kolab3 locks up every 2 days

[Mat Cantin]

On 2013-03-04 16:23, Mat Cantin wrote:
> On 2013-03-01 6:25, JP Horst wrote:
>> Am Freitag, 1. März 2013, 15:12:22 schrieb Jeroen van Meeuwen:
>>> On 2013-03-01 12:00, JP Horst wrote:
>>>> Hello
>>>> 
>>>> as reported before on the German kolab-users list before:
>>>> Almost every 2 days postfix runs into authentication problems with
>>>> ldap (on
>>>> Centos6.3). After the first occurrence Kolab3 locks-up from user
>>>> perspective.
>>>> Only a manual restart via root solves the matter for the next two
>>>> days.
>>>> This way Kolab3 is too unstable for production environment.
>>>> 
>>>> The maillog shows the following:
>>>> 
>>>> (...snip...)
>>>> 
>>>> Helpful suggestions are appreciated.
>>> 
>>> It could be the Directory server is running out of file descriptors 
>>> -
>>> this should show in /var/log/dirsrv/slapd-*/errors
>>> 
>>> Please see [1] for upstream documentation on tweaking the system to
>>> optimize 389's performance.
>>> 
>>> https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Directory_Server/8.2/
>>> html/Performance_Tuning_Guide/system-tuning.html
>>> 
>>> Kind regards,
>>> 
>>> Jeroen van Meeuwen
>> 
>> Yes that seems to be the problem.
>> However, this is not a question of optimization but of having a
>> running server
>> at all. Our kolab3 system is used by 9 people. This number should not
>> be a
>> bottleneck.
>> 
>> Thank you for the link though. I will have a look at the
>> documentation.
>> 
>> Kind regards,
>> Jens-Peter
>> 
>> _______________________________________________
>> Kolab-users mailing list
>> Kolab-users at kolab.org
>> https://www.intevation.de/mailman/listinfo/kolab-users
> 
> I have this problem with one of my production servers, very annoying.
> What I've setup as a workaround is this script that runs every night. 
> To
> be honest, it still sometimes locks up during the day and requires me 
> to
> manually run the scripts, but it's better than restarting the server.
> 
> #!/bin/bash
> /etc/init.d/kolab-saslauthd stop
> /etc/init.d/kolabd stop
> /etc/init.d/dirsrv restart
> /etc/init.d/kolabd start
> /etc/init.d/kolab-saslauthd start
> 
> I've upped the file descriptors on this server, but it just delays the
> inevitable. Hope someone figures out a patch soon...
> 
> --
> Mat
> 
> _______________________________________________
> Kolab-users mailing list
> Kolab-users at kolab.org
> https://www.intevation.de/mailman/listinfo/kolab-users

I just wanted to follow up with this problem. It seems my server did 
this twice this morning in the span of 4 hours... the access logs might 
shed some light on the problem, but I'm not really sure what I'm looking 
at.

Right before the crash, the logs are filled with thousands of lines of 
this:
[08/Mar/2013:04:24:49 -0800] conn=94601 op=6 SRCH 
base="cn=kolab,cn=config" scope=2 
filter="(&(associatedDomain=mail.mydomain.ca))" attrs="associatedDomain"
[08/Mar/2013:04:24:49 -0800] conn=94601 op=6 RESULT err=0 tag=101 
nentries=0 etime=0

Once I run my script (shown above), the logs have a few hundred lines 
of this:
[08/Mar/2013:08:22:05 -0800] conn=94518 op=3 fd=985 closed - U1
[08/Mar/2013:08:22:05 -0800] conn=94501 op=3 UNBIND

Again, this server is only hosting 5 users who all use the roundcube 
interface, IMAPS, and activesync. Not really sure why the conn and fd 
counts are so high.

--
Mat