Slurpd Object Leak Patch

[Hans Oesterholt-Dijkema <hdnews@gawab.com>]

Dear All,

While our volume of objects in LDAP is increasing
we encountered problems doing replication on our AIX
platform.  In certain situations we need  to replicate
a large number of objects in a relativly short time.

Slurpd keeps taking in objects from the replication log,
until it is taking up 256MB of memory, after which
AIX kills the process. Because the replication log
keeps growing, restarting slurpd doesn't help, it just
dies again.

As we are running openldap in a production environment,
this behaviour was not acceptable for us. We cannot have
a slurpd dying, because we depend on it to replicate a
master server to several slave servers.

So we've made a patch on openldap-stable 2.2.23. Now
it is possible to configure 'replica-max-object-queue-depth' in
slapd.conf. It defaults to 2500 objects. This makes slurpd
keeping the object queue to about 2500 objects (it will still
slowly increase if an LDAP server is not available (one object
per 2 seconds)).

Is anyone interested in a 'memory stable' slurp daemon?
We would like to commit a patch to openldap to have
this feature available to the openldap community (and to
ourselves of course) in future versions of openldap.

Best regards,

Hans Oesterholt-Dijkema & Bas ten Berge
(NL).