[systemd-devel] Debugging systemd shutdown problem on openSuSE 12.3

[Rainer Krienke]

Am 14.05.2013 12:00, schrieb Andrey Borzenkov:
> On Tue, May 14, 2013 at 11:43 AM, Rainer Krienke <krienke at uni-koblenz.de> wrote:
>>
>> Next I tried to reenable autofs, reboot, then  manually stopped the
>> service (systemctl stop autofs) and then reboot. Again the system hangs
>> but this time I still have a remote ssh open and it was not killed so I
>> could run journalctl to get the log:
>>
>> http://userpages.uni-koblenz.de/~krienke/tmp/systemd/journalctl
>>
>> What is striking to me is that the logs shows:
>>
>> May 14 09:28:27 rzinstal3 systemd[1]: Unit autofs.service entered failed
>> state
>>
>> This was exactly the time I manually stopped autofs, which worked.
> 
> No, it did not.
> 
> May 14 09:28:25 rzinstal3 systemd[1]: Stopping Automounts filesystems
> on demand...
> May 14 09:28:25 rzinstal3 systemd[1]: About to execute /sbin/killproc
> -t 45 -p $PIDFILE automount
> May 14 09:28:25 rzinstal3 systemd[1]: Forked /sbin/killproc as 3361
> May 14 09:28:25 rzinstal3 systemd[1]: autofs.service changed running -> stop
> May 14 09:28:26 rzinstal3 killproc[3361]: killproc: Usage:
> May 14 09:28:26 rzinstal3 systemd[1]: Received SIGCHLD from PID 3361 (killproc).
> May 14 09:28:26 rzinstal3 systemd[1]: Got SIGCHLD for process 3361 (killproc)
> May 14 09:28:26 rzinstal3 systemd[1]: Child 3361 died (code=exited,
> status=2/INVALIDARGUMENT)
> May 14 09:28:26 rzinstal3 systemd[1]: Child 3361 belongs to autofs.service
> May 14 09:28:26 rzinstal3 systemd[1]: autofs.service: control process
> exited, code=exited status=2
> 
> 
> It failed to properly stop and had to kill.
> 
>  One


Thanks for the hint. I tried to find out whats wrong with autofs.

The reason that autofs failed was not that it had problems mounting or
umounting but was rooted in /etc/nsswitch.conf, where long before
systemd was present in openSuSE I had entered ldap as an additional
source of maps and those old "sysv init" days this caused no problems
(eg in openSuSE 12.1 using still sysv init):

automount:      files nis ldap

Actually autmount gets its maps from NIS and our LDAP server is not yet
providing automounter maps and the ldap client isn't configured either
for this. For this reason automount reports an error since it cannot get
maps from ldap:

May 14 13:44:54 rzinstal3 automount[3496]: get_query_dn: lookup(ldap):
query failed for (&(objectclass=nisMap)(nisMapName=auto.direct)):
Invalid DN syntax

So I simply removed the ldap entry from nsswitch.conf because its
actually not needed at the moment. After this things now work and my
system reboots.

After all I still do not understand why this small error prevented
systemd from rebooting the whole system. autofs reported an error but on
the other hand it successfully umounted existing NFS mounts and
terminated the automount process. There was no hanging mount, just the
exit status != 0.

After all the question in my eyes is if this is a bug or a feature of
systemd?

Thanks
Rainer
-- 
Rainer Krienke, Uni Koblenz, Rechenzentrum, A22, Universitaetsstrasse  1
56070 Koblenz, http://userpages.uni-koblenz.de/~krienke, Tel: +49261287 1312
PGP: http://userpages.uni-koblenz.de/~krienke/mypgp.html,Fax: +49261287
1001312


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4697 bytes
Desc: S/MIME Kryptografische Unterschrift
URL: <http://lists.freedesktop.org/archives/systemd-devel/attachments/20130514/224c4638/attachment-0001.bin>