I used pth-1.3.5, and built it with
--enable-optimize
--enable-syscall-soft
--disable-shared
Solaris 2.6 with the recommended patches.
-----Original Message-----
From: Wolfgang Hottgenroth [mailto:who_o@hottis.de]
Sent: Tuesday, July 18, 2000 1:48 PM
To: Vinod Nair; openldap-software@OpenLDAP.org
Subject: Re: slapd and Pth
RE: slapd and PthHi,
do you remember something more? Unfortunately it seems to be not enough ...
Wolfgang
----- Original Message -----
From: Vinod Nair
To: 'Wolfgang Hottgenroth' ; openldap-software@OpenLDAP.org
Sent: Tuesday, July 18, 2000 8:17 PM
Subject: RE: slapd and Pth
I seem to remember having to specify --with-yielding-select=no
to get around this.
Vinod.
-----Original Message-----
From: Wolfgang Hottgenroth [mailto:who_o@hottis.de]
Sent: Tuesday, July 18, 2000 10:00 AM
To: openldap-software@OpenLDAP.org
Subject: slapd and Pth
Yesterday I sent a question about a stopping output of slapd with Pth. I
observed the problem there under Solaris 2.6. Now I crosschecked it with
Linux (Kernel 2.2.15, Pentium) and found the same odd thing.
Did anyone of you has either good or bad experience with slapd and Pth?
Thanks,
Wolfgang
---x-------------------------------------------------------------------
Are there any problems known with Pth 1.3.6 and openldap 1.2.11? We run it
under Solaris 2.6, the backend is the BerkeleyDB, db-2.7.7. (I searched the
archive but didn't found anything helpful for me.)
I just compiled in this userland thread library to achieve better response
times for concurrent lookups, each with large result sets. But what I found
is, that the output of those large result sets stop at a random time and
will continue when I issue a lookup with another client.
I gave: (round about 400,000 entries in the server)
ldapsearch -b "dc=uu.net" -D "cn=root,dc=uu.net" -w secret -L
"objectclass=*"
I got a couple of entries, then it stops.
I gave from another xterm:
ldapsearch -b "dc=uu.net" -D "cn=root,dc=uu.net" -w secret -L -z 10
"objectclass=*"
I got the entries and the output for the first lookup will continue for a
while, then stops again. Reproducable.
I traced with "slapd -d 1" and found slapd stucking in "send_search_entry"
in file result.c. It waits at "ldap_pvt_thread_cond_wait( &conn->c_wcv,
&active_threads_mutex );"
I attached the file. The debug mark 14 is the last thing I saw when the
output stops.
Be aware: the output stops not for ever. It starts again when I issue a
lookup with another client.
Can anyone help?
Dipl. Ing. Wolfgang Hottgenroth
Product Engineering UUNET Europe, Middle East & Africa
Tel: +49 231 972 1231 Sebrathweg 20
Fax: +49 231 972 1180 44149 Dortmund
mailto: who@de.uu.net http://www.uunet.de
Dipl. Ing. Wolfgang Hottgenroth
Product Engineering UUNET Europe, Middle East & Africa
Tel: +49 231 972 1231 Sebrathweg 20
Fax: +49 231 972 1180 44149 Dortmund
mailto: who@de.uu.net http://www.uunet.de