[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Why /etc/init.d/.depend.start did not caught editing to /etc/insserv.conf?



Regid Ichira wrote:
> Bob Proulx writes:
> > Regid Ichira wrote:
> > >   I have installed the ntp package.  I edited /etc/insserv.conf, and
> > > added +ntp to the $time line.
> > > 
> > >     $ grep time /etc/insserv.conf
> > >     # The system time has been set correctly
> > >     $time           +hwclock +ntp
> > 
> > Why do you think this manual action is needed?  Why isn't simply
> > installing the ntp package sufficient?
> 
>   I tried to avoid adding $all for my case. $all was suggested by myself,
> in bts#634215: Having init.d/cron run early breaks the semantics of @reboot
> time specification.

Looking at http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=634215 I
see that you have found a problem related to @reboot with cron
starting earlier using the dependency based booting than it did before
with the legacy based manually specified number booting.  But there
are some missing details.

I assume that you have an action in @reboot from cron that is now
being run before ntp is started and that your hardware clock isn't
functioning?  Because if your hardware clock is functioning then time
should be pretty close to correct even without ntp running.  But I
know if your hardware clock is dead, such as from a dead battery, then
time will be very incorrect until ntp is finished running.  I assume
that your @reboot action is having a problem because it is running in
this intermediate time?  Could you clarify?

> > Won't adding that line create a circular dependency loop between
> > rsyslog and ntp?  rsyslog needs $time.  ntp needs $syslog.
> 
>   I tried that.  It was refused by insserv because of the circular loop.

Right.  And that shows that there is a problem.  These types of
problems have always existed.  I have run into them usually with DNS
nameservers and networking.  Everyone wants DNS nameservice, everyone
wants logging, everyone wants networking and no one is willing to
stage for the bootstrap process.  For the most part Debian has taken a
"head in the sand" approach of ignoring the problem.  But with
dependency based booting this issue is forced.  I think dealing with
the issue is better.  But that does mean that things are going to be
broken until it is really resolved in some reasonable way.  Welcome to
the front line of the battle.

For what it is worth I think your customization of cron to add a
dependency is the right way to go.  Did you try adding ntp to the
"Should-Start" list?  I think that would be better.

Bob

Attachment: signature.asc
Description: Digital signature


Reply to: