A Forum reader recently asked:
“I would like to get some advice on adding a old server back to the replica ring. Our environment is Red Hat Linux and eDirectory 18.104.22.168. We had 2 servers in the replica ring. We had to remove one server (server 1) from the ring for couple weeks for hardware upgrading.
I removed the server 1 from the replica ring using “ndsrepair -P -Ad”. I did not remove server object and other server1 related objects (such as LDAP server and LDAP group) from the tree. Right now I have the server 1 back with old data. I do not know what will be the best way to add the server 1 back to the ring.
Before I did adding a empty server to the ring by using “ndsconfig add”. I would like to know whether the data in server 2 will get corrupted if I just add the server 1 back to the ring using ndsconfig since server 1 contains old data. Shall I delete all the objects on server 1 before I add it back to the ring? If “ndsconfig add” is not a
good option, what command shall I use?
I do notice I get a “625″ error when I run “ndsrepair -T” right now on server 1. Next time if we need to do something similar, shall I remove the server object and all the related objects when I remove the server from replica ring?”
And here is the response from Aaron Burgemeister …
You’re not talking just about removing it from the replica ring, but from the tree. If you have had your server offline for this long and you first removed its entry in the replica ring (apparently without cleaning the server itself) then you should probably complete the instance’s removal from the server and add it back in with `ndsconfig
add` but first you must destroy the instance on the box (the one removed already). This shouldn’t be too hard really.
1. Find the instances.<uid> file (probably instances.0) in /etc/opt/novell/eDirectory/conf/.edir and see where the nds.conf file is based on the contents.
2. Find that nds.conf file and then find the ‘vardir’ and ‘dibdir’ parameters. Those paths will exist.
3. Delete ‘dibdir’ and then delete ‘vardir’ (or at least rename them to something else).
4. When done, delete/rename your nds.conf file and delete the appropriate line from your instances.
Now when you run ndsmanage you should not see any instances, which is good.
5. Before adding the server back to the tree, delete all of this server’s objects from the tree (LDAP/NCP/SAS/KMOs/etc.). They should all have this server’s name in their name, so you can tell them apart from other servers that may, for some reason, be in the same context.
6. Go ahead and create a new one with ndsmanage or ‘ndsconfig add’ as you suggested.
If you are on eDir 8.8.x you should have that instances.<uid> file, regardless of the number of instances you have. Now that I look back, you have 22.214.171.124 – so yes, just rename nds.conf and the directories indicated by ‘dibdir’ and ‘vardir’ in that .conf file, and then create a new instance with ‘ndsconfig add’.