Re: executing ntpdate on boot - seems it doesn't work
-----BEGIN PGP SIGNED MESSAGE-----
Rick Thomas wrote:
> On Feb 17, 2007, at 5:34 PM, Franck Joncourt wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>> Rolf Bode-Meyer wrote:
>>> I currently try to figure out if ntpdate is called on boottime in my
>>> system or not.
>>> It *should* be called when the network interfaces come up (ifup),
>>> therefore the /etc/network/if-up.d/ntpdate is present. And it's indeed
>>> called when I manually call ifup -a after booting--an entry in the
>>> syslog then shows something like "adjust time server ... offset ...".
>>> But I don't see such a syslog entry for boottime, so I fear there's
>>> something wrong. Any ideas what that could be or how to be sure
>>> everything is ok?
>> If it manually works, maybe you can add more lines to your
>> /etc/network/if-up/ntpdate file in order to track down where the
>> probleme comes from.
> Try turning on bootlogd (change "No" to "Yes" in
> /etc/default/bootlogd). That will copy everything that goes onto the
> console (from the "S05" point on in rsS.d) into /var/log/boot .
I did not know that, thanks. Maybe you can update your /etc/default/rcS
file with the following option : VERBOSE=yes, too. You should get more
information at boot time. I do not know whether it is going to help or
not, but you can give it a try.
>>> And another oddity: ifup is called by the network script which is
>>> rcS.d/S40networking. So if everything works well, ntpdate sets the
>>> system clock at S40. But *after* that S50hwclock.sh calls hwclock
>>> --hctosys which sets the system clock to the hardware clock.
>>> So doesn't hwclock needs to be called before ntpdate?
>> According to me you are right, hwclock should be start before ntpdate,
>> since ntpdate sets the system clock, and as you said, hwclock sets the
>> hardware clock from the system clock. It would be odd to do it in a
>> different way. I have checked my rcS.d directory, and I have :
>> S11hwclock and S40networking.
> That's (S11hwclock.sh) where hwclock gets called on my Etch test machine
> too. I have no S50hwclock.sh on that machine.
> But I *do* have S18hwclockfirst.sh *and* S50hwclock in /etc/rcS.d on my
> Sarge server. So, did you upgrade this machine from Sarge?
By the way, I am running Sid.
> Actually, if you don't use dynamic networking (as on a laptop with WiFi
> and modems and such -- you can't tell where your next internet
> connection is coming from) then the current recommendation from the NTP
> maintainers is to use ntp, not ntpdate. The latest ntp included in Etch
> has the ability to sync the system clock quickly on reboot, thus making
> ntpdate unnecessary. The upstream NTP development group (Dave Mills et
> al) would like to have ntpdate go the way of the dodo-bird. The last
> remaining place where it's got a serious application is on machines with
> intermittent network connections.
GPG server : pgpkeys.mit.edu
Fingerprint : C10E D1D0 EF70 0A2A CACF 9A3C C490 534E 75C0 89FE
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
Yahoo! Messenger - with free PC-PC calling and photo sharing. http://uk.messenger.yahoo.com