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

Re: slapd crashes under Solaris 8




On Tue, 29 Aug 2000, Fazal Majid wrote:

>I have been stress-testing slapd under Solaris 8 Intel by running 2
>concurrent ldapadds with 100.000 entry LDIF files each. After a few
>minutes, the system will crash with a SIGSEGV in _reap in the thread
>library.

Is that 100,000 or 100 entries?  I have been doing the same for
FreeBSD, in FreeBSD you will get a bad memory leak that will consume
all of your resources until either you kill it or it just craps out.
With FreeBSD I have been doing 8000 chunks at a time, and going up
to around 50,000 ldapadds and deletes.

Threads seem to be the problem with adding more than 1,500 entries
that have about five or six attributes=values to each, aside from
the needed. 

FreeBSD lacking in the threads shindig, seems to stabalize with slapd
if configured --without-threads, and then compiled.  But without
threads, its very obvious that it can only do one thing at a time, and
the priority is with reading the database.  Which is ok, that makes sense.

Linux sofar, between FreeBSD and Linux(Mandrake 7.0), has the most
impressive speed between the two to me.  At home working on a project from
work, decided to setup LDAP on a linux box, looped a search, large 30,000
entry files, ldapadd < entry_file ; ldapdelete < entry_file... so on.

Try compiling --without-threads and see if that helps out.  But sadly
if threads are working nicely, there is a significant performance improvment.

Jaron Omega