Bug#388328: [pkg-ntp-maintainers] Bug#388328: boot logging fix for ntpdate in /etc/init.d

[Rick Thomas]

On Sep 21, 2006, at 5:47 PM, Peter Eisentraut wrote:

> Again, check the package in unstable.  That's the direction.

Thanks for the pointer.

I've looked at the ntpdate_4.2.2+dfsg.2-2 stuff now, and I have a  
couple of comments on that.

	The first is just an extension of this original bug-report:

It does not make sense to use the "-s" option with ntpdate before  
syslogd has been started.  The if-up.d stuff will most commonly be  
run during system initialization.  The order of things in /etc/rcS.d  
is such that the network initialization will occur before starting  
syslogd (as it must, if syslog is redirected to another host).

But on the other hand, it does make sense to use the "-s" option if  
ntpdate is being called after syslogd is started -- as, for example,  
when bringing up a new network interface without a reboot.

So the logic in /etc/network/if-up.d/ntpdate needs to be a bit like  
the login in the old /etc/init.d/ntpdate in that it knows whether  
it's being called as part of system initialization or on-the-fly  
after initialization.


	The second has to do with interaction of ntpdate with the ntp daemon:

Maybe I've missed something, but i see no code that makes sure the  
ntp daemon is stopped when running ntpdate.  My experience is that a  
running ntpd can get badly confused if you run ntpdate at the same time.

Maybe the logic is that, now that ntpd can handle it's own clock  
startup, there's no need to have the two packages installed at the  
same time?  If so, would it be desirable to have them officially  
conflict?


Enjoy!

Rick