(ITS#4551) refreshOnly fails if provider is ever unreachable

[Darren.Gamble@sjrb.ca]

Full_Name: Darren Gamble
Version: 2.3.20
OS: Linux (CentOS4)
URL: 
Submission from: (NULL) (204.209.209.129)


Good day,

We are evaluating OpenLDAP for our use, and have two machines configured with
OpenLDAP 2.3.20 (CentOS4).  One machine is i386, which is a provider, and the
other is x86_64, which is a consumer.  We are testing refreshOnly as the
replication type.

We have set up both servers without much trouble, loaded data, and confirmed
that replication works as expected when the consumer comes up, and when we
change data.

But, when we simulated network problems, we were able to break replication
fairly easily.

If the consumer is ever unavailable (slapd is not running, or if the host is not
reachable), then the consumer will log the failure ("do_syncrep1:
ldap_sasl_bind_s failed (-1)") but never, ever try again.  Replication will be
broken until someone manually restarts the consumer's slapd (assuming the
provider is available at that time, of course).  If this is done, then the
consumer will resync again right away, and seems to work until the next problem
occurs.

Here's the syncrepl config section (low interval for testing purposes):

syncrepl rid=123
        provider=ldap://A.B.C.D:389
        type=refreshOnly
        interval=00:00:05:00
        searchbase="o=Shaw Cablesystems,c=CA"
        scope=sub
        schemachecking=on
        updatedn="cn=consumeradmin,o=Shaw Cablesystems,c=CA"
        bindmethod=simple
        binddn="cn=readuser,o=Shaw Cablesystems,c=CA"
        credentials=xxxxxxx

Please let us know if you need any more information.  Thanks!