On Thu, Aug 23, 2012 at 11:24:55AM +0200, Franky Van Liedekerke wrote:
On 2012-08-23 11:18, Jakub Hrozek wrote:
>On Thu, Aug 23, 2012 at 11:07:31AM +0200, Franky Van Liedekerke
>wrote:
>>Hi,
>>
>>I just had a weird situation: one of my servers suddenly no longer
>>allowed me to log in (pam auth via sssd).
>>Looking in the log for sssd, I had this message once every minute:
>>
>>[sssd[pam]] [pam_dp_reconnect_init] (0): Could not reconnect to LDAP
>>provider.
>>
>>Also, "getent passwd" no longer showed any ldap users.
>>Everything seemed correct, also the ldap servers, so I just
>>restarted the sssd daemon and all was well again.
>>Shouldn't the retry options of sssd just do that?
>>I have this as sssd config (obfuscated a bit):
>>
>
>Most likely this means the Data Provider had crashed. Can you
>check syslog
>for messages that would indicate a crash?
>
>The Data Provider should have respawned and the PAM Provider
>should have
>reconnected, though..I just ran a quick-n-dirty test locally,
>killed the
>sssb_be process and the sssd_pam reconnected fine for me..
>
>Unfortunately it seems that debugging is completely off according to
>your config file, I assume that there is nothing interesting in
>/var/log/sssd/*.log ?
In sssd_pam, I have these messages starting from Aug 19 (every minute):
(Sun Aug 19 05:09:28 2012) [sssd[pam]] [pam_dp_reconnect_init] (0):
Could not reconnect to LDAP provider.
but in syslog I only have this one line from Aug 20:
Aug 20 16:46:45 bqsma0001ap sssd[be[LDAP]]: LDAP connection error:
(null)
Ah, I was really expecting that the system would report that sssd_be
crashed. In general, running abrtd should catch all the crashes without
admin intervention.
For the rest there's no log (since of course it's a prod
system). Is
there a logging level I can safely set that will not fill my
partition?
debug_level = 6 or maybe 7 (1.5.x doesn't support the new hexa debug levels
yet) might be a good start.
I wouldn't worry too much about the logs filling the partition up, the
SSSD ships with a logrotate file and only level 9 gets really really
noisy.
Also, for completeness: version sssd-1.5.1-66.el6_2.3.x86_64 on
CentOS 6.2. Maybe some patch after that version fixed it, since 1.8
is already released.
Hmm, the bug I had in mind initially (#743841) was already fixed in the
version you are running. The 1.8 branch has quite a few fixes and
enhancements, I can recommend an upgrade from 1.5.x on a RHEL6 system.