Desiree Klecker wrote:

to update entry persistend in a master slave openldap environment the following steps are performed using Java :

1. The entry is removed.
2. The entry is created again.

This does not result in faulty behavior using a single/simple OpenLdap directory. Using a master slave configuration this results in unpredictable errors claiming that the item still exists (the remove statement has ran immediately before) if it is recreated.

I cannot see why the master slave environment should not behave in a tranparent way. It seems the remove statement was not executed in complete allthough the call is synchronous.

Which additional information are required to analyse the problem?

Do similar problems exist/ is this a known problem?

Nobody can answer any meaningful without knowing the OpenLDAP version(s) used and the actual setup.

Note that some implementors tend to ignore read-after-write issues with replication setups where write requests go to the provider and read requests are answered by the consumer.

Ciao, Michael.