Hello,
We are testing openldap-2.3.38 with db-4.5.20. Our tests are oriented to measure capacity and stability of LDAP daemon.
3 clients machines perform search queries against the ldap server. We have used two types of clients: library NET::LDAP in perl, and C LDAP library directly. The problem is reproduced whit both types of clients
We are observed that the ldap server crash after 24-48 hours after working with high load (1000-2000 searches/second). In log file there is the following message:
connection_read(24): no connection! connection_read(18): no connection! connection_read(20): no connection! connection_read(19): no connection! connection_read(23): no connection! connection_read(18): no connection! connection_read(16): no connection! connection_read(22): no connection! connection_read(19): no connection!
Using TCPDump we have observed that clients do not finish the search witch an unbind operation. Simply TCP connection is closed by a FIN packet. We suppose that this is the cause than produces the message in the LDAP log.
The main issue is than after 24-48 hours the LDAP daemon dies. We have also tested openldap-2.3.30.
Our configures are:
Berkeley DB config.status 4.5.20 configured by ../dist/configure, generated by GNU Autoconf 2.60, with options "'--prefix=/opt/lib/db-4.5.20' '--enable-compat185' '--enable-cxx'"
configured by ./configure, generated by GNU Autoconf 2.59, with options "'--prefix=/opt/openldap-2.3.38' '--exec-prefix=/opt/openldap-2.3.38' '--disable-syslog' '--disable-ipv6' '--enable-slapd' '--enable-crypt' '--enable-bdb' '--enable-syncprov' '--with-threads' '--with-tls' 'CPPFLAGS=-I/opt/lib/db/include -I/opt/lib/openssl/include' 'LDFLAGS=-L/opt/lib/db/lib -L/opt/lib/openssl/lib
The machine is a linux Redhat Enterprise 4.5.
--On Monday, November 05, 2007 2:10 PM +0100 Antonio Guirado Puerta antonio.guirado@f-integra.org wrote:
Hello,
We are testing openldap-2.3.38 with db-4.5.20. Our tests are oriented to measure capacity and stability of LDAP daemon.
3 clients machines perform search queries against the ldap server. We have used two types of clients: library NET::LDAP in perl, and C LDAP library directly. The problem is reproduced whit both types of clients
We are observed that the ldap server crash after 24-48 hours after working with high load (1000-2000 searches/second). In log file there is the following message:
connection_read(24): no connection! connection_read(18): no connection! connection_read(20): no connection! connection_read(19): no connection! connection_read(23): no connection! connection_read(18): no connection! connection_read(16): no connection! connection_read(22): no connection! connection_read(19): no connection!
That's not a crash, that says there's no connection to be read. What would actually be useful, is the last data logged by slapd before the connection_read errors. Given that I've had slapd run for months under a high read rate without any such issue, I'm guessing there something specific to your setup going on here. Unfortunately, what you've submitted here doesn't really give any data useful for determining what that is.
--Quanah
--
Quanah Gibson-Mount Principal Software Engineer Zimbra, Inc -------------------- Zimbra :: the leader in open source messaging and collaboration