RE: slapd stops processing thread jobs (ITS#2774)

[hyc@highlandsun.com]

> -----Original Message-----
> From: owner-openldap-bugs@OpenLDAP.org
> [mailto:owner-openldap-bugs@OpenLDAP.org]On Behalf Of
john.matthews@crossml.com

> Full_Name: John Matthews
> Version: 2.1.22
> OS: Linux 2.4.20
> URL: ftp://ftp.openldap.org/incoming/
> Submission from: (NULL) (203.12.97.112)
>
>
> I have been investigating a slapd problem where after a few
> days it will stop
> responding to new client requests. Rather oddly the problem
> always occurs at
> about 1am on one of our LDAP servers. Using a debug level
> that includes
> LDAP_DEBUG_CONNS and with some extra debug output in
> connection.c (see below) I
> have tracked the problem down into the threading code tpool.c.

> Notice how each new operation is deferred in
> connection_input() because
> conn->c_conn_state is "stuck" in a SLAP_C_BINDING state. The number of
> "active_threads" in the debug ouput increases for each new request,
> connection_op_activate() never gets called again and of
> course the LDAP clients
> never get any further response from the ldap server .. until slapd is
> restarted.

This trace does not imply to me a bug in tpool.c, it looks more like some
part of the Bind operation processing failed since the SLAP_C_BINDING state
is usually reset before control returns to the connection handler loop. Try
running slapd with the debug level set higher to see what actually happened
during each client request. There isn't enough information in your log to
even guess what requests are being sent.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support