Solaris 10 x86 u8 openldap-2.4.23 db-4.8.30.NC dovecot-1.2.7

Can I have some clarification on how failover works when using OpenLDAP library as a client.

If a piece of software is passing multiple space-separated hosts to ldap_init() (or initialize) calls, but receives "Server is busy" error during a search request, should the LDAP library not handle that for the caller automatically/transparently? Ie, switch to next working LDAP server, and perform the ldap_search.

Or is it closer to that ldap_init will find the first working IP (the failover is early in the flow), but if an already started search later fails, it is up to the software to retry/reconnect? It only guarantees that the initial connection is valid.

What is the general practise here?

Thanks,

Lund