Quanah Gibson-Mount quanah@zimbra.com писал(а) в своём письме Mon, 14 Oct 2013 19:53:07 +0400:
--On Monday, October 14, 2013 12:27 PM +0400 Igor Zinovik zinovik.igor@gmail.com wrote:
In my setup write operations are directed only to single node of a cluster. I would like to ask for advice about my deletion plan. 0. Make backup of current cn=config and data
- Disable replication via deletion olcServerID in cn=config and setting
olcMirrorMode to FALSE 2. Direct write operations to third node 3. Stop slapd instance on first node 4. Edit configuration (remove custom schema from file, renumber siblings in cn=schema,cn=config) 5. Load new configuration into /etc/openldap/slapd.d 6. Start slapd instance on first node 7. Check that schema is absent in cn=config on first node 8. Repeat steps from 3-7 on second node 9. Direct write operations to first node 10. Repeat steps from 3-7 on third node 11. Turn on replication in cluster
Since delete on schema is not even supported in the 2.4 series, you have to remove it via slapcat/slapadd anyhow.
I clearly understand that this is the only way to achive my goal.
I would note that if there are zero references to the schema, and you are not replicating cn=config, then simply stopping slapd, updating the
There is no references to schema, but I'm *replicating* cn=config between nodes. That is way I ask how to do this correctly.
LDIF and reloading, should be sufficient on each server (i.e., no need to mess with serverID or olcMirrorMode).