replication of multiple databases

[Daniel Tiefnig <openldap@qmail.infonova.at>]

Hi there,

I've found something, that might be considered a bug, or maybe just 
"user error"(due to wrong configuration).
AFAIK someone usually has to use one slurpd for each backend database to 
replicate. I'm now trying to replicate two databases, using just one 
slurpd. I'm doing thise merely by tricking slurpd into what I expect it 
to do.

slapd.conf looks like that:

database        bdb
replica         host=slave.host:port
                binddn="cn=Replicator,ou=unit1,dc=company,dc=com"
                bindmethod= ...
replogfile      /path/to/my/single/replogfile
suffix          "cn=Replicator,ou=unit1,dc=company,dc=com"
...

database        bdb
replica         host=slave.host:port
                binddn="cn=Replicator,ou=unit1,dc=company,dc=com"
                bindmethod= ...
replogfile      /path/to/my/single/replogfile
suffix          "cn=Replicator,ou=unit2,dc=company,dc=com"
...

Now, what i _expected_ was slurpd watching /path/to/my/single/replogfile 
until slapd writes something to it, then checking whether the entry 
belongs to a replica-configured database and doing the replication if 
so. What i _experienced_ now, was slurpd sending each update to one of 
the backends twice to the slave. (Not sure whether really _each_, but 
most of them at least.)

If I remove the second replica directive in slapd.conf, everything wents 
fine, and both backends are replicated correctly. (It seems.)
And that last thing is actually the part that I think shouldn't happen. 
AFAICT slurpd should ignore updates to ou=unit2, even if they are 
written to /path/to/my/single/replogfile. ("skipping foo (not mine)" or 
something like that, like it is done for replicas that are unknown by 
hostname or portnumber or so.)

Comments? (I'm also interested whether I'll shoot myself into the foot 
anyway by planning the use of such a setup for a production 
environment.)


lg,
daniel