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

Re: How to disable services at startup... and keep them so.

First, apologies to Shawn. I didn't pay attention when responding.
Time to have dinner I think.

On Wed, Dec 8, 2010 at 3:56 PM, shawn wilson <ag4ve.us@gmail.com> wrote:
> On Wed, Dec 8, 2010 at 4:43 PM, Camaleón <noelamac@gmail.com> wrote:
>> On Wed, 08 Dec 2010 13:35:25 -0700, Bob Proulx wrote:
>> > shawn wilson wrote:
>> >> It's generally advised that you use update-rc.d to do this. However,
>> >> IIRC this is the exact same thing that the command does (nothing more)
>> >> so you should be fine.
>> >
>> > I see a lot of advice to use update-rc.d to manipulate the symlinks.
>> > That is fine.  But it isn't required.  It is requored for packages to
>> > use update-rc.d by policy.  But it isn't required for people.  You
>> > wouldn't want packages all to do their own thing in the postinst scripts
>> > because then it would then all be done inconsistently and many would be
>> > buggy.  Therefore packages are required to use the update-rc.d tool as a
>> > consistent interface to update symlinks.  That way they don't introduce
>> > random bugs and changes to the scheme can be implemented all in one
>> > place.  But that is packages and not people.
>> (...)
>> I am open to any/better alternative.
>> In fact, what this thread has shown us is that there is not a standard
>> method (let's call it "a common way") for doing a simple task like is
>> disabling a script from running and keep its current status.
>> I was looking for a "Debian way" for handling this, not just with Network
>> Manager but with all the scripts.
>> True is that "man update-rc.d" suggests using tools like "sysv-rc-conf"
>> but this tool is no even installed by default, so, how does one can give
>> credit to such tools if they are not part of the base system? >:-)
>> Other people in this thread has suggested the manipulation of "/etc/
>> init.d/*" scripts headers and then re-injecting them with "insserv",
>> which is of course another option... so, what is the recommended/
>> preferred way of doing this? "update-rc.d", "sysv-rc-conf", "insserv",
>> (other)...?
> the 'debian way' from what i've always understood is to use update-rc.d to
> do this. what Bob said was a pretty interesting way of doing things
> (changing the script and commenting so that you know why you did something).
> i've personally just gone and created and killed the symlinks and noted in
> the readme on each directory of what i did. i also keep a weekly backup of
> etc (a cron job on another server that does something like tar -cR /etc |
> gzip -c | scp - backup@host:/backups) and look at backups or changes in my
> readme's if i need to.
> what we are saying is that 'the debian way' and the manual way are
> essentially the same thing. so i suppose just use what you are comfortable
> with.

The move to insserv & lsb headers to deal with concurrent boot issues
has thrown a bit of dust into the eyes of update-rc.d. I think if the
user is working on a concurrent boot system, as squeeze is, they ought
to consider handling scripts using insserv, /etc/insserv.conf and the
lsb headers for each script in /etc/init.d (AFAIK insserv uses
update-rc.d to ensure symlinks in various runlevels match lsb

It will all be moot in a year or so anyway, once systemd gets tested
out on Fedora. Probably not in time for wheezy, but perhaps wheezy+1.

Reply to: