[Fedora-directory-users] A multi-master disaster

[Edward Capriolo]

Sorry for the funny title. I THINK* this may be a bug of multi-master
replication agreement created from the Windows Version of the Console
Tool but I was hoping someone could shed some light on this. I have
set up multi master replication before and did not run into this
issue.


I go about the normal process described in the documentation to setup
a multi-master replication agreement for only one suffix of my
directory server. dc_edops_dc_com.
Enable changelog
Select multi-master replication
Give each server a unique id

I initialize the side of the connection with data. (I also have tried
using ldif to get both sides in sync before the transfer. )

On the side I initialize from I get these messages in the error log.

[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Workers
finished; cleaning up...
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Workers cleaned up.
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Cleaning up
producer thread...
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Indexing
complete.  Post-processing...
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Flushing caches...
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Closing files...
[02/Jul/2008:18:25:19 -0400] - import dc_edops_dc_com: Import
complete.  Processed 42 entries in 2 seconds. (21.00 entries/sec)
[02/Jul/2008:18:25:19 -0400] NSMMReplicationPlugin -
multimaster_be_state_change: replica dc=edops,dc=com is coming online;
enabling replication
[02/Jul/2008:18:25:20 -0400] NSMMReplicationPlugin -
repl_set_mtn_referrals: could not set referrals for replica
dc=edops,dc=com : 32

At this point actually multi-master replication is enabled, but the B
side of the connection has some strange referral to the A side. When I
connect to side B of the replication it seems like there is one more
level there.

Does anyone understand why this simple setup would not be working?