[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: openldap core dumping under load



On Thu, 27 Mar 2003, Matthew Schumacher wrote:

> I should note that openldap 2.1.16/sendmail-8.12.8 is performing very
> well in my environment on slackware linux.  I have a 1.2GHZ dell server
> with a raid5 array (I know, not the best setup for mail) but was still
> able to turn about 2100 messages a minute for 8 straight hours without
> fail.  I'm using the 2.4.20 kernel with the ext3 patches and also the
> new 2.0.3 megaraid driver.
>
> I don't use the full-on sendmail-ldap routing table, but I have a couple
> of rules that lookup data in ldap and my virtusertable is also mapped to
> ldap.
>
> Howard, any idea when the fixes in ITS# 2404 will make it to a release?
>   I really try to avoid CVS releases in production.
>

You can download the patch for the 2.1.16 release:
http://www.openldap.org/devel/cvsweb.cgi/~checkout~/libraries/libldap_r/tpool.c?rev=1.15.2.7&hideattic=1&sortbydate=0

-Igor

> Thanks,
>
> schu
>
> Igor Brezac wrote:
> > I am not sure if this is related, but I noticed that sendmail using ldapi
> > fails to connect to the ldap server at random times under moderate to
> > heavy loads.  I changed to ldap://localhost/ and problem when away.  This
> > is not a very busy sendmail installation.
> >
> > -Igor
> >
> > On Wed, 26 Mar 2003, Howard Chu wrote:
> >
> >
> >>I believe part of the response problem may be due to how sendmail interacts
> >>with LDAP, and part due to how slapd processes its input. Sendmail tends to
> >>spew a ton of queries out in rapid succession, and slapd tends to try to read
> >>all of its input before it processes any of it. There are two macros in
> >>slapd/connection.c that affect slapd's behavior: CONNECTION_INPUT_LOOP and
> >>DATA_READY_LOOP. If you comment out both of them, then slapd will read only
> >>one message at a time. Try that, it may even out the response time.
> >>
> >>Of course, pummeling a 350MHz PII with a 2.4GHz P4 may be part of your
> >>problem too...
> >>
> >>  -- Howard Chu
> >>  Chief Architect, Symas Corp.       Director, Highland Sun
> >>  http://www.symas.com               http://highlandsun.com/hyc
> >>  Symas: Premier OpenSource Development and Support
> >>
> >>
> >>>-----Original Message-----
> >>>From: Gordon Shumway [mailto:gshumway@cityextreme.com]
> >>
> >>>Thanks Howard.
> >>>
> >>>That certainly stopped the core dump.
> >>>
> >>>The only issue now seems to be CPU. During these times of
> >>>unusually high
> >>>load, slapd seems to use enough of the CPU to cause responses
> >>>to be slow.
> >>>Slow enough for IMAP ldap auths to timeout. Everything comes
> >>>good once the
> >>>mail flow slows.
> >>>
> >>>Maybe I just need a faster box.
> >>
> >>
> >
>
>

-- 
Igor