Adding another Master is not an option because for the end result the IP address and dns record of the master should be same. I did try to swap the hardware on the existing Master after stopping writes and making sure that the db was in sync. After recreating replication agreements with the hubs, it is now complaining that the hubs have a different generation id, hence won't replicate. 




On Dec 16, 2012, at 6:52 AM, "Dan Lavu" <dan@lavu.net> wrote:

You do not have to init the database as long as they are in sync, you can just 'send updates'. If all the slaves are 1.1.2, you'd want to eventually upgrade those machines too, have you considered just adding another master (master/master) then sliding the original master out? Should equate to zero downtime. 

On Dec 14, 2012, at 6:53 PM, Shardul Kerkar <SKerkar@accessline.com> wrote:

Hi Folks,
 
I have recently  been tasked with moving a Single Ldap Master from a dying machine to a spanking new blade. After doing some research it appears to me that the optimum way to do this will be installing a fresh instance of the application on the new server, import the database and then recreate and reinitialize all the hubs and replicas. The problem I face is that this work place has a humongous LDAP database will 3 mil+ entries. Re-initialization is taking upto 3 hours in some cases. With 5 hubs and 20 replicas to reinitialize, the downtime is unacceptable to the client.
 
If I stop writes to the Master, then export the database to the new box and recreate the New-Master-Hub replication after removing the old Master , will I still need to re-initialize the hubs? Is there any way to do this swap without reinitializing or fooling the hubs and reps into thinking that they are still talking to the same Master albeit on a new machine (same ip address/dns).
 
The client is still using ver. 1.1.2 on Centos 5.4
 
Thanks,
Shar Ker
--
389 users mailing list
389-users@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/389-users

<ATT00001..c>