If someone delete data by doing a wrong operation, and this deleting
replicated in all the replicas, I want to restore the lost data from a
backup, and I want the entries to have the same nsuniqueid than before. Of
course, I want this import to be replicated to all the replicas. If
possible, I want to be able to restore a subtree of the directory, and not
The way I handle this in a multi-master setup is:
1. Restore machine 1 from backup
2. Re-initialize machines 2-4 from machine 1
I have written a perl script (and a module) which completely handles an
MMR restore job over-the-wire. Restoring a 4-way MMR setup with 10k
entries takes less than 5 minutes.
I am the process of trying to open-source this tool (among others), but
it may take a long time or may not happen at all. You can find out how
to write it yourself by driving the operations with the admin console
and sniffing the traffic with ethereal.
Only suffixes can be restored from backup. If you have built your
subtrees as suffixes, then you can restore at the subtree level.