I am no expert and English is not my mother tongue either, but here is what I understand:
In a multi-master configuration, the system may not be able to reconcile the databases if a partition has occurred and multiple writes/delete occurred.
"If a network is partitioned and multiple clients start writing to each of the "masters" then reconciliation will be a pain; it may be best to simply deny writes to the clients that are partitioned from the single master"
In MirrorMode, you still have two masters replicating to each other, but you use a third component -the load balancer- to redirect all requests to only one of the masters. If this master fails, then it switches to the other server. Seems to me that the masters are aware of this configuration because of the directive "mirrormode on"
In the proposed Failover Configuration, all requests are first sent to a cluster of slaves. Those slaves process all read requests and redirect the write requests to a "cluster" of masters; but in fact the load balancer is sending all write requests to a single master unless it fails, in which case it sends the request to the second master.
-Jorge
________________________________________ From: openldap-software-bounces+jmedina=e-dialog.com@OpenLDAP.org [mailto:openldap-software-bounces+jmedina=e-dialog.com@OpenLDAP.org] On Behalf Of Liutauras Adomaitis Sent: Friday, July 18, 2008 2:22 AM To: openldap-software@openldap.org Subject: explain diff between multimaster and mirror mode
Since English is not my mother tong, can someone explain me what is ment in http://www.openldap.org/doc/admin24/replication.html#Mixture%20of%20both %20Pull%20and%20Push%20based by saying "MirrorMode is not what is termed as a Multi-Master solution. This is because writes have to go to one of the mirror nodes at a time"? Is it that I have always write to the same master? Can I do writes some times to one sometimes to another, but not to both at the same time? Which way Multi-Mater or Mirror-Mode is easier to recover from broken replication. On Master-Slave I just delete Slave db and restart ldap, this is acceptible for me because there is no much data to replicate.
I'm using Master-Slave replication now, but it seems that it is not good for some reason, which I don't know and can't explain. So trying to find an alternative, but don't want to learn on my own mistakes on production :)
Liutauras