Re: Critical extension and Internal (implementation specific) errors

[Quanah Gibson-Mount <quanah@stanford.edu>]

--On Monday, November 14, 2005 5:35 PM -0200 Daniel Kobayashi Imori 
<daniel@bastion.com.br> wrote:

> There are many parameters configuration and factors in Berkeley DB and
> OpenLDAP about performance, I know. However, it worked very well when
> one database is running in one daemon. And when many databases are
> running in one daemon, is it cause many performance problems too?
>
> In my case, databases (BDBs) are smalls, with 50 entries or less, but
> the number of databases is up to 15!
>
> I believe that it's not a common configuration, but I need to keep this
> configuration because many reasons... And I believe that the problems
> appears because this uncommon configuration...
>
> Am I right? Someone can help me?
>
> Sorry for my bad english...

Daniel,

In that case, even the BDB defaults may be sufficient... That is an 
extremely tiny set of data for a database.  I'm personally not familiar 
with the behavior in that environment, you may indeed have hit a bug.  I 
would still think that "threads 100" is overly high, although with 15 DB's, 
it might work out to something like "8*15" which would be more than 100 
threads, so I'm not entirely sure.

It might be useful to see about grabbing some debugging information from 
slapd (start it with -d 7) and dump it to a log file until the error 
happens, and then send that output into an ITS at 
<http://www.openldap.org/its/> for evaluation from a developer.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/Shared Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html