Clean up a replica ring after a server delete (Linux)

Below are instructions to clean up a replica ring after a server has been forced out of the tree.  This is helpful if you unexpectedly loose a server (or eDirectory becomes corrupt and you can't restore it).

This process is also helpful if a replica add operation failed or hangs and you need to clean up the replica ring.  Sometimes it is easier to remove damaged replicas and add them back to the server rather than try and fix them.

1.  delete the objects from the tree that are specific to the server going away. (NCP server object, SAS objects for that server, LDAP objects..etc..)

Then identify each server in the tree that holds a replica in common with the server to be removed. In other words if you had a copy of [root] the the failed server then all other servers that have a copy of [root] must be manually cleaned.   Do the following in a shell as root:

2.  ndsrepair -P -Ad

3. select the Partition you want to clean the server out of (just select the number)

4.  select 10. View Replica Ring  option under the Replica Options screen

5.  select the number of the server you want to remove.

6.  select 6 from the Server Options menu "remove this server from the replica ring"

7.  authenticate, using cn=admin.o=container   (note the period an not a comma)

8.  on the next screen you must type I Agree to continue, don't type the single quotes

With these steps you have forced the current server to know that the failed server is no longer in the replica ring.  This operation does not synchronize.  (well, ok that's not entirely true.  If you do this process on the master of the replica and then as you go to the other servers to continue you may find that one or more have already removed the server from the replica ring, that's just a bonus and should not be expected)

This is just one way of cleaning up after a mess.