Re: scaling openldap for performance

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

--On Friday, June 04, 2004 8:59 PM -0700 Trevor Warren 
<trevorwarren@yahoo.com> wrote:

> Morning Gibson,

My name is Quanah, not Gibson.

> --- Quanah Gibson-Mount <quanah@stanford.edu> wrote:
> > What version of OpenLDAP are you using?
> [snip]
>
>  The latest, 2.2.11.
>
> > What database backend are you using?
> [snip]
>
>  back-bdb. Do you suggest any other method for scaling
> to 30million users with significant updates/sec.
>
> > What version of the database backend are you using?
> [snip]
>
>  berkeley 4.2.52
>
> > Does it have all
> > applicable patches?
> [snip]
>
>  There just seem to be 2 patches available for the
> same wrt:
>
>  > Long running apps hanging
>  > Replication locks.
>
>  Our test scenario goes for a toss wrt 20 users  in a
> 15mins test. Now would you call that long??.

No I wouldn't call that long.  Are you using a multiprocessor master? 
There was a bug in 2.2.11 that would cause slurpd to lock up on 
multiprocessor machines.

> > Is your database backend configured via its
> > configuration mechanisms?
> [snip]
>
>  Neh...am gonna be looking at that too. You have any
> pointers on where to start reading up please?.

I hardly expect any system that has not properly configured the BDB backend 
to work very well.  It's about as useful as trying to run an OS without 
installing any RAM.

I suggest you read the slapd.conf manpage about:

checkpoint
cachesize

<http://www.stanford.edu/services/directory/openldap/configuration/slapd-conf-replica.html> 
has some more information on those, and:

<http://www.stanford.edu/services/directory/openldap/configuration/bdb-config.html> 
about configuring the DB_CONFIG file.

I also suggest reading:

<http://www.openldap.org/lists/openldap-software/200405/msg00413.html>

--Quanah

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