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

Re: slapd grows without bounds

Well, here's the latest:

Systems are now FreeBSD 3.4-stable, cvsup'd and
compiled Feb 22, using FreeBSD uthreads.  (db-2.7.7,
openldap-1.2.9 both compiled from scratch (not ports)
following directions/readmes found at openldap/sleepycat).
Previous compile was Feb 14.

I am unhappy to say the the problem has not relented
even slightly. I tried two different boxes, and the result
is the same. Again, configuring --without-threads
causes slapd to behave properly.

Forgive my ignorance, but how can I find the leakage?
(Could you point me to some tools and docs?). Never
had to do this before.


"Kurt D. Zeilenga" wrote:

> At 09:25 PM 2/21/00 -0500, Dave H wrote:
> >Well, you're right... I should have confirmed this. Compiling
> >with --without-threads definitely causes slapd to behave
> >the way it should. The question remains... what now?
> >Something is definitely broken. Is this an Openldap
> >problem or a FreeBSD problem (or both)?
> Are you using FreeBSD uthreads or FreeBSD LinuxThreads?
> If using LinuxThreads, use uthreads.
> In the past, OpenLDAP on FreeBSD 3.x releases have been rock
> solid.  Note, however, that FreeBSD 3-STABLE and 4-CURRENT
> have both undergone significant change in their uthread
> subsystem.  Apparantly one of these changes has induced a
> leak, likely a thread leak.  To sort out the exact cause
> (and the cure) will require further investigation.  Given
> that the leak appears to be significant in size, it should
> be fairly easy to find using common diagnostic tools.
> >Even if it's just a matter of tweaking something at compile
> >time, the solution should find it's way into the docs.
> Well, the first thing I suggest you do is make sure you
> have the latest FreeBSD-3.4-STABLE installed.  My quick
> tests against 3.4-STABLE uthreads and OpenLDAP running
> on 3.4-STABLE uthreads have all not shown leakage.
>         Kurt