> >How about calling the init.d script from /etc/network/interfaces?
> >
> >I mean, there is little need for having the deamon running while
> >offline and providing that your clock does not drift too much, that
> >should do it.
> >
> That is hard to configure reliable because of stuff like openvpn that
> also needs a varying timespan to settle etc..
> So I would really like to know a way to tell ntpd to not give up on a
> server entry if it is not reachable/resolvable on the first try:
>
> Jan 23 21:06:27 europa ntpd_initres[5255]: host name not found:
> 0.debian.pool.ntp.org ???
> Jan 23 21:06:27 europa ntpd_initres[5255]: couldn't resolve
> `0.debian.pool.ntp.org', giving up on it ???

If you really wish to have ntpd running all the time, you can still
add/remove peers dynamically.
ntpdc -c addserver peer_address [keyid] [version] [prefer]
ntpdc -c unconfig peer_address

I don't there is any way to get away from the synchronization time
though. So running ntpd without peers versus not running it at all
should not make any difference except if your clock is very biased and
you stay offline for a long time. I also do not know what happen at
startup to correct for the skew of the rtc during the shutdown. If it is
taken care of by ntpd at start, then you have one less reason to leave
it on...

jacques


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFFtp/IoYczEVPGNVERAglHAJ4/XmtBvGHWY5ZNCBp4MAX3HzhZIACguU3r
bIbO9AZjokL/KZUVkJ2AhD4=
=/AY+
-----END PGP SIGNATURE-----

On Jan 23, 2007, at 2:53 PM, Bruno Voigt wrote:

> Jacques Normand wrote:
>>> I'm running debian/unstable on my laptop and often the LAN/WLAN
>>> is not connected (yet)
>>> when the system is starting up - including NTPD.
>>>
>>> NTPD then seems to discard all unreachable server entries and
>>> ends up with no peers left.
>>> In some googled doc I found the ntp.conf option "dynamic" to tell
>>> it that some peers may become available later on,
>>> but the debian ntpd doesnt't seem to understand it - or I don't
>>> know how to use it correctly.
>>>
>>> What is the best way to configure the ntpd in such an environment ?
>>>
>>
>> How about calling the init.d script from /etc/network/interfaces?
>>
>> I mean, there is little need for having the deamon running while
>> offline and providing that your clock does not drift too much, that
>> should do it.
>>
> That is hard to configure reliable because of stuff like openvpn
> that also needs a varying timespan to settle etc..
> So I would really like to know a way to tell ntpd to not give up on
> a server entry if it is not reachable/resolvable on the first try:
>
> Jan 23 21:06:27 europa ntpd_initres[5255]: host name not found:
> 0.debian.pool.ntp.org │
> Jan 23 21:06:27 europa ntpd_initres[5255]: couldn't resolve
> `0.debian.pool.ntp.org', giving up on it │

Try replacing the symbolic name in one of the "server" lines in /etc/
ntp.conf with an absolute IP address. Then comment out (or simply
remove) all the other servers to minimize clutter. It might also be
a good idea to add the "minpoll 10" option to the server you keep.

E.g.

The default /etc/ntp.conf file has (just the server lines):

server 0.debian.pool.ntp.org iburst
server 1.debian.pool.ntp.org iburst
server 2.debian.pool.ntp.org iburst
server 3.debian.pool.ntp.org iburst

Replace them with the single line:

server xx.yy.zz.ww minpoll 10 iburst

Where xx.yy.zz.ww is chosen from one of the numeric IP addresses
returned by doing

host pool.ntp.org

This will avoid the "can't resolve, so giving up" phenomenon because
the IP address is already known and doesn't need to be resolved.
It will also tell ntpd to poll no more than often than once every
1024 seconds (about 17 minutes), which will reduce the frequency of
useless ntp packets that can't be delivered. Without the "minpoll
10" it will default to "minpoll 6" which will poll every 64 seconds
(about 1 minute). If you don't usually stay connected to the
internet for longer than a half-hour or so at a time, try "minpoll 8"
which will poll every 256 seconds (4-5 minutes). The object is to
get a few polls out to the server and back during each period of
internet connectivity.

The downside is that you have to pick one of the pool.ntp.org servers
and stick with it. If it goes down or leaves the net, you have to be
aware of that possibility and be ready to choose a different server
when it happens. Using a symbolic DNS address (default for those
with reliable internet connections, but not a viable option for you)
will get a random choice from a known-good list each time ntpd is
started up.

I've deliberately avoided giving a particular IP address (used
xx.yy.zz.ww, instead) to avoid having a bunch of novices bombarding
some unfortunate single server in the case that this advice gets
posted to a wiki and used by lots of folks. So, "some assembly
required" is the price of life in the real world.


Hope this s!

Rick

On Tue, 23 Jan 2007, Jacques Normand wrote:
> you stay offline for a long time. I also do not know what happen at
> startup to correct for the skew of the rtc during the shutdown. If it is
> taken care of by ntpd at start, then you have one less reason to leave
> it on...

ntp can do two things. It can "step" time, which is dangerous if it makes
the clock go backwards in an already running system (i.e. not during early
startup), or it can slew time (make the clock slow down a bit or go a bit
faster) until the time is correct.

If the difference is too big, slewing is not possible, as slewing cannot
make the clock run that much faster or that much slower than real time
for obvious reasons.

If ntp is slewing time, you better don't kill it until it is done.

Anyway, if the machine is not going to be connected to a network with
timeservers most of the time, install the "chrony" package instead of ntp.
"chrony" is designed for this kind of situation, where the machine stays a
lot of time offline. It will keep time correct better than NTP in such a
situation, and it will be easier on the time servers too.

ntp really *is* for machines that are constantly in contact with either
other ntp servers, or directly connected to a reference clock (GPS, atomic
clock, etc). For all other uses, you have "chrony".

As for ntpdate, it is something you should just use when the system is in
single user mode or during early startup.

--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On 23.01.07 01:49, Bruno Voigt wrote:
> I'm running debian/unstable on my laptop and often the LAN/WLAN is not
> connected (yet) when the system is starting up - including NTPD.
>
> NTPD then seems to discard all unreachable server entries and ends up with
> no peers left. In some googled doc I found the ntp.conf option "dynamic"
> to tell it that some peers may become available later on, but the debian
> ntpd doesnt't seem to understand it - or I don't know how to use it
> correctly.
>
> What is the best way to configure the ntpd in such an environment ?

I use chrony on my laptop. It supports online and offline mode, however in
sarge it only has up/down scripts in ppp subdirectories (not very good since
I usually plug in ethernet). However I'm quite satisfied with it and when
needed I run those scripts manually. I was thinking about creating ip-up.d
and ip-down.d scripts (very elegant with ifplugd) which will set proper
mode, and probably the provided NTP server (from dhcp or ppp config) too.

Maybe something alike is in etch already...
--
Matus UHLAR - fantomas, uhlar@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
One OS to rule them all, One OS to find them,
One OS to bring them all and into darkness bind them


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Fri, Jan 26, 2007 at 07:31:08PM +0100, Matus UHLAR - fantomas wrote:
> On 23.01.07 01:49, Bruno Voigt wrote:
> > I'm running debian/unstable on my laptop and often the LAN/WLAN is not
> > connected (yet) when the system is starting up - including NTPD.
> >
> > NTPD then seems to discard all unreachable server entries and ends up with
> > no peers left. In some googled doc I found the ntp.conf option "dynamic"
> > to tell it that some peers may become available later on, but the debian
> > ntpd doesnt't seem to understand it - or I don't know how to use it
> > correctly.
> >
> > What is the best way to configure the ntpd in such an environment ?
>
> I use chrony on my laptop. It supports online and offline mode, however in
> sarge it only has up/down scripts in ppp subdirectories (not very good since
> I usually plug in ethernet). However I'm quite satisfied with it and when
> needed I run those scripts manually. I was thinking about creating ip-up.d
> and ip-down.d scripts (very elegant with ifplugd) which will set proper
> mode, and probably the provided NTP server (from dhcp or ppp config) too.
>


I used to run chrony (with difficulty) but ran into a problem with my
new box: it couldn't access the hwclock. If I told it not to, (so that
the hwclock shutdown script could work), it really messed up my time.
So I switched to ntp. I access the net with ppp and put a script into
ip-up.d to restart ntp when the link comes up. Yes, ntp will jump
instead of skew since a skew can take a __very__ long time to
accomplish.

With fewer and fewer people using dial-up I feel a bit orphaned at times
and ntp/chrony is one of them.

Doug.


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Douglas Allan Tutty <dtutty@porchlight.ca> writes:

> I used to run chrony (with difficulty) but ran into a problem with my
> new box: it couldn't access the hwclock. If I told it not to, (so that
> the hwclock shutdown script could work), it really messed up my time.
> So I switched to ntp. I access the net with ppp and put a script into
> ip-up.d to restart ntp when the link comes up. Yes, ntp will jump
> instead of skew since a skew can take a __very__ long time to
> accomplish.
>
> With fewer and fewer people using dial-up I feel a bit orphaned at times
> and ntp/chrony is one of them.

You might want to look into something like what I do. I use ntpdate
to set the time daily over ppp, but then I manually track the offset
and use adjtimex to adjust the kernel clock. I have it set well
enough that the average slew is about 30 msec/day. I think this is
basically what chrony does automatically, but I was using this before
chrony was available.

--
Carl Johnson carlj@peak.org


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Jan 27, 2007, at 2:19 PM, Douglas Allan Tutty wrote:

> I used to run chrony (with difficulty) but ran into a problem with my
> new box: it couldn't access the hwclock. If I told it not to, (so
> that
> the hwclock shutdown script could work), it really messed up my time.
> So I switched to ntp. I access the net with ppp and put a script into
> ip-up.d to restart ntp when the link comes up. Yes, ntp will jump
> instead of skew since a skew can take a __very__ long time to
> accomplish.


If you're running "etch", you might want to try the etch "ntpdate"
package. It's configured to run ntpdate and jump the clock to the
correct time every time a network interface is brought up. You
should probably not run ntp at the same time -- in this mode, the two
will fight about who controls the system clock.

Rick


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Sun, Jan 28, 2007 at 12:56:02AM -0500, Rick Thomas wrote:
>
> On Jan 27, 2007, at 2:19 PM, Douglas Allan Tutty wrote:
>
> >I used to run chrony (with difficulty) but ran into a problem with my
> >new box: it couldn't access the hwclock. If I told it not to, (so
> >that
> >the hwclock shutdown script could work), it really messed up my time.
> >So I switched to ntp. I access the net with ppp and put a script into
> >ip-up.d to restart ntp when the link comes up. Yes, ntp will jump
> >instead of skew since a skew can take a __very__ long time to
> >accomplish.
>
>
> If you're running "etch", you might want to try the etch "ntpdate"
> package. It's configured to run ntpdate and jump the clock to the
> correct time every time a network interface is brought up. You
> should probably not run ntp at the same time -- in this mode, the two
> will fight about who controls the system clock.
>


The ntp docs say that ntpdate will (as they put it, after an appropriate
period of mourning) be obsolete now that ntp has the -q option.

Doug.


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Monday 22 January 2007 16:49, Bruno Voigt wrote:
> Hi,
> I'm running debian/unstable on my laptop and often the LAN/WLAN is not
> connected (yet)
> when the system is starting up - including NTPD.
>
> NTPD then seems to discard all unreachable server entries and ends up
> with no peers left.
> In some googled doc I found the ntp.conf option "dynamic" to tell it
> that some peers may become available later on,
> but the debian ntpd doesnt't seem to understand it - or I don't know how
> to use it correctly.
>
> What is the best way to configure the ntpd in such an environment ?

Just configure ntpd as normal, but add an hourly cronjob which does
"/etc/init.d/ntp restart >/dev/null".

A client has a server on PPPOE where ntpd always fails at the outset.
Using this, rather than ntpdate, results in the clock staying in sync
without stepping. The hwclock is good enough for ntpd after reboot
(or DSL outage) until the next time the cronjob runs.

--Mike Bird


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Jan 28, 2007, at 3:34 PM, Mike Bird wrote:

> On Monday 22 January 2007 16:49, Bruno Voigt wrote:
>> Hi,
>> I'm running debian/unstable on my laptop and often the LAN/WLAN is
>> not
>> connected (yet)
>> when the system is starting up - including NTPD.
>>
>> NTPD then seems to discard all unreachable server entries and ends up
>> with no peers left.
>> In some googled doc I found the ntp.conf option "dynamic" to tell it
>> that some peers may become available later on,
>> but the debian ntpd doesnt't seem to understand it - or I don't
>> know how
>> to use it correctly.
>>
>> What is the best way to configure the ntpd in such an environment ?
>
> Just configure ntpd as normal, but add an hourly cronjob which does
> "/etc/init.d/ntp restart >/dev/null".
>
> A client has a server on PPPOE where ntpd always fails at the outset.
> Using this, rather than ntpdate, results in the clock staying in sync
> without stepping. The hwclock is good enough for ntpd after reboot
> (or DSL outage) until the next time the cronjob runs.
>
> --Mike Bird

I like that idea. It avoids ntpdate, which (for better or for worse,
is deprecated by the NTP developers and will not be getting timely
updates in the future) and it gives ntpd a chance to do it's starting
transient thing (fairly recently added) and thus get the parameters
right if it can talk to the Internet servers. And it does no
(serious) harm if you're disconnected from the Internet when the cron
runs.

Worth a try!

Rick


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Jan 28, 2007, at 3:34 PM, Mike Bird wrote:

> On Monday 22 January 2007 16:49, Bruno Voigt wrote:
>> Hi,
>> I'm running debian/unstable on my laptop and often the LAN/WLAN is
>> not
>> connected (yet)
>> when the system is starting up - including NTPD.
>>
>> NTPD then seems to discard all unreachable server entries and ends up
>> with no peers left.
>> In some googled doc I found the ntp.conf option "dynamic" to tell it
>> that some peers may become available later on,
>> but the debian ntpd doesnt't seem to understand it - or I don't
>> know how
>> to use it correctly.
>>
>> What is the best way to configure the ntpd in such an environment ?
>
> Just configure ntpd as normal, but add an hourly cronjob which does
> "/etc/init.d/ntp restart >/dev/null".
>
> A client has a server on PPPOE where ntpd always fails at the outset.
> Using this, rather than ntpdate, results in the clock staying in sync
> without stepping. The hwclock is good enough for ntpd after reboot
> (or DSL outage) until the next time the cronjob runs.
>
> --Mike Bird

I like that idea. It avoids ntpdate, which (for better or for worse,
is deprecated by the NTP developers and will not be getting timely
updates in the future) and it gives ntpd a chance to do it's starting
transient thing (fairly recently added) and thus get the parameters
right if it can talk to the Internet servers. And it does no
(serious) harm if you're disconnected from the Internet when the cron
runs.

Worth a try!

Rick


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Sun, Jan 28, 2007 at 12:34:10PM -0800, Mike Bird wrote:
> On Monday 22 January 2007 16:49, Bruno Voigt wrote:
> > Hi,
> > I'm running debian/unstable on my laptop and often the LAN/WLAN is not
> > connected (yet)
> > when the system is starting up - including NTPD.
> >
> > NTPD then seems to discard all unreachable server entries and ends up
> > with no peers left.
> > In some googled doc I found the ntp.conf option "dynamic" to tell it
> > that some peers may become available later on,
> > but the debian ntpd doesnt't seem to understand it - or I don't know how
> > to use it correctly.
> >
> > What is the best way to configure the ntpd in such an environment ?
>
> Just configure ntpd as normal, but add an hourly cronjob which does
> "/etc/init.d/ntp restart >/dev/null".
>
> A client has a server on PPPOE where ntpd always fails at the outset.
> Using this, rather than ntpdate, results in the clock staying in sync
> without stepping. The hwclock is good enough for ntpd after reboot
> (or DSL outage) until the next time the cronjob runs.
>


Or, if you have some scriptable way to tell if the net is up (a ping to
the ntp server?), you could put a script as the last to run in the
network ip-up.d. Have it sit there polling (intermittant ping?) the
server every minute until it gets a response then /etc/init.d/ntp
restart, then exit.

This would avoid the net-up just after the cron job goes then having to
wait an hour before ntp is connected.

Doug.


--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org