[Pkg-openldap-devel] Bug#688797: [NS] Bug#688797: Dead lock on BDB and partial stop on slapd

[Jose Manuel dos Santos Calhariz]

On Wed, Oct 17, 2012 at 03:50:32PM +0100, Jose Manuel dos Santos Calhariz wrote:
> On Wed, Oct 17, 2012 at 06:46:11AM -0700, Quanah Gibson-Mount wrote:
> > 
> > 
> > --On October 17, 2012 2:35:30 PM +0100 Jose Manuel dos Santos
> > Calhariz <jose.calhariz at netvisao.pt> wrote:
> > 
> > >On Wed, Oct 10, 2012 at 03:18:50PM +0100, Jose Manuel dos Santos Calhariz
> > >wrote:
> > >>> (...)
> > >>
> > >>Is openldap 2.4.31 current enough?
> > >
> > >It even happen with 2.4.33 :-(  It was compiled with BDB (4.7.25 + all
> > >patches).  Waiting for the next stop to collect debug information.
> > 
> > Ok.  did you see the information Howard requested you gather via his
> > comment in the bug?
> 
> For some reason I did't receive that comment.  Let me give the
> information that I have now.

Another partial stop with 2.4.33.  Again what stopped was the
synchronization with the LDAP master.  A watchdog detected the
situation and restarted the slapd daemon.


> 
> >> Can you give some more background on what has happened just prior to this 
> >> hang? 

The slaves are running the normal day workload until the thread that
process the replication stops processing the updates.  We have a
watchdog that detect that the slave is no longer up to date with the
master and restart the slave.

I have the exact time the daemon was restarted, 15:10:42, and the
full log with -d "Stats Stats2 Sync Trace"

> >> Do you have slapd logs (running with TRACE debug level, -d1 or -s1) 
> >> leading up to the hang? In particular, I'm interested in knowing whether you 
> >> see a message like:
> >>   <= index_entry_%s( %ld, \"%s\" ) failure
> 
> I will change the debug level to include trace.

Looking into the logs I only see:
    <= index_entry_%s( %ld, \"%s\" ) success

> 
> >> just before. Also, are you using cn=config on these servers, and did you just 
> >> reconfigure the index of the relevant attribute prior to the hang? (E.g., in 
> >> your trace, the istPersonServices attribute.)
> 
> Yes, we use cn=config, but didn't do any changes to any index.
> 
> > 
> > Thanks,
> > Quanah
> > 
> > 
> 
>    Thanks,
> Jose Calhariz
> 
> _______________________________________________
> ns-list mailing list
> ns-list at dsi.ist.utl.pt
> https://mlists.ist.utl.pt/mailman/listinfo/groups.ciist.ns-list

Thanks,
Jose Calhariz


-- 
--
Nihil est in intellectu quod non ante fuerit in sensu.
(Nada está no intelecto que não tenha passado antes pelos sentidos)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-openldap-devel/attachments/20121019/96391dbb/attachment.pgp>