Re: Repeated Search+Modify fails after 1024 cycles, slapd seg faults. (ITS#1447)

[simon@ateb.co.uk]

Hello again...

I've also noticed that once the server has stopped returning results after the
1024th search and modify
it is something to do with objectclasses. The query
(&(objectclass=posixaccount)(uid=test1024)) returns no result, but the query
(uid=test1024) does return the dn for the record correctly (including the
objectclass: posixaccount attribute!). The first query will not work until
(&(objectclass=posixaccount)(uid=test2045)) but the second works for every uid.

If we remove the index for objectclass altogether then the server works fine all
the time but queries are too slow :-(

Our target time per search on a db with two main objectclasses, with 5000
records in each is .025 seconds with objectclass indexing on. With it off this
search time is 4 seconds per record.

There must be a problem somewhere in the indexing of objectclasses.

Hope this sheds more light on our problem.

Regards
Simon

P.s Several people have mailed me to say they have the same problem on RH7.1 and
SuSE 7.2 with openldap 2.0.18.


----- Original Message -----
From: "Pierangelo Masarati" <openldap-its@OpenLDAP.org>
To: <simon@ateb.co.uk>
Sent: 16 November 2001 16:16
Subject: Re: Repeated Search+Modify fails after 1024 cycles, slapd seg faults.
(ITS#1447)


> sounds like a thread problem, though you state it occurs also when
> compiling --without-threads. Can you confirm that?
>
> Pierangelo.
>
>