Hello Connie,

On Thu, 11 Aug 2005, Connie Sieh wrote:

> Stephan,
>
> On Thu, 11 Aug 2005, Stephan Wiesand wrote:
>
>> Hi,
>>
>> yes, I reported this for the first 3.0.5 release candidate, and the
>> solution is indeed to remove the LSB init info altogether. Trying
>> to invalidate it may make things worse - try replacing "BEGIN INIT INFO"
>> with "NO BEGIN INIT INFO", then chkconfig --reset on SL 4.1. Don't do this
>> on production systems, I have no clue how init reacts on the resulting
>> mess.
>>
>> And yes, afs is affected. The impact is however quite limited at least on
>> 3.0.5. On 4.1 it's a bit worse (the K01afs in all run levels may
>> terminate some processes much too early and not as cleanly as the should).
>> This had escaped my attention. We should probably do something about it.
>
> Suggestions?

Yes. We could rebuild openafs with init info removed from the rc script.
No need to update any package except openafs itself, nor to rebuild the 
modules, so it's not a big deal. It should then probably be released as an 
update for the other platforms as well.

An according srpm is here:

  /afs/ifh.de/user/w/wiesand/public/www/SL4/openafs-1.3.82-5.SL.src.rpm
  http://www-zeuthen.desy.de/~wiesand/SL4/openafs-1.3.82-5.SL.src.rpm

Or we could release an SL_ tweak rpm (in contrib?) with a trigger like 
this:

  %triggerin -- openafs

  sed -ie '/### BEGIN INIT INFO.*/,/### END INIT INFO.*/d' /etc/init.d/afs

  if [ -e /etc/rc3.d/S99afs -o -e /etc/rc3.d/S50afs ]; then
      /sbin/chkconfig afs reset || :
  fi

You can find a complete SL_afs_chkconfig_hotfix-1.0-1.src.rpm in the 
same location above. This would also work on 3.0.5 - and it undoes the
damage done by chkconfig before. Something similar could be done for
the pcp scripts if the chkconfig bug is a problem there as well.

> Still have time to fix 41 x86_64.

Cheers,
 	Stephan

> -Connie Sieh
>>
>> Cheers,
>>  	Stephan
>>
>> On Thu, 11 Aug 2005, Valery Mitsyn wrote:
>>
>>> Hi All,
>>>
>>> for services in /etc/rc.d/init.d with SuSE compatible part like:
>>> ### BEGIN INIT INFO
>>> .
>>> .
>>> .
>>> chkconfig <service> reset
>>> will set both start and stop priority to 50 on SL305,
>>> and start=99, stop=50 on SL41. If one broke compatibilities
>>> by (for example) inserting addition "-" in
>>> # Required-Start:
>>> # Required-Stop:
>>> things will going better.
>>> On SL41.
>>> --------------------------------------------------
>>> chkconfig pcp reset
>>>
>>> find /etc/rc.d -name \*pcp
>>> /etc/rc.d/rc4.d/S99pcp
>>> /etc/rc.d/rc6.d/K50pcp
>>> /etc/rc.d/rc5.d/S99pcp
>>> /etc/rc.d/rc0.d/K50pcp
>>> /etc/rc.d/init.d/pcp
>>> /etc/rc.d/rc3.d/S99pcp
>>> /etc/rc.d/rc1.d/K50pcp
>>> /etc/rc.d/rc2.d/S99pcp
>>>
>>> perl -pi -e "s|# Required-|# Required--|g" \
>>> /etc/rc.d/init.d/pcp
>>>
>>> chkconfig pcp reset
>>>
>>> find /etc/rc.d -name \*pcp
>>> /etc/rc.d/rc4.d/S95pcp
>>> /etc/rc.d/rc6.d/K05pcp
>>> /etc/rc.d/rc5.d/S95pcp
>>> /etc/rc.d/rc0.d/K05pcp
>>> /etc/rc.d/init.d/pcp
>>> /etc/rc.d/rc3.d/S95pcp
>>> /etc/rc.d/rc1.d/K05pcp
>>> /etc/rc.d/rc2.d/S95pcp
>>> --------------------------------------------------
>>> services affected on SL41:
>>> grep -l '# Required-' /etc/init.d/*
>>> /etc/init.d/afs
>>> /etc/init.d/pcp
>>> /etc/init.d/pmie
>>> and on SL305:
>>> /etc/init.d/afs
>>> /etc/init.d/kdc
>>> /etc/init.d/smartd
>>>
>>> Can anyone confirm these observations, or this weirdness
>>> comes with my environment only?
>>>
>>>
>>
>>
>

-- 

  ----------------------------------------------------
| Stephan Wiesand  |                                |
|                  |                                |
| DESY     - DV -  | phone  +49 33762 7 7370        |
| Platanenallee 6  | fax    +49 33762 7 7216        |
| 15738 Zeuthen    | mobile +49 171 317 6367        |
| Germany          | email  [log in to unmask] |
  ----------------------------------------------------